jpad journal

AND option

OR option

AMYLOID AND APOE STATUS OF SCREENED SUBJECTS IN THE ELENBECESTAT MISSIONAD PHASE 3 PROGRAM

 

C. Roberts1, J. Kaplow2, M. Giroux2, S. Krause2, M. Kanekiyo2

 

1. Eisai Ltd., Hatfield, UK; 2. Eisai Inc., Woodcliff Lake; NJ, USA

Corresponding Authors: Claire Roberts, Eisai Ltd., Hatfield, UK, Email: claire_roberts@eisai.net, Phone: +44 8456 761 590

J Prev Alz Dis 2021;
Published online January 25, 2021, http://dx.doi.org/10.14283/jpad.2021.4

 


Abstract

BACKGROUND/OBJECTIVES: Elenbecestat, an oral BACE-1 inhibitor that has been shown to reduce Aβ levels in cerebrospinal fluid, was investigated in two global phase 3 studies in early AD. Here we report on differences observed in characteristics of APOE ε4 and amyloid positive subjects in the large screening cohort.
DESIGN: Screening was performed in 5 sequential tiers over a maximum of 80 days, as part of placebo controlled, double blind phase 3 studies.
SETTING: Subjects were evaluated at sites in 7 regions (29 countries).
PARTICIPANTS: Overall, 9758 subjects were screened.
INTERVENTION: All screened subjects that were eligible received either placebo or 50 mg QID elenbecestat post randomisation.
MEASUREMENTS: Gender, disease staging, APOE ε4 status, amyloid status, amyloid positron emission tomography (PET) standard uptake value ratio (SUVr) and amyloid PET Centiloid (CL) values were determined for screened subjects; by country and region.
RESULTS: In this program, 44% of subjects were APOE ε4 positive. Frequency of females was similar in both APOE ε4 positive and negative groups. However, early mild AD subjects were slightly higher in the APOE ε4 positive group compared with the APOE ε4 negative group. 56% of subjects were amyloid positive. The mean age in the amyloid positive group was slightly higher than the amyloid negative group. The gender distribution was similar between amyloid groups. A lower number of mild cognitive impairment was observed in the amyloid positive group along with a higher number of early mild AD. APOE ε4 positive subjects were higher in amyloid positive group compared to the amyloid negative group. China had the lowest APOE ε4 and amyloid positivity rates with Western Europe and Oceania performing best. Subjects received florbetapir, florbetaben or flutemetamol amyloid PET tracer. Amyloid negative and positive subjects CL values were normally distributed around their respective means of 1.5 CL and 83 CL. However, there was an appreciable overlap in the 20-40 CL range.
CONCLUSIONS: In this large cohort of cognitively impaired subjects, subject demographics characteristics were comparable regardless of APOE genotype or amyloid positivity. APOE ε4 positivity and amyloid positivity varied by country and by geographical region.

Key words: MissionAD, Amyloid, APOE, elenbecestat.


Introduction

Alzheimer’s disease (AD) is characterised by the deposition of amyloid-beta (Aβ) aggregates and neurofibrillary tangles in the brain (1). The amyloid cascade hypothesis proposes that Aβ aggregates trigger the spreading of tau-related neurofibrillary tangles and subsequent neuronal degeneration (2). Recently, it has been shown that Aβ accumulation precedes tau and seems to accelerate neocortical tau pathology (3). In addition, amyloid accumulation has been shown to be poorly correlated with cognitive impairment (4), while tau accumulation shows a high correlation to neurodegeneration and loss of cognitive function (5). Aβ is produced when amyloid precursor protein (APP) is cleaved sequentially by β-site APP–cleaving enzyme 1 (BACE-1; also referred to as β-secretase) and γ-secretase (6). Inhibition of BACE-1 is a potential therapeutic strategy for slowing the progression of Alzheimer’s disease by reducing the production of Aβ.
Elenbecestat is an oral BACE-1 inhibitor that has been shown to reduce the Aβ level in the cerebrospinal fluid (CSF). Aβ (1-x) is a compilation of all amyloid isoforms that are cleaved by the BACE enzyme at the 1 position, that include at least the first 24 amino acids. CSF Aβ (1-x) decreased by 62% compared to placebo in healthy volunteer phase 1 study (E2609-A001-002) at 50mg; increasing to 74% at 100mg and 85% at 400mg (7). This was confirmed in patients with mild cognitive impairment (MCI) and early mild AD in an elenbecestat phase 2 study (Study E2609-G000-201)[8] indicating an average of 69% decrease in CSF Aβ(1-x) with 50mg of elenbecestat.
Elenbecestat was investigated in two global phase 3 studies (E2609-G000-301 or MissionAD1; E2609-G000-302 or MissionAD2) in early AD. The population consisted of subjects with a diagnosis of MCI due to AD and no more than 25% diagnosed as early stage mild dementia due to AD.
The studies were recommended to terminate early by the programme Data Safety Monitoring Board in September 2019, following an unfavourable benefit-risk profile of elenbecestat. At the time of termination approximately 959 subjects had reached 12 months of treatment (500 placebo, 459 elenbecestat). There was no evidence of potential efficacy and the adverse event profile was worse than placebo. At that time elenbecestat was the only BACE inhibitor programme ongoing, following similar outcomes with BACE inhibitor studies with verubecestat (9-10), lanabecestat (11), atabecestat (12-13), and umibecestat (14).
At the time of the early termination the MissionAD studies were fully recruited with 2212 randomised subjects. This has resulted in a large cohort of 9758 screened subjects, establishing eligibility through cognitive assessments, laboratory assessments, MRI safety and amyloid status; representing a large volume of information for the Alzheimer’s research community. Here we report on differences observed in characteristics of APOE ε4 and amyloid positive subjects in this large screening cohort. Preliminary data was presented at AAIC and CTAD conferences during the past two years (15-17).

 

Methods

The MissionAD screening process was performed in 5 sequential tiers over a maximum of 80 days. Cognitive assessments, medical history and clinical diagnosis were determined in tier 1. Questionnaires were administered in tier 2 to establish baseline quality of life and to assess any suicidality risk. Laboratory assessments, including APOE ε4 status, were conducted in tier 3. An MRI scan was done at tier 4 and finally an amyloid PET scan or a CSF sample was taken to determine amyloid status in tier 5.

Amyloid PET Scans

Three amyloid PET tracers were used for amyloid assessment depending on availability of the tracer manufacturing facilities: florbetapir (Amyvid™), florbetaben (Neuraceq™) or flutemetamol (Vizamyl™). Florbetaben was prioritised if more than one tracer was available at an imaging facility. Florbetaben and flutemetamol emission acquisitions require a 20-minute scan, 90-110-minute post injection. Florbetapir emission acquisition requires a 20 minute scan, 50-70 minutes post injection.
Amyloid PET status was assessed centrally by visual read, by a radiologist blinded to cognitive status. The label of each tracer defines the number of positive regions to claim positivity on a visual read. All scans were analysed by readers trained on the guidelines established by the manufacturer. Reading for visual positivity includes the following regions for analysis:
• Florbetapir (Amyvid™) Frontal cortex (excluding midline), medial frontal cortex (including Anterior cingulate), parietal cortex (excluding midline), medial parietal cortex (precuneus and/or posterior cingulate), temporal cortex, and occipital cortex.
• Florbetaben (Neuraceq™) Lateral temporal, frontal lobes, posterior cingulate/precuneus, parietal lobes.
• Flutemetamol (Vizamyl™) Frontal lobes (axial & sagittal views), posterior cingulate and precuneus (sagittal & coronal views), temporal lobes – lateral regions (axial views – coronal views as supportive), parietal lobes – lateral regions (coronal views – axial views as supportive), striatum (axial views – sagittal views as supportive).

Amyloid PET SUVr data was calculated as mean composite SUVr (a simple average of cingulate, frontal, parietal, and temporal corticies) and mean composite SUVr including occipital region (a simple average of cingulate, frontal, parietal, temporal, and occipital corticies) using whole cerebellum as reference region.
In addition, amyloid PET data was combined across the three tracers using the Centiloid methodology for the mean composite SUVr. The Centiloid project has driven the creation of a 100-point scale termed “Centiloid (CL),” which is an average value of zero in “high certainty” amyloid negative subjects and an average of 100 in “typical” AD patients (Klunk et al., 2015; http://www.gaain.org/centiloid-project).

Centiloid Conversion

The following equations were used to calculate the mean composite SUVr in Centiloid for each tracer as derived by Bioclinica (18, 19): Florbetapir: 205.72 * mean composite SUVr – 209.63; Florbetaben: 175.57 * mean composite SUVr – 173.21; Flutemetamol: 145.58 * mean composite SUVr – 139.29. All tracers: Mean composite SUVr with whole cerebellum region used as reference region is used.

CSF Amyloid Analysis

CSF sample collection was an option in this study, to determine baseline amyloid burden status. This could replace or be in addition to a baseline amyloid scan. During the study two platforms were utilized for analysis; Aβ(1-42) <250 pg/mL from Alzbio3 run at the ADNI core lab (Dr. Leslie Shaw) and the Lumipulse™ platform from Fujirebio, using a total tau:Aβ(1-42) ratio greater than 0.37 to indicate positive amyloid status. The Lumipulse CSF cutpoint was established (20); note: >0.37 ratio was set prior to the incorporation of the issued IRMM Aβ(1-42) reference standard.

APOE Status Determination

APOE genotyping was performed using a real time PCR Taqman assay, developed and performance determined by Brooks Life Sciences. The subjects were categorized into APOE ε4 positive if they had at least one ε4 allele, and negative if they did not have an ε4 allele.

Results

Overall, 9758 subjects were screened in MissionAD, of which 5710 had a known APOE status, 4121 had a known amyloid status (PET or CSF), 4077 had a known APOE and amyloid status, 3492 had a known amyloid status and available SUVr data, and 2212 subjects were randomised (Table 1). The majority of subjects, that reached tiers 3-5, had a MMSE score ≥24, CDR Global score of 0.5, and a cognitive impairment of ≥1 standard deviation from age-adjusted norms in the International Shopping List Task.

Table 1. Recruitment status

 

APOE Genotype

Of the 5,710 screened subjects with known APOE genotype, 44% were APOE ε4 positive (Table 2). The mean age in both the APOE groups was 71 years. Frequency of females was similar in both APOE ε4 positive and negative groups (51% and 52%). Early mild AD subjects were slightly higher in the APOE ε4 positive group compared with the APOE ε4 negative group (15% and 10%).

Table 2. Demography of APOE ε4 and amyloid positive and negative screened subjects

*The percentages are based on the subjects with ApoE4 status. **The percentage are based on the subjects with amyloid status (PET or CSF). Other percentages are based on each group.

 

Amyloid Status

Of the 4,077 screened subjects with known APOE and amyloid status (as determined either by PET or CSF), 56% were amyloid positive (Table 2). The mean age in the amyloid positive group was slightly higher (72 years c.f. 69 years). The gender distribution was similar between amyloid groups. A lower number of MCI was observed in the amyloid positive group (84% and 90%) along with a higher number of early mild AD (16% and 8%). APOE ε4 positive subjects were higher in amyloid positive group compared to the amyloid negative group (64% and 22%).

APOE and Amyloid Status Varies Depending on Region & Country

29 countries spanning the 7 regions of North America (Canada, USA), South America (Argentina, Chile, Mexico), Western Europe & Oceania (Australia, Austria, Denmark, Finland, France, Germany, Greece, Italy, Portugal, South Africa, Spain, UK), Eastern Europe (Bulgaria, Croatia, Czech Republic, Hungary, Poland, Russia, Slovakia), China, Japan, and Other Asia (Singapore, South Korea, Taiwan) participated in the studies.
Mean APOE ε4 positivity across the countries was 48%. This was lowest in Mexico < Taiwan < Greece < Bulgaria < Slovakia, but care needs to be used where numbers are too low to interpret. The highest rate was in France > Australia > Finland > Hungary > South Africa (Table 3). Mean amyloid positivity was 64% across the countries. This was lowest in Singapore < Croatia < Poland < China < Taiwan and highest in Greece > France > Italy > Hungary > Australia (Table 3). Grouping the countries into regions reduced variability. The mean APOE ε4 positivity and amyloid positivity across the regions was 46% and 59%, respectively. China had the lowest APOE ε4 and amyloid positivity rates, while Western Europe and Oceania the highest rates (Figure 1).

Table 3. APOE and amyloid status by country

Figure 1. APOE and amyloid status by region (Dotted red lines indicate regional means)

 

Amyloid PET Status

53% of Subjects screened with a known amyloid status and amyloid PET SUVr data were amyloid positive. Overall, 1563 (45%) subjects were APOE ε4 positive, 1816 (52%) were female, and 3028 (87%) were MCI and 420 (12%) mild early AD (Table 2).

Amyloid PET SUVr

In total, 386, 2548 and 558 subjects received florbetapir, florbetaben or flutemetamol amyloid PET tracer, of which 218 (56%), 1292 (51%) and 330 (59%) were determined visually to be amyloid positive, respectively. Amyloid PET mean composite (SD; min – max) SUVr for amyloid negative subjects were 1.03 (0.11; 0.80 – 1.48), 0.99 (0.08; 0.70 – 1.45) and 0.98 (0.11; 0.80 – 1.38) for florbetapir, florbetaben or flutemetamol, respectively. Amyloid PET mean composite (SD; min – max) SUVr for amyloid positive subjects were 1.41 (0.18; 0.94 – 1.94), 1.47 (0.22; 0.80 – 2.22) and 1.49 (0.19; 0.95 – 1.99) for florbetapir, florbetaben or flutemetamol, respectively. Including the occipital region in the mean composite SUVR had very little impact on the data.

Amyloid PET Conversion to Centiloid Mean Composite

Centiloid (CL) conversion, combining data for the 3 tracers, calculated a mean (SD) of 1.5 (15.2) CL for amyloid negative subjects (n=1652) and 82.7 (36.7) CL for amyloid positive (n=1840). No differences were observed in the individual CL values between the tracers.

Mean Composite Centiloid Distribution

Amyloid negative and positive subjects CL values were normally distributed around their respective means of 1.5 CL and 83 CL. However, there was an appreciable overlap in the 20-40 CL range; minimum and maximum values of -50 and 95 CL for amyloid negative and 33 and 217 CL for amyloid positive subjects (Figure2).
There was no obvious impact of gender on the distribution of CL, with mean (SD) composite of 85 (35.6) CL and 81 (37.8) CL for amyloid positive, and 3.2 (16.1) CL and 0 (14.0) CL for amyloid negative, for females and males respectively.
More amyloid positive subjects were APOE ε4 positive (65%) compared to APOE ε4 negative (35%), with the mean 10 CL higher in APOE ε4 positive (86 CL, SD 34.2) than APOE ε4 negative (76 CL, SD 40.4) in the amyloid positive population. A majority (77%) of the APOE ε4 positive subjects were also amyloid positive.
Of the 420 mild early AD subjects 69% were amyloid positive, while in the 3028 MCI subjects 51% were amyloid positive. There was a shift in the mean by approximately 9 CL in the amyloid positive cohort, 81 CL (SD 36.6) for MCI subjects and 90 CL (SD 36.8) for mild early AD subjects.

Figure 2. Distribution of mean composite Centiloid for amyloid negative and positive subjects

 

Discussion

Alzheimer’s Disease clinical trials, due to the complexity required to identify the correct population, include many screening assessments that encompass not only traditional elements but also multiple cognitive, biomarker and imaging assessments. Especially in large global studies, there is an expectation of a high screen failure rate, while tiered screening is often required in order to reduce unnecessary burden on subjects and control costs.
For MissionAD the screening approach was to conduct the cognitive assessments and study specific clinical diagnosis in the first tier to eliminate the highest proportion of subjects not suitable for the trial. Approximately 41% of subjects screen failed due to cognitive or suicidality assessments. About 17% further subjects failed due to lab assessments. Finally, in the last tier a further 19% of subjects screen failed when amyloid status was determined to be negative. Overall, the screen fail rate in the early AD population (MCI and early mild AD) of MissionAD was 77%. This screen failure rate is in the same range as reported in studies with similar populations. For example, the verubecestat APECS study in prodromal AD had a SF rate of 68% (10), the AMARANTH and DAYBREAK studies with lanabecestat in early and mild AD had SF rates of 68% and 70%, respectively (11). The slightly higher rate in MissionAD may be explained by the focus on recruiting a higher proportion of MCI subjects (>75%) in this program.
When assessed at the end of the screening cascade, amyloid positivity (amyloid PET visual read or amyloid CSF) was a global average of 56%. Of note, the positivity rates varied from country to country. It is important to understand where rates may be low, so that recruitment enrichment strategies can be employed in those areas. Countries where the positivity rates were low may be reflective of populations that more readily report subjective cognitive decline or the “worried well”, lack of access to technology, or lack of access to care. Socioeconomic status and different healthcare systems may result in a greater focus on subjective cognitive symptoms. Lowest rates of positivity in MissionAD were observed in North America and China. APOE ε4 positivity also varied by country and was lowest in North America and China. Summation of these components supported amyloid positivity as highly correlated with APOE ε4 positivity.
These data reinforce the need to be aware of the variability in APOE ε4 and amyloid positivity when designing clinical trials, and the importance of having strategies available to mitigate. Various approaches were employed in this trial to maximise eligible subjects, however, the screen failure rate for MissionAD was still 77% (2212 randomised out of the 9758 screened subjects).
This study allowed the use of three different amyloid PET tracers with florbetaben being used in 75% of the subjects in this study. We presented the SUVr data as a mean composite with and without the occipital region for all tracers. However, it should be noted that per the manufacturers label the occipital region is appropriate only for florbetaben. Data indicate that the tracers were comparable and could be grouped for comparisons by using the Centiloid methodology supporting the use of multiple tracers in a global trial to expand countries able to participate.
Amyloid negative and positive subjects Centiloid values were normally distributed with an appreciable overlap in the 20-40 CL range. Distribution was shifted to higher Centiloid values by APOE ε4 positive status and mild early AD diagnosis. Similar results have been shown by other groups when examining large cohort studies (21) and have revealed similar, bimodal distributions of CL in the population studied (22). These results reveal similar factors influenced amyloid positivity as was evident with amyloid PET visual read.
In this large cohort of cognitively impaired subjects, just under half were positive for APOE ε4 and just over half were amyloid positive. APOE ε4 positivity and amyloid positivity varied by country and by geographical region. Subject demographics characteristics were comparable regardless of APOE genotype or amyloid positivity. A higher APOE ε4 or amyloid positivity rate generally reflected a higher identification of early AD subjects in most regions. APOE ε4 positive subjects are more likely to have elevated amyloid.

 

Acknowledgments: Study participants and all sites that took part in MissionAD. Editorial support provided by JD Cox, PhD and Mayville Medical Communications.

Funding: Funding for the studies, analyses, and editorial support was provided by Eisai Inc.

Disclosures: All authors are, or were, employees of Eisai Inc or Eisai Ltd.

Ethical standards: The study was performed in full compliance with International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use and all applicable local Good Clinical Practice and regulations.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

References

1. Masters CL, Bateman R, Blennow K, Rowe CC, Sperling RA, Cummings JL. Alzheimer’s disease. Nat Rev Dis Primers 2015;1:15056.
2. Selkoe DJ, Hardy J. The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO Mol Med 2016;8:595–608.
3. Hanseeuw BJ, Betensky RA, Jacobs HI, et al. Association of Amyloid and Tau With Cognition in Preclinical Alzheimer Disease: A Longitudinal Study. JAMA Neurol. 2019;76(8):915-924.
4. Jack CR, Knopman DS, Jagust WJ, et al. Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol. 2010;9(1):119-28.
5. Duyckaerts C, Brion JP, Hauw JJ, Flament-Durand J. Quantitative assessment of the density of neurofibrillary tangles and senile plaques in senile dementia of the Alzheimer type. Comparison of immunocytochemistry with a specific antibody and Bodian’s protargol method. Acta Neuropathol. 1987;73(2) :167-170.
6. Yan R, Vassar R. Targeting the β secretase BACE1 for Alzheimer’s disease therapy. Lancet Neurolog. 2014 ;13:319–29.
7. Albala B, Kaplow J, Lai R, Matijevic M, Aluri J, Satlin A. CSF Amyloid Lowering in Human Volunteers after 14 days Oral Administration of the Novel BACE1 Inhibitor E2609. 2012. AAIC abstract published July 2012 (oral presentation).
8. Majid O, Gee M, Albala B, et al. Dose-related reductions of CSF amyloid β (1-x) by E2609, a novel BACE1 inhibitor, in patients with mild cognitive impairment due to Alzheimer’s disease (AD) and mild-to-moderate AD dementia. CTAD 2016;San Diego: P3-28 (poster presentation).
9. Egan MF, Kost J, Tairot PN et al. Randomised trial of verubecestat for mild-to-moderate Alzheimer’s disease. N. Eng. J. Med. 2018;378(18): 1691-1703.
10. Egan MF, Kost J, Voss T, et al. Randomized Trial of Verubecestat for Prodromal Alzheimer’s Disease. N Engl J Med. 2019;380(15):1408-1420.
11. Wessels AM, Tariot PN, Zimmer JA, et al. Efficacy and Safety of Lanabecestat for Treatment of Early and Mild Alzheimer Disease: The AMARANTH and DAYBREAK-ALZ Randomized Clinical Trials. JAMA Neurol. 2020 Feb 1;77(2):199-209.
12. Henley D, Raghavan N, Sperling R, Aisen P, Raman R, Romano G. Preliminary Results of a Trial of Atabecestat in Preclinical Alzheimer’s Disease. N Engl J Med. 2019;380(15):1483-1485.
13. Novak G, Streffer JR, Timmers M, et al. Long-term safety and tolerability of atabecestat (JNJ-54861911), an oral BACE1 inhibitor, in early Alzheimer’s disease spectrum patients: a randomized, double-blind, placebo-controlled study and a two-period extension study. Alzheimers Res Ther. 2020;12(1):58.
14. Graf A, Borowsky B, Tariot P, et al. Alzheimer’s Prevention Initiative Generation Program: Update And Next Steps. 2019. CTAD abstract [OC11].
15. Roberts C, Kaplow J, Giroux M, Krause S, Kanekiyo M. Amyloid burden assessed by three amyloid PET tracers in the elenbecestat MissionAD phase 3 program. 2020. AAIC abstract published, July 2020 (poster presentation).
16. Roberts C, Kanekiyo M, Albala B. Amyloid positive subject characteristics in the elenbecestat MissionAD phase 3 program. 2019. CTAD abstract published, December 2019 (poster presentation).
17. Roberts C, Kanekiyo M, Kaplow J, Albala B. APOE4 status and amyloid burden differences across regions in the elenbecestat MissionAD phase 3 program. 2019. AAIC abstract published, July 2019 (poster presentation).
18. Klunk WE, Koeppe RA, Price JC, et al. The Centiloid Project: standardizing quantitative amyloid plaque estimation by PET. Alzheimers Dementia. 2015;11(1):1-15.
19. Adamczuk K, Sampat M, Bracoud L, Runkle M, Gorman B, Scott D. Centiloid scale in practice: Effects of different SUVr reference regions and comparison of centiloid cut-offs. Alzheimer’s & Dementia. 2019;15(7) suppl: 223-224.
20. Kaplow J, Vandijck M, Gray J, et al. Concordance of Lumipulse cerebrospinalfluid t-tau/Aβ42 ratio with amyloid PET status. Alzheimer’sDement. 2020;16:144–152.
21. Mormino E, Betensky RA, Hedden T, et al. Amyloid and APOE ε 4 interact to influence short-term decline in preclinical Alzheimer’s disease. Neurology. 2014;82:1760-1867.
22. La Joie R, Ayakta N, Seeley WW, Borys E et al. Multisite study of the relationships between antemortem [11C]PiB-PET Centiloid values and post-mortem measures of Alzheimer’s disease neuropathology. Alzheimer’s & Dementia. 2018; 1-12.

OVERTREATING ALZHEIMER’S DISEASE

 
M. Canevelli1, N. Vanacore2, A. Blasimme3, G. Bruno1, M. Cesari4
 

1. Sapienza University, Rome, Italy; 2. National Center for Disease Prevention and Health Promotion, National Institute of Health, Rome, Italy; 3. Department of Health Sciences and Technology, ETH Zürich, Switzerland; 4. IRCCS Istituti Clinici Maugeri, University of Milan, Italy.

Corresponding Author: Matteo Cesari, MD, PhD, Geriatric Unit, IRCCS Istituti Clinici Scientifici Maugeri, Via Camaldoli 64 – 20138 Milan – Italy, Phone: +39 02 50725136, Twitter: @macesari, email: macesari@gmail.com

J Prev Alz Dis 2020;
Published online December 31, 2020, http://dx.doi.org/10.14283/jpad.2020.74

 


Abstract

The management of frailty in older persons is not easy, implying interventions beyond the simple prescription of medications. Biological complexity, multimorbidity, polypharmacy, and social issues often hamper the possibility to directly translate the evidence coming from research into clinical practice. Frailty indeed represents the most relevant cause of the “evidence-based medicine issue” influencing clinical decisions in geriatric care. Today, patients with Alzheimer’s disease (AD) are much older and frailer than some decades ago. They also tend to have more drugs prescribed. In parallel, research on AD has evolved over the years, hypothesizing that anticipating the interventions to the earliest stages of the disease may provide beneficial effects (to date, still lacking). In this article, we argue that, by focusing exclusively on “the disease” and pushing to anticipate its detection (sometimes even before the appareance of its clinical manifestations) may overshadow the person’s values and priorities. Research should be developed for better integrating the concept of aging and frailty in the design of clinical trials in order to provide results that can be implemented in real life. On the other hand, clinicians should be less prone to the easy (but unsupported by evidence) pharmacological prescription.

Key words: Overtreatment, dementia, cognition, geriatrics, prevention.


 
The management of frailty in older persons is not easy, mainly because it requires interventions beyond the simple prescription of medications (1–5). Biological complexity, multimorbidity, polypharmacy, and social issues often hamper the possibility to translate the evidence coming from research into clinical practice directly. Frailty indeed represents the most relevant cause of the “evidence-based medicine issue”, which has traditionally complicated geriatric medicine (3, 6, 7). Moreover, the outcomes often used in clinical trials conducted in adult and/or non-frail individuals are often meaningless in older persons with frailty (8, 9). What is more, the priorities of older persons with chronic conditions are usually found on aspects that are not necessarily related to the cure of a given disease or the life extension (10, 11). Instead, considerations about the quality of life play a critical role that is routinely neglected in medical decision-making (8). Lack of regard for patient-centered care as well as the absence of evidence specifically supporting the clinical management of frail older persons often lead to the need of making decisions with the addition of a good dose of common sense.
At the same time, the increasing diagnostic capacity and the hyper-specialization of medicine (arguably a correlate of a culture that privileges functional integrity over life quality and emotional fulfillment) (12) have substantially contributed to the widespread phenomena of overdiagnosis and overtreatment (13). In particular, these problems of modern medicine stem from the incapacity to deal with the individual’s multidimensional complexity and the lack of integrated models of care. The former, caused by the heterogeneity of the underlying aging process (14, 15), is responsible for the “one-size-fits-all” paradigm pervasively promoted in medicine. The latter causes an approach to the patient that proceeds by siloes, failing to 1) realize the complexity of the system (16, 17), and 2) adequately integrate available clinical information with the person’s life plans, expectations, and preferences (3). After all, it is well-established that, while the robustness of scientific evidence tends to fade in the oldest and frailest individuals, these are paradoxically the ones who are the most exposed to polypharmacy (18). In this scenario, we cannot ignore the detrimental aspect of the defensive medicine19. Clinicians today identify the prescription of tests and medications as the final goal of their mission, and patients are not aware that sometimes “less is more” (20).
The field of Alzheimer’s disease (AD) has substantially been exposed to these risks. AD is a condition of old age. Today, AD patients are much older and frailer than some decades ago (21). They also tend to have more drugs prescribed. The complexity of the condition is also exemplified by the substantial role played by social factors (e.g., literacy) in its clinical expression (22).
Research on AD has been evolving over the years, hypothesizing that the anticipation of the interventions to the earliest stages of the disease may provide those beneficial effects that, to date, are still lacking. For this reason, a myriad of preclinical conditions has been developed. Constructs as mild cognitive impairment (23), prodromal AD (24), subjective cognitive decline (25)… have been designed for research purposes, in particular for defining conditions to target with experimental drugs. Although sometimes explained the rationale behind the design of these pre-AD conditions, their adoption in the clinical setting has always been quite rapid – a phenomenon known as “diagnostic creep” (26). This leads to an earlier diagnosis of AD (and its different forms), generating a more extended life “with the disease”, more heterogeneity in the biological/clinical profile, more patients to treat, and higher costs for the healthcare systems. By focusing exclusively on “the disease”, the push to anticipate its detection even before its preclinical manifestations appear (27) overshadows the person’s values and priorities.
Furthermore, it also seems that the risk of preciously treating individuals presenting early signs/symptoms of AD is never considered for its potential of reducing the effect size of the intervention. In other words, there is the possibility that an effective molecule might appear less effective than it is simply because administered to a population including false positive (i.e., non-diseased) individuals. Moreover, anticipating the diagnosing of a condition that, to date, has still no disease-modifier treatment is fraught with a host of not yet settled ethical issues (28–30). In particular, does an anticipated AD diagnosis really expand individual autonomy and self-determination (as many seem to argue), or is it instead conducive to self-depreciation, stigmatization, and further social isolation?
Moreover, let us consider some of the implications of treating AD preclinical phases as actual clinical entities. Recently, Egan et al. (31) published the results from a randomized, double-blind, placebo-controlled trial aimed at evaluating the effect of verubecestat, a beta-site amyloid precursor protein-cleaving enzyme-1 (BACE-1) inhibitor, in prodromal AD. Beyond the negative effect of the drug on cognitive function, what stood out from the report was the theoretically unexpected, high proportion of participants (i.e., 46.0 %) taking cholinesterase inhibitors or memantine at the study baseline. Such a high prevalence in the treatment of prodromal AD is unjustified and quite worrisome, given the likely inclusion of many potentially reversible cases among the participants (32). It is furthermore not negligible that this overtreatment occurs with medications that:
1) have proven to be ineffective at delaying/halting the progression to overt dementia (33);
2) may expose the individual to adverse drug reactions and even worsen the person’s functioning (34); and
3) are not approved for such use by international regulatory agencies.
The absence of alternative therapeutic options cannot justify these interventions, which should be considered as ethically, clinically, and economically inappropriate/unsustainable. Research should thus be developed for better integrating the concept of aging and frailty in the design of clinical trials in order to provide results that can be implemented in real life (8, 35–38). On the other hand, clinicians should be less prone to the easy (but unsupported by evidence) pharmacological prescription.

 

Conflicts of interest: Marco Canevelli is supported by a research grant of the Italian Ministry of Health (GR-2016-02364975) for the project “Dementia in immigrants and ethnic minorities living in Italy: clinical-epidemiological aspects and public health perspectives” (ImmiDem). Matteo Cesari has received honoraria for presentations at scientific meetings and/or research funding from Nestlé. No specific conflict of interest declared by the other authors.

 

References

1. de Kerimel J, Tavassoli N, Lafont C, et al. How to Manage Frail Older Adults in the Community? Proposal of a Health Promotion Program Experienced in a City of 16,638 Inhabitants in France. J Frailty Aging. 2018;7(2):120-126. doi:10.14283/jfa.2017.47
2. Sinclair AJ, Abdelhafiz A, Dunning T, et al. An International Position Statement on the Management of Frailty in Diabetes Mellitus: Summary of Recommendations 2017. J Frailty Aging. 2018;7(1):10-20. doi:10.14283/jfa.2017.39
3. Cesari M, Marzetti E, Thiem U, et al. The geriatric management of frailty as paradigm of “The end of the disease era”. Eur J Intern Med. 2016;31:11-14. doi:10.1016/j.ejim.2016.03.005
4. Negm AM, Kennedy CC, Thabane L, et al. Management of Frailty: A Systematic Review and Network Meta-analysis of Randomized Controlled Trials. J Am Med Dir Assoc. 2019;20(10):1190-1198. doi:10.1016/j.jamda.2019.08.009
5. Muscedere J. The Need to Implement Frailty in the International Classification of Disease (ICD). J Frailty Aging. 2020;9(1):2-3. doi:10.14283/jfa.2020.2
6. Balducci L. Evidence-based or evidence-biased medicine? J Med Pers. 2013;11:99-100.
7. Evans J. Evidence-based and evidence-biased medicine. Age Ageing. 1995;24(6):461-463.
8. Canevelli M, Remoli G, Toccaceli Blasi M, et al. Ongoing Research Protocols for the Pharmacological Treatment of Neuropsychiatric Symptoms in Dementia. J Frailty Aging. Published online July 7, 2020. doi:10.14283/jfa.2020.36
9. Guralnik J, Bandeen-Roche K, Bhasin S a. R, et al. Clinically Meaningful Change for Physical Performance: Perspectives of the ICFSR Task Force. J Frailty Aging. 2020;9(1):9-13. doi:10.14283/jfa.2019.33
10. Fried TR, Street RL, Cohen AB. Chronic Disease Decision Making and “What Matters Most.” J Am Geriatr Soc. 2020;68(3):474-477. doi:10.1111/jgs.16371
11. Renne I, Gobbens RJ. Effects of frailty and chronic diseases on quality of life in Dutch community-dwelling older adults: a cross-sectional study. Clin Interv Aging. 2018;13:325-334. doi:10.2147/CIA.S156116
12. Cesari M, Pérez-Zepeda M, Marzetti E. Frailty and multimorbidity: different ways of thinking about geriatrics. J Am Med Dir Assoc. 2017;18(4):361-364. doi:10.1016/j.jamda.2016.12.086
13. Canevelli M, Bruno G, Remiddi F, et al. Spontaneous Reversion of Clinical Conditions Measuring the Risk Profile of the Individual: From Frailty to Mild Cognitive Impairment. Front Med (Lausanne). 2017;4:184. doi:10.3389/fmed.2017.00184
14. Cesari M. Frailty and Aging. J Frailty Aging. 2012;1:3-6.
15. Cesari M, Vellas B, Gambassi G. The stress of aging. Exp Gerontol. 2013;48(4):451-456. doi:10.1016/j.exger.2012.10.004
16. Kiiti Borges M, Oiring De Castro Cezar N, Silva Santos Siqueira A, Yassuda M, Cesari M, Aprahamian I. The relationship between physical frailty and mild cognitive impairment in the elderly: a systematic review. J Frailty Aging. 2019;8(4):192-197. doi:10.14283/JFA.2019.29
17. Maltais M, de Souto Barreto P, Rolland Y, Vellas B. The Association of ApoE ε4 Status with Lower Limb Function and Handgrip Strength in Older Adults. J Frailty Aging. 2019;8(2):62-66. doi:10.14283/jfa.2019.7
18. Muth C, Blom JW, Smith SM, et al. Evidence supporting the best clinical management of patients with multimorbidity and polypharmacy: a systematic guideline review and expert consensus. J Intern Med. 2019;285(3):272-288. doi:10.1111/joim.12842
19. Snyder Sulmasy L, Weinberger SE. Better care is the best defense: high-value clinical practice vs defensive medicine. Cleve Clin J Med. 2014;81(8):464-467. doi:10.3949/ccjm.81a.14037
20. Dam A, Datta N, Mohanty U, Khanna P. Pharmaceutically Less and Holistically More. JAMA Intern Med. 2013;173(11):948. doi:10.1001/jamainternmed.2013.475
21. Canevelli M, Cesari M, Trobia F, et al. Sociodemographic and Clinical Changes Over Time of Individuals Evaluated for Cognitive Disturbances: Good or Bad News. J Am Med Dir Assoc. 2015;16(12):1095-1096. doi:10.1016/j.jamda.2015.09.013
22. Arce Rentería M, Vonk JMJ, Felix G, et al. Illiteracy, dementia risk, and cognitive trajectories among older adults with low education. Neurology. 2019;93(24):e2247-e2256. doi:10.1212/WNL.0000000000008587
23. Petersen R. Clinical trials for early (pre-dementia) Alzheimer’s disease: a case for mild cognitive impairment. J Nutr Health Aging. 2010;14(4):304-305.
24. Dubois B, Feldman HH, Jacova C, et al. Advancing research diagnostic criteria for Alzheimer’s disease: the IWG-2 criteria. Lancet Neurol. 2014;13(6):614-629. doi:10.1016/S1474-4422(14)70090-0
25. Jessen F, Amariglio R, van Boxtel M, et al. A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease. Alzheimers Dement. 2014;10(6):844-852. doi:10.1016/j.jalz.2014.01.001
26. McGregor MJ, Martin D. Testing 1, 2, 3: is overtesting undermining patient and system health? Can Fam Physician. 2012;58(11):1191-1193, e615-617.
27. Yassine HN. Targeting prodromal Alzheimer’s disease: too late for prevention? Lancet Neurol. 2017;16(12):946-947. doi:10.1016/S1474-4422(17)30372-1
28. Mattsson N, Brax D, Zetterberg H. To know or not to know: ethical issues related to early diagnosis of Alzheimer’s disease. Int J Alzheimers Dis. 2010;2010. doi:10.4061/2010/841941
29. Howe EG. Ethical Issues in Diagnosing and Treating Alzheimer Disease. Psychiatry (Edgmont). 2006;3(5):43-53.
30. Galvin JE. Screening for Mild Cognitive Impairment: There is the Will but Is There a Way? J Prev Alzheimers Dis. 2020;7(3):144-145. doi:10.14283/jpad.2020.16
31. Egan MF, Kost J, Voss T, et al. Randomized Trial of Verubecestat for Prodromal Alzheimer’s Disease. New England Journal of Medicine. 2019;380(15):1408-1420. doi:10.1056/NEJMoa1812840
32. Canevelli M, Grande G, Lacorte E, et al. Spontaneous Reversion of Mild Cognitive Impairment to Normal Cognition: A Systematic Review of Literature and Meta-Analysis. J Am Med Dir Assoc. Published online August 5, 2016. doi:10.1016/j.jamda.2016.06.020
33. Raschetti R, Albanese E, Vanacore N, Maggini M. Cholinesterase inhibitors in mild cognitive impairment: a systematic review of randomised trials. PLoS Med. 2007;4(11):e338. doi:10.1371/journal.pmed.0040338
34. Schneider LS, Insel PS, Weiner MW. Treatment With Cholinesterase Inhibitors and Memantine of Patients in the Alzheimer’s Disease Neuroimaging Initiative. Arch Neurol. 2011;68(1):58-66. doi:10.1001/archneurol.2010.343
35. Archibald M, Kitson A, Frewin D, Visvanathan R. Transdisciplinary Research In Frailty: Knowledge Translation To Inform New Models of Care. J Frailty Aging. 2017;6(2):62-64. doi:10.14283/jfa.2017.6
36. Rowe J, Fried L. Incorporating frailty into clinical practice and clinical research. J Frailty Aging. 2013;2(3):126-127. doi:10.14283/jfa.2013.17
37. Beard JR. Linking Geroscience and Integrated Care to Reinforce Prevention. J Prev Alzheimers Dis. 2020;7(2):68-69. doi:10.14283/jpad.2020.13
38. Takeda C, Guyonnet S, Sumi Y, Vellas B, Araujo de Carvalho I. Integrated Care for Older People and the Implementation in the INSPIRE Care Cohort. J Prev Alzheimers Dis. 2020;7(2):70-74. doi:10.14283/jpad.2020.8

 

PSYCHOMETRIC PROPERTIES OF THE CLINICAL DEMENTIA RATING – SUM OF BOXES AND OTHER COGNITIVE AND FUNCTIONAL OUTCOMES IN A PRODROMAL ALZHEIMER’S DISEASE POPULATION

 

F. McDougall1, C. Edgar2, M. Mertes3, P.Delmar3, P. Fontoura3, D. Abi-Saab3, C.J. Lansdall3, M. Boada4,5, R. Doody1,3

 

1. Genentech, South San Francisco, USA; 2. Cogstate Ltd, London, UK; 3. F. Hoffmann-La Roche Ltd, Basel, Switzerland; 4. Research Center and Memory Clinic, Fundació ACE, Institut Català de Neurociències Aplicades, Universitat Internacional de Catalunya, Barcelona, Spain; 5. Networking Research Center on Neurodegenerative Diseases (CIBERNED), Instituto de Salud Carlos III, Madrid, Spain.

Corresponding Author: Fiona McDougall, Genentech 620 E Grand Ave, South San Francisco, CA 94080, USA, mcdougall.fiona@gene.com

J Prev Alz Dis 2020;
Published online December 21, 2020, http://dx.doi.org/10.14283/jpad.2020.73

 


Abstract

BACKGROUND: The Clinical Dementia Rating–Sum of Boxes (CDR-SB) has been proposed as a primary outcome for use in prodromal AD trials. However, the psychometric properties of this, and of other commonly used measures, have not been well-established in this patient population.
OBJECTIVE: To describe the psychometric properties of commonly used efficacy measures in a clinical trial of prodromal AD.
SETTING: Data were gathered as part of a two-year clinical trial.
PARTICIPANTS: Patients had biomarker confirmed prodromal AD.
MEASUREMENTS: Clinical Dementia Rating (CDR), Functional Activities Questionnaire (FAQ), Alzheimer’s Disease Assessment Scale – Cognition Subscale 11 and 13 (ADAS-Cog), Mini Mental State Exam (MMSE), and Free and Cued Selective Reminding Test (FCSRT-IR [words]). Assessments were conducted at least every 24 weeks.
RESULTS: For the CDR-SB, test-retest reliability was good (intra-class correlation coefficient [ICC]=0.83); internal consistency was 0.65 at baseline but above 0.8 at later assessments. Relationships between the CDR-SB and other measures were as expected (higher correlations with more closely related constructs), and the CDR-SB differentiated between patients with different severities of dementia (-2.9 points difference between CDR-Global Score 0.5 and 1, P<.0001). Floor and ceiling effects on the CDR-SB total score were minimal; however, at baseline there were ceiling effects in the personal care domain. Further detail is provided on the psychometric properties of ADAS-Cog, MMSE, FCSRT-IR and FAQ in this population.
CONCLUSION: The psychometric properties of the CDR-SB are adequate in prodromal AD and continued use is warranted in clinical trials. However, there remains scope for improvement in the assessment of functional constructs and development of novel measures should continue.

Key words: Clinical dementia rating, prodromal Alzheimer’s disease, psychometric testing.


 

Introduction

A number of clinical trials of potential disease-modifying treatments in Alzheimer’s Disease (AD) are now targeted at early stage disease, where it is thought that there will be the greatest benefit to patients. By targeting the disease at the prodromal and mild dementia stages (also referred to in this paper as “early AD”), it is hoped to slow progression before extensive, irreversible neurodegeneration occurs. Since AD may be viewed as a continuum with preclinical, prodromal and dementia stages (mild, moderate and severe), the dementia diagnosis itself may be an important milestone in progression, but not one that represents a natural or stark differentiating boundary in terms of underlying pathophysiology. Diagnostic criteria for prodromal AD (pAD) (1) and mild cognitive impairment (MCI) due to AD (2) are now established; however, existing outcome measures to assess efficacy were mainly developed and validated for overt dementia and so may be unsuitable for clinical trials in this earlier patient population.
The FDA and EMA have both called for novel approaches to assess efficacy, recognizing the limitations of existing instruments in the earliest stages of AD. The FDA guidance and EMA guidelines (3, 4) have stated that clinical trials in the dementia stage of AD should use a co-primary approach, in which a treatment should demonstrate efficacy on both a cognitive measure and a functional measure (3, 5). This has been described as intending to ensure “that a clinically meaningful effect was established by a demonstration of benefit on the functional measure and that the observed functional benefit was accompanied by an effect on the core symptoms of the disease as measured by the cognitive assessment” (3). However, in the early stages of AD (stages 3 and 4), spanning pAD and mild AD dementia (mAD), it is recognized in both the FDA draft guidance (3) and the EMA guideline (6) that measurement may be more challenging. As independent research has shown, current assessment tools may have limited sensitivity due to ceiling effects and slow rates of progression (7, 8). Co-primary outcomes are not well established at this early stage, and whilst the principle behind the co-primary approach still holds, it has been suggested by regulators and others that application in practice could be achieved by integrated cognitive and functional endpoints, such as the Clinical Dementia Rating – Sum of Boxes (CDR-SB) score (6, 9-11). The CDR is intended to measure “the influence of cognitive loss on the ability to conduct everyday activities” (12). Whilst it has been hypothesized that the CDR may be broken down into ‘cognitive’ and ‘functional’ items (10, 13), the original intent was a unitary underlying construct (14) . Thus, it may be that observed statistical relationships supporting separate cognitive and functional items result from other properties, such as disease severity, or the use of information from the patient versus that from the caregiver-informant.
Studies have consistently reported high internal consistency for the CDR-SB across the AD spectrum, including clinically defined prodromal populations (CDR-GS = 0.5) (10, 15). Inter-rater reliability of the CDR-SB in a prodromal population is unclear, with some studies reporting low inter-rater agreement in populations with earlier non-biomarker confirmed AD dementia (13, 14, 16). Although many clinical trials in early/prodromal AD have used the CDR-SB as a primary endpoint, including studies of crenezumab (NCT02670083, NCT03114657), gantenerumab (NCT03443973, NCT03444870), aducanumab (NCT02484547, NCT02477800), BAN2401 (NCT03887455), and verubecestat (NCT01953601), a comprehensive assessment of the psychometric properties of the CDR-SB in this population is lacking.
To our knowledge, there are no studies describing the test-retest reliability according to gold-standard intra-class correlation coefficient (ICC) for a biomarker-confirmed prodromal population; a critical gap in the evidence needed to support use of the CDR-SB as a primary endpoint in AD clinical trials.
Investigation of the psychometric properties of commonly used outcome measures in the pAD clinical trials population is a critical step in confirming that assessments are fit for purpose, and for identifying potential gaps/areas for further development. Here, we describe traditional psychometrics, including test-retest reliability, of cognitive and functional assessments in a pAD trial population from SCarlet RoAD (NCT01224106; WN25203), a Phase 3, multicenter, randomized, double-blind, placebo-controlled study. In addition to amnestic MCI, subjects recruited to SCarlet RoAD were required to have evidence of amyloid pathology as demonstrated by low levels of Aβ(1–42) in cerebrospinal fluid (CSF). We also explore suitability of the CDR-SB as a single primary endpoint. Such properties should be established for the planned context of use (17, 18), and this paper is intended to do so for multinational, pAD, clinical trials. Furthermore, estimates are dataset dependent and a range of published estimates across different contexts of use may be informative. There are three main points of distinction from prior studies on this topic; some studies have used data from a single country only (15), while others have used observational cohorts (10) and defined AD based on clinical rather than biomarker criteria (10). This paper therefore adds to the existing literature by providing a comprehensive evaluation of the psychometric properties of key clinical outcome assessments in a multinational, clinical trial, biomarker confirmed prodromal AD population. The analysis presented is based on data from the SCarlet RoAD trial that evaluated low dose gantenerumab in patients with prodromal AD.

 

Methods

Data source and patients

The data were gathered as part of a Phase III, multicenter, randomized, double-blind, placebo-controlled, parallel-group, two-year study to evaluate the effect of subcutaneous gantenerumab (RO4909832) on cognition and function in prodromal Alzheimer’s disease (pAD) conducted across 24 countries.
The primary objective of the trial was to evaluate the effect of gantenerumab versus placebo on the change from baseline to week 104 in the Clinical Dementia Rating scale Sum of Boxes (CDR-SB). All measures were translated and linguistically validated as per industry guidelines (19). CDR raters received comprehensive training prior to study start and refresher trainings at regular interval during the study. The information to make each rating was obtained through a semi-structured interview of the subject and a reliable informant. The study required that each subject have a study partner who, in the investigator’s judgment, had frequent and sufficient contact with the subject so as to be able to provide accurate information regarding the subject’s cognitive and functional abilities and who agreed to accompany the subject to clinic visits for scale completion. As far as possible, raters and study partners remained unchanged during the conduct of the study. Other assessments were rated by qualified site staff who were trained and, when necessary, certified to administer the assessments. Whenever possible, the CDR rater did not assess the other cognitive scales.
Inclusion criteria were modelled on International Working Group (IWG) criteria, which redefined AD as a clinicobiological syndrome that can be identified prior to the onset of dementia by an amnestic syndrome of the hippocampal type and supportive evidence from biomarkers (20). Key inclusion criteria were: age 50-85; recent gradual decline in memory (informant); abnormal memory function based on the Free and Cued Selective Reminding Test (FCSRT: free recall <17, or total recall <40, or [free recall <20 and total recall <42]); Mini-Mental Status Exam (MMSE) score ≥24; and Clinical Dementia Rating – global score (CDR-GS) of 0.5 with memory box score of 0.5 or 1; CSF Aβ(1-42) ≤600 ng/L as measured by the central laboratory.

Study Design and Outcome measures

The study consisted of an 8-week screening period, a double-blind treatment phase of 100 weeks, a final assessment at week 104, and follow-up visits at 16 and 52 weeks after the last dose. Participants were recruited from clinical sites, some of which were memory centers. Subjects meeting all eligibility criteria during screening were randomized 1:1:1 to receive either placebo,
105 mg, or 225 mg gantenerumab subcutaneously every four weeks. Assessments at screening included the CDR, the Functional Activities Questionnaire (FAQ), the Alzheimer’s Disease Assessment Scale – Cognition Subscale 11 and 13 item version (ADAS-Cog 11, ADAS-Cog 13 respectively), the MMSE, and the Free and Cued Selective Reminding Test (FCSRT-IR [words]) (Table 1). These assessments (listed in Table 1) were also conducted at baseline, and at 24-weeks interval (at weeks 24, 52, and 76) including final assessment at Week 104. The MMSE, ADAS-Cog, and FCSRT were obtained at additional 12 weeks intervals (at weeks 12, 36, 64, and 88), and the CDR was also obtained at weeks 64 and 88.

Table 1. Clinical Outcome Assessments

ClinRO: Clinician-rated Outcome Assessment; PerfO: Performance-based Outcome Assessment

 

Clinical Dementia Rating scale

The CDR was originally developed as a staging tool to categorize dementia severity into normal, questionable, mild, moderate or severe. Clinicians rate the severity of symptoms across six domains following a semi-structured interview with the subject and a reliable informant or collateral source (e.g., family member). There are three cognition domains (Memory, Orientation, Judgment & Problem Solving) and three functional domains (Community Affairs, Home & Hobbies, and Personal Care) (12, 21). The response options for each domain describe five degrees of impairment: 0=None; 0.5=Questionable (not present in the Personal care domain); 1=Mild; 2=Moderate; 3=Severe. The CDR-Global score, which determines dementia stage, is rated from 0-3. The Sum of Boxes score is a continuous measure of dementia severity and ranges from 0-18. For completeness, we report the psychometric properties of the CDR-Cognition and CDR-Function domains individually, acknowledging that the CDR was not intended for this purpose and that each domain contains few items (3 items) for traditional analyses (e.g. Chronbach’s α).

Alzheimer Disease Assessment Scale-Cognition (ADAS-Cog)

The ADAS-Cog-11 (22) is a structured scale that evaluates memory (word recall, word recognition), reasoning (following commands), language (naming, comprehension), orientation, ideational praxis (placing letter in envelope) and constructional praxis (copying geometric designs). Ratings of spoken language, language comprehension, word finding difficulty, and ability to remember test instructions are also obtained. The test is scored in terms of errors, with higher scores reflecting poorer performance. Scores can range from 0 (best) to 70 (worst). The 13-item version also includes Delayed Word Recall and Number Cancellation tasks, with scores ranging from 0 to 85.

Mini Mental State Exam (MMSE)

The MMSE consists of a set of standardized questions to evaluate possible cognitive impairment and help stage the severity level of this impairment. The questions target five areas; orientation, short term memory retention, attention, short term recall and language (23). The MMSE is scored as the number of correctly completed items with lower scores indicative of poorer performance and greater cognitive impairment. The total score ranges from 0 (worst) to 30 (best).

Functional Activities Questionnaire (FAQ)

The FAQ is an informant-based assessment in which caregivers rate abilities on 10 activities of daily living (ADLs) (24). The 10 items are scored as Dependent = 3, Requires assistance = 2, Has difficulty but does by self = 1, Normal = 0. The total score ranges from 0 to 30 with higher scores indicating worse functioning. The FAQ has demonstrated good sensitivity and specificity in differentiating MCI from very mild AD, by reflecting very mild functional impairment (25) .

Free and Cued Selective Reminding Test – Immediate Recall (FCSRT-IR)

The FCSRT-IR is a measure of memory under conditions that control attention and cognitive processing in order to obtain an assessment of memory unconfounded by normal age-related changes in cognition (26, 27). The FCSRT-IR used cards with four written words corresponding to a specific category cue, with immediate recall after each card followed by a cued recall using the category cue (28). Abnormal memory function according to FCSRT-IR was defined as a free recall score<17 (sum of free recall items), a total recall score<40 (sum of free recall and cued recall items), or a free recall score<20 and total recall score<42. FCSRT-IR performance has been associated with preclinical and early dementia in several longitudinal epidemiological studies. The CDR, ADAS-Cog, MMSE and FAQ may be viewed as composites, where a total score is based on the sum of item responses and individual items are intended to assess different cognitive and/or functional domains or concepts. Whilst total scores are also derived for FCSRT, items/words are not interpreted as individually meaningful.

Statistical methods

All screening and baseline analyses were conducted on the total sample. Analyses that included Week 52 and/or Week 104 data were conducted in the placebo group only to remove the potential impact of treatment from the evaluation of psychometrics. Patients were included in the analysis if they had completed measures at a given time point.

Test-retest reliability

Test–retest reliability is used to assess the degree to which a measure provides stable scores over time, assuming that the underlying condition of patients has not changed. This aspect of reliability was evaluated by intra-class correlation coefficients (ICCs, Shrout & Fleiss classification Random set 2, 1) (29) between the screening and baseline visits i.e. an interval approximately 8 Weeks (up to 12 Weeks was allowed for FCSRT-IR). Subjects were expected to remain clinically stable over this interval, whereas for longer intervals (baseline to Week 52, Week 52 to Week 104), clinical progression would be expected. Intra-class correlation coefficients that exceed 0.70 are generally assumed to be adequate (30).

Internal consistency

Internal consistency refers to the degree of association between the individual items that comprise a composite measure, and was measured by Cronbach’s α, which generally increases as the inter-correlation amongst test items increases (31). As a general rule, >0.7 is considered an appropriate target for internal consistency (30, 32, 33). Internal consistency was not calculated for the FCSRT-IR outcomes since these are essentially single item constructs.

Construct validity

Construct validity refers to the extent to which a measure adequately assesses an intended concept and may be evaluated by the association to other measures of both similar and different concepts. Relationships between the measures were examined in cross-section, using scores at baseline and change from baseline scores at Week 104. Spearman correlation coefficients (with Fisher’s adjustment) were used to test the correlation between continuous variables. It was expected that objective cognitive measures would be inter-correlated (≥0.4), as would functional measures, but that correlation between cognition and function measures may be lower. The following thresholds were used to assess the strength of the relationship: <0.2: Weak, ≥0.2 to <0.4: Modest, ≥0.4 to <0.6: Moderate, ≥0.6 to 0.8: Strong; ≥0.8: Very strong (34, 35).

Ability to detect change (responsiveness to decline)

As AD is a progressive neurodegenerative disease, decline over time may be used as a way to assess ability to detect change. As an effect size metric, standardized response means (SRM) for the change from baseline in the placebo arm were calculated as SRM = mean change divided by the standard deviation of change, at Week 104. For convenience, we considered values ≥0.2 to <0.5 as low and ≥0.5 to <0.8 as moderate responsiveness (36) (Table 3). Ceiling and floor effects were determined according to the proportion that received the highest and lowest scores at baseline.

Known groups validity

The difference between CDR Global score = 0.5 (Questionable dementia) and CDR Global Score = 1 (Mild dementia) groups were calculated, for each of the variables. This evaluation was conducted at Week 52 (with the exception of FCSRT for which Week 104 was used as the only available time-point), since this maximized the sample size in both CDR-GS = 0.5 and CDR-GS = 1 groups; CDR-GS of 0.5 was an inclusion criterion at screening and a reduced sample size was available at Week 104. Independent samples t-tests were used to assess the statistical significance of the between groups differences. A significant difference between groups is generally considered to reflect reasonable known groups validity.

 

Results

Patients

Seven hundred and ninety-seven subjects received allocated treatment (All Patients), mean age 70.4 years (SD 7.2), mean years of education 12.5 years (SD 4.5), 43.2% male. Two-hundred and sixty-six were randomized to placebo, with 104 completing the Week 104 visit (Placebo Arm), mean age 68.5 years (SD 6.8), mean years of education 12.3 years (SD 4.7), 43.8% male. The clinical characteristics of both populations are summarized in Table 2. Countries with the highest enrollment (>3%) included: the United States (14.3%), Spain (12.6%), Canada (7.3%), the United Kingdom (7.1%), Germany (6.9%), Italy (6.9%), France (6.4%), Australia (6.1%), Mexico (5.9%), Argentina (4.5%), and the Netherlands (4.1%).

Table 2. Clinical characteristics

*For FAQ, ADAS-Cog11, and ADAS-Cog13 n=221; †For FAQ, ADAS-Cog11, and ADAS-Cog13 n=105, FCSRT-IR assessments n=100; ADAS-Cog-11: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 11 item version, ADAS-Cog-13: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 13 item version, CDR-Cognition: Clinical Dementia Rating Cognition domain, CDR-Function: Clinical Dementia Rating Function domain, CDR-SB: Clinical Dementia Rating-Sum of Boxes, FAQ: the Functional Activities Questionnaire, FCSRT-IR: Free and Cued Selective Reminding Test – Immediate recall, MMSE: Mini Mental State Exam, SD: standard deviation.

 

Floor and Ceiling Effects

At the total score level, floor and ceiling effects were within acceptable ranges (Figure 1). At the item level, for all composite measures (i.e. CDR-SB, FAQ, ADAS-Cog and MMSE), notable ceiling effects (≥20% of patients at ceiling) were evident, showing that a large proportion of the enrolled pAD patient population were unimpaired in several of the items and/or domains assessed by these instruments (Figure 1). In addition, delayed word recall (ADAS-Cog and MMSE recall items) showed evidence of a floor effect.

Figure 1. Floor and Ceiling effects by item and total scores at baseline

All Patients. Dashed line represents threshold for notable floor or ceiling effect.

 

Test-retest reliability

Test–retest reliability for the total scores was generally >0.7, with the exception of ADAS-Cog11 (0.67), MMSE (0.52) and FCSRT-IR Cued Recall (0.68) (Table 3).

Table 3. Intraclass correlation coefficients, internal consistency, responsiveness and clinical validity

*All p values less than .0001; evaluation was conducted at Week 52 for all measures, with the exception of FCSRT for which Week 104 was used, in order to maximize n; ^Negative value due to scoring direction (lower score = worse cognition); α, standardized Cronbach’s alpha; ADAS-Cog-11: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 11 item version, ADAS-Cog-13: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 13 item version, CDR-Cognition: Clinical Dementia Rating Cognition domain, CDR-Function: Clinical Dementia Rating Function domain, CDR-SB: Clinical Dementia Rating-Sum of Boxes, FAQ: the Functional Activities Questionnaire, FCSRT-IR: Free and Cued Selective Reminding Test – Immediate recall, MMSE: Mini Mental State Exam. Internal consistency not appropriate for FCSRT.

 

Internal consistency

Internal consistency at screening and baseline was <0.5 for CDR-Cognition and CDR-Function, 0.65 for CDR-SB, 0.63 and 0.68 for ADAS-Cog11, and ADAS-Cog13, respectively; and 0.8 for FAQ. Chronbach’s α tended to increase over the study, exceeding 0.7 for most measures at later timepoints (Table 3). The exception was the MMSE, which had very low internal consistency at baseline, rising to 0.66 at Week 104.

Construct validity

Inter-correlation of scores at baseline and change from baseline to Week 104 are reported in Table 4. CDR-SB was most strongly correlated with FAQ (0.6 at baseline and change from baseline). However, CDR-SB and FAQ were not strongly correlated with the cognitive measures, ADAS-Cog, MMSE or FCSRT (all correlations ≤0.4), with the exception of the correlation between change in CDR-SB and ADAS-Cog13 change at Week 104 (0.5). Both CDR ‘cognition’ and ‘function’ items were equally well correlated with function as measured by FAQ. However, a low degree of correlation was seen between CDR function and ADAS-Cog for the baseline scores only. As expected, the objective cognitive tests, ADAS-Cog, MMSE and FCSRT tended to be more highly correlated with each other.

Table 4. Inter-correlation of scores

ADAS-Cog-11: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 11 item version, ADAS-Cog-13: the Alzheimer’s Disease Assessment Scale – Cognition Subscale 13 item version, CDR-Cognition: Clinical Dementia Rating Cognition domain, CDR-Function: Clinical Dementia Rating Function domain, CDR-SB: Clinical Dementia Rating-Sum of Boxes, FAQ: the Functional Activities Questionnaire, FCSRT-IR: Free and Cued Selective Reminding Test – Immediate recall, MMSE: Mini Mental State Exam.

 

Responsiveness to decline/Ability to detect change

Sensitivity to change was evaluated as the SRM (Table 3). Of the total scores, ADAS-Cog and FCSRT-IR were the least responsive, whilst CDR-SB, FAQ and MMSE were the most responsive. Importantly, CDR-SB and CDR cognition were free from floor and ceiling effects at Week 104, which may influence SRM. CDR function showed 25.6% at ceiling, FAQ 9.1% at ceiling and 0.3% at floor and MMSE 1.6% at ceiling, suggesting modest impact on SRM.

Known Groups Validity

For the evaluation of known groups validity, large and statistically significant (P<0.0001) differences were evident between subjects with a CDR-Global score of 0.5 versus those with a score of 1, for all measures
(Table 3). CDR-SB, CDR-Function and CDR-Cognition all had Cohen’s effect sizes of greater than 2 (≥small effect size).

 

Discussion

The Clinical Dementia Rating was devised as a global dementia-staging tool, taking into account results of clinician testing of cognitive performance and a rating of cognitive behavior in everyday activities, in six major categories of cognitive performance. Impairment is scored as decline from the person’s previous level due to cognitive loss alone, not impairment due to other factors, such as physical impairment, depression, or personality change (21). The CDR is considered to be a face valid measure of “the influence of cognitive loss on the ability to conduct everyday activities” (12). The CDR-SB has gained prominence in recent times as a single primary endpoint for clinical trials in early AD. Results from the crenezumab discontinued Ph III trial show a robust decline in CDR-SB score over 24 months in patients with both prodromal and mild AD, suggesting that, when enriching for “fast progressors” who are impaired on the FCSRT , the CDR-SB is sensitive to decline in early (prodromal-mild) AD (37). Whilst psychometric properties of the CDR-SB have been explored in early AD dementia (10), they have not been explored in a biomarker confirmed pAD population, and test-retest reliability, critical to repeated assessments, has not previously been published to our knowledge.
These data further support the psychometric properties of the CDR-SB, in demonstrating adequate test-retest reliability, a good degree of internal consistency especially over time, and also construct validity in terms of association to instrumental activities of daily living measured by the FAQ. Importantly, these properties are now confirmed in a pAD clinical trial population. Although there was evidence for ceiling effects in individual domains/items at the baseline assessment, this did not have a major impact on sensitivity to decline, and CDR-SB showed a greater degree of responsivity than ADAS-Cog and FCSRT-IR. However, SRM was lower than previously reported over two years in an early AD population derived from ADNI data (0.71 in this report, versus 1.03 in ADNI), which may result from differences in inclusion criteria (10). There were floor and ceiling effects at the item level for all composite measures; this may be an important consideration with respect to the coverage of relevant concepts and for potential sensitivity to disease progression in the early stages of the disease. Floor effects suggest that even at the early stages of disease, delayed free recall assessments may be markedly impaired, again impacting potential sensitivity.
Previous reports have focused on inter and intra-rater reliability. The novel finding for test-retest is of particular value, given the importance of reliability in clinical trial use. Strong test-retest reduces measurement error, which increases the likelihood of detecting true treatment effects. There was an improvement in internal consistency from baseline for all measures over the course of the study. One possible explanation for improved internal consistency may be ’other’ reliability, such as improved intra-rater reliability and reliability of subject and informant report, as all parties become more familiar with the scales and have more data available to inform them. In addition, regression to the mean, or disease progression bias could result in greater homogeneity of scores between items over time. The internal consistency for CDR-SB at Week 52 and 104 (Cronbach’s alpha 0.84 and 0.90, respectively) was similar to that observed in the French REAL.FR cohort study of patients with very mild-to-moderate AD (0.88) (15).
Specific to construct validity of the CDR, Tractenberg et al previously observed that in an AD dementia population, change in ‘cognitive’ items showed a modest correlation with change in MMSE and a low correlation with change in ADL, and ‘functional’ items the opposite pattern (13). Along with results from principal components analyses, this was seen as supportive of separate cognitive and function domains. In the present data, a correlation was observed between both the CDR cognition and function domains and FAQ, for both baseline and change scores (all 0.5). This may be seen as supportive of overall convergent validity with the FAQ (25). Inter-correlation of CDR-SB and FAQ may be driven by measurement of function and some direct overlap in item content and the use of informant report in both assessments. Cedarbaum et al (2013), found correlations with FAQ tended to be higher than with ADAS-Cog11 or ADAS-Cog13 for both cognitive (0.63, 0.55, and 0.59, respectively) and function domains (0.58, 0.42, and 0.45, respectively) in subjects with early or mild AD at baseline in the ADNI study (10). In addition, although their factor analysis showed some support for separate domains, there was overlap for “Judgment and problem solving” and “Community affairs” items in several cases, and a differential pattern based on disease severity was observed. Thus, the CDR may not capture function and cognition as separate domains but still address both, consistent with the original unitary measurement concept (“the influence of cognitive loss on the ability to conduct everyday activities”). Furthermore, low internal consistency reliability of these scores suggests there may be too few items for them to be reliable as separate measures.
For the FAQ and ADAS-Cog, adequate test-retest reliability and a good degree of internal consistency were observed. Both measures demonstrated construct validity in terms of association to related measures (CDR to FAQ and ADAS-Cog to MMSE and FCSRT). This is an important finding for the FAQ, given the lack of validation evidence for the scale beyond discriminative ability (38). Though there was evidence for ceiling effects with individual items at the baseline assessment for both measures, this did not have a major impact on sensitivity to decline for FAQ as the SRM for FAQ (0.73) was greater than for the other measures.
For the MMSE, assessment of psychometric properties was impacted by its use as a screening criterion, with only scores of between 24 and 30 out of 30 possible at screening. This would initially reduce range of scores and variance, decreasing power to obtain high alpha coefficients and impact the ability to adequately assess scale properties at the screening and baseline assessments in particular. Thus, caution is warranted in the interpretation of the results. Overall MMSE did show good sensitivity to decline, comparable to CDR-SB and FAQ (SRM=−0.71), with orientation to time as the single greatest contributory item (SRM=−0.63). This prominence of orientation as sensitive to decline across the different scales is consistent with other data, which has shown orientation to be sensitive to disease progression and important for inclusion in novel composite outcomes (39).
For the FCSRT-IR, adequate test-retest reliability was also observed, though this was also utilized as a screening inclusion criterion. Whilst the measures of free, cued, and total recall were relatively free from ceiling and floor effects, this did not translate to sensitivity to decline and SRMs were −0.2 for cued, −0.5 for free and −0.46 for total recall. Therefore, there may be limited additional value in FCSRT-IR as a longitudinal outcome measure in this patient population.
There are some limitations which could impact the generalizability of these findings. The study population was derived from a clinical trial, in which CDR-Global Score was one of the inclusion criteria, and thus we cannot rule out the possibility that this influenced the reporting of the CDR domains at baseline. Additionally, the CDR-Global score was used to define questionable dementia and mild dementia for the known groups validity analysis of CDR-SB. This may have impacted the analysis, as the CDR-SB and global score may be interrelated. Although industry standards were followed with regards to translation, we did not formally evaluate whether psychometric properties differed by culture or language. Furthermore, this was a biologically homogenous population with low levels of Aβ(1-42) and different results may be found in a more heterogeneous sample. This study population may have been subject to selection bias due to initial study recruitment methods (e.g. site selection), as well as individual interest in participating in a clinical trial. Loss to follow-up will also have impacted the representativeness of longitudinal analyses. Finally, further work is required to establish what constitutes a meaningful change on the CDR-SB in prodromal AD.
In conclusion, CDR-SB showed adequate psychometric properties in the pAD population and its sensitivity to decline over time further support its utility as a clinical trial outcome measure. In addition, its conceptual basis as a measure of the influence of cognitive loss on the ability to conduct everyday activities was supported by the construct validity data. These data reinforce the continued use of CDR-SB as a single primary outcome measure in early AD clinical trials, such as the phase III gantenerumab GRADUATE program and the phase III BAN2401 Clarity AD trial. In addition, validity and reliability of the other assessments, particularly the FAQ, is further supported.
Efforts to develop novel cognitive and functional assessments free from ceiling and floor effects and with greater sensitivity to change in this population should continue. However, given the adequate psychometrics of the CDR-SB, its clinical relevance, and the lack of a clearly established relationship between other objective cognitive endpoints and clinical benefit, at least for patients with early AD, there is good reason to continue to employ the CDR-SB in treatment trials.

 

Acknowledgements: The authors would like to acknowledge to important contributions of the SCarlet RoAD Investigators, Patients and their Families participating globally in Argentina, Australia, Belgium, Brazil, Canada, Chile, the Czech Republic, Denmark, Finland, France, Germany, Italy, Mexico, the Netherlands, Poland, Portugal, Russia, Spain, South Korea, Sweden, Switzerland, Turkey, the United Kingdom and the United States. This work was supported by F. Hoffmann-La Roche Ltd, Basel, Switzerland.

Conflicts of interest: FM is an employee of Genentech Inc., South San Francisco, USA. MM, PD, PF, DAS, and CJL are employees of F. Hoffmann-La Roche Ltd, Basel, Switzerland. PD owns stock in F. Hoffmann-La Roche Ltd. RD is an employee and owns stock in Genentech Inc. and F. Hoffmann-La Roche Ltd. MB has received grants from Merck & Co., Inc. related to the submitted work (paid to the institution); she has received grants from Araclon, Biogen Research Ltd, Bioberica, Grifols, Lilly S.A, Merck Sharp & Dohme, Nutricia SRL, Oryzon Genomics, Piramal Imaging Ltd, Schwabe Farma Iberica SLU and Merck & Co, Inc. within the last 36 months outside the submitted work (paid to the institution); she has served as a consultant or provided scientific advisory board services and/or given lectures for Roche, Araclon, Bioberica, Grifols, Kyowa Hakko Kirin, Laboratorios Servier, Lilly, S.A., Merck Sharp & Dohme, Nutricia SRL, Schwabe Farma Iberica SLU.

Ethical Standards: Institutional Review Boards (IRBs) approved the SCarlet RoAD study, and all participants gave informed consent before participating.

Data sharing statement: Qualified researchers may request access to individual patient-level data through the clinical study data request platform: https://vivli.org. Further details on Roche’s criteria for eligible studies are available here: https://vivli.org/members/ourmembers. For further details on Roche’s Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see here: https://www.roche.com/research_and_development/who_we_are_how_we_work/clinical_trials/our_commitment_to_data_sharing.htm

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

References

1. Dubois B, Feldman HH, Jacova C, Hampel H, Molinuevo JL, Blennow K, et al. Advancing research diagnostic criteria for Alzheimer’s disease: the IWG-2 criteria. Lancet Neurol. 2014;13(6):614-29.
2. Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7(3):270-9.
3. FDA. Early Alzheimer’s Disease: Developing Drugs for Treatment Guidance for Industry, Draft Guidance. 2018.
4. EMA. Discussion paper on the clinical investigation of medicines for the treatment of Alzheimer´s disease and other dementias. In: CHMP, editor. London: European Medicines Agency; 2014. p. 33.
5. FDA. Multiple Endpoints in Clinical Trials. In: (CDER) CfDEaR, (CBER) CfBEaR, editors. Rockville, MD2017.
6. EMA. Guideline on the clinical investigation of medicines for the treatment of Alzheimer’s disease In: CHMP, editor. London2018.
7. Jekel K, Damian M, Wattmo C, Hausner L, Bullock R, Connelly PJ, et al. Mild cognitive impairment and deficits in instrumental activities of daily living: a systematic review. Alzheimers Res Ther. 2015;7(1):17.
8. Wang J, Logovinsky V, Hendrix SB, Stanworth SH, Perdomo C, Xu L, et al. ADCOMS: a composite clinical outcome for prodromal Alzheimer’s disease trials. J Neurol Neurosurg Psychiatry. 2016.
9. FDA. Guidance for Industry Alzheimer’s Disease: Developing Drugs for the Treatment of Early Stage Disease DRAFT GUIDANCE. In: CDER, editor. 2013.
10. Cedarbaum JM, Jaros M, Hernandez C, Coley N, Andrieu S, Grundman M, et al. Rationale for use of the Clinical Dementia Rating Sum of Boxes as a primary outcome measure for Alzheimer’s disease clinical trials. Alzheimers Dement. 2013;9(1 Suppl):S45-55.
11. Kozauer N, Katz R. Regulation of drugs for early Alzheimer’s disease. N Engl J Med. 2013;369(3):288.
12. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993;43(11):2412-4.
13. Tractenberg RE, Weiner MF, Cummings JL, Patterson MB, Thal LJ. Independence of changes in behavior from cognition and function in community-dwelling persons with Alzheimer’s disease: a factor analytic approach. J Neuropsychiatry Clin Neurosci. 2005;17(1):51-60.
14. Morris JC, Ernesto C, Schafer K, Coats M, Leon S, Sano M, et al. Clinical dementia rating training and reliability in multicenter studies: the Alzheimer’s Disease Cooperative Study experience. Neurology. 1997;48(6):1508-10.
15. Coley N, Andrieu S, Jaros M, Weiner M, Cedarbaum J, Vellas B. Suitability of the Clinical Dementia Rating-Sum of Boxes as a single primary endpoint for Alzheimer’s disease trials. Alzheimer’s & dementia : the journal of the Alzheimer’s Association. 2011;7(6):602-10.e2.
16. Rockwood K, Strang D, MacKnight C, Downer R, Morris JC. Interrater reliability of the Clinical Dementia Rating in a multicenter trial. J Am Geriatr Soc. 2000;48(5):558-9.
17. FDA. Patient-Reported Outcome Measures: Use in Medical Product Development to Support Labeling Claims:Guidance for Industry. 2009.
18. Powers JH, 3rd, Patrick DL, Walton MK, Marquis P, Cano S, Hobart J, et al. Clinician-Reported Outcome Assessments of Treatment Benefit: Report of the ISPOR Clinical Outcome Assessment Emerging Good Practices Task Force. Value Health. 2017;20(1):2-14.
19. Wild D, Grove A, Martin M, Eremenco S, McElroy S, Verjee-Lorenz A, et al. Principles of good practice for the translation and cultural adaptation process for patient-reported outcomes (PRO) measures: report of the ISPOR task force for translation and cultural adaptation. Value in Health. 2005;8(2):94-104.
20. Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P, Cummings J, et al. Research criteria for the diagnosis of Alzheimer’s disease: revising the NINCDS-ADRDA criteria. Lancet Neurol. 2007;6(8):734-46.
21. Berg L. Clinical Dementia Rating (CDR). Psychopharmacol Bull. 1988;24(4):637-9.
22. Rosen WG, Mohs RC, Davis KL. A new rating scale for Alzheimer’s disease. Am J Psychiatry. 1984;141(11):1356-64.
23. Folstein MF, Folstein SE, McHugh PR. «Mini-mental state». A practical method for grading the cognitive state of patients for the clinician. Journal of psychiatric research. 1975;12(3):189-98.
24. Vellas B, Andrieu S, Sampaio C, Coley N, Wilcock G, European Task Force G. Endpoints for trials in Alzheimer’s disease: a European task force consensus. Lancet Neurol. 2008;7(5):436-50.
25. Pfeffer RI, Kurosaki TT, Harrah CH, Jr., Chance JM, Filos S. Measurement of functional activities in older adults in the community. J Gerontol. 1982;37(3):323-9.
26. Grober E, Buschke H. Genuine memory deficits in dementia. Developmental Neuropsychology. 1987;3:13-36.
27. Grober E, Hall C, McGinn M, Nicholls T, Stanford S, Ehrlich A, et al. Neuropsychological strategies for detecting early dementia. J Int Neuropsychol Soc. 2008;14(1):130-42.
28. Grober E, Sanders AE, Hall C, Lipton RB. Free and cued selective reminding identifies very mild dementia in primary care. Alzheimer Disease and Associated Disorders. 2010;24(3):284.
29. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86(2):420-8.
30. Nunnally JC, Bernstein IH. Psychometric Theory. 3rd. ed. ed. New York: McGraw-Hill; 1994.
31. Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika. 1951;16(3):297-334.
32. DeVon HA, Block ME, Moyle-Wright P, Ernst DM, Hayden SJ, Lazzara DJ, et al. A psychometric toolbox for testing validity and reliability. Journal of Nursing scholarship. 2007;39(2):155-64.
33. DeVellis RF. Scale development: Theory and applications: Sage publications; 2016.
34. Swinscow T. Correlation and regression. Statistics at square one. 9 ed1997.
35. Bartz AE. Basic Statistical Concepts. 4 ed. Upper Saddle River, NJ: Merrill; 1999.
36. Middel B, Van Sonderen E. Statistical significant change versus relevant or important change in (quasi) experimental design: some conceptual and methodological problems in estimating magnitude of intervention-related change in health services research. International Journal of Integrated Care. 2002;2.
37. Sink K, Djakovic S., Smith JW, Hu N, Mackey H, Ostrowitzki S, et al. FCSRT inclusion criteria support recruitment of a population with early Alzheimer’s disease likely to progress over 24 months: results from the CREAD trial. Clinical Trials in Alzheimer’s Disease; San Diego, USA2019.
38. Kaur N, Belchior P, Gelinas I, Bier N. Critical appraisal of questionnaires to assess functional impairment in individuals with mild cognitive impairment. Int Psychogeriatr. 2016;28(9):1425-39.
39. Vellas B, Bateman R, Blennow K, Frisoni G, Johnson K, Katz R, et al. Endpoints for Pre-Dementia AD Trials: A Report from the EU/US/CTAD Task Force. J Prev Alzheimers Dis. 2015;2(2):128-35.

MOLECULAR SUBTYPING OF MILD COGNITIVE IMPAIRMENT BASED ON GENETIC POLYMORPHISM AND GENE EXPRESSION

H.-T. Li1, S.-X. Yuan1, J.-S. Wu2, X.-Z. Zhang3, Y. Liu4, X. Sun1 and For the Alzheimer’s Disease Neuroimaging Initiative†

1. State Key Laboratory of Bioelectronics, School of Biological Science and Medical Engineering, Southeast University, Nanjing, P. R. China; 2. School of Geography and Biological Information, Nanjing University of Posts and Telecommunications, Nanjing, P.R. China; 3. Department of Biostatistics, Rollins School of Public Health, Emory University, Atlanta, USA; 4. The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, Nanjing, P.R. China

Corresponding Author: Xiao Sun, State Key Laboratory of Bioelectronics, School of Biological Science and Medical Engineering, Southeast University, Nanjing 210096, P. R. China, xsun@seu.edu.cn

J Prev Alz Dis 2020;
Published online November 23, 2020, http://dx.doi.org/10.14283/jpad.2020.65

 


Abstract

Background: Alzheimer’s Disease (AD) is a neurodegenerative brain disease in the elderly. Recent studies have revealed the heterogeneous nature of AD. Mild Cognitive Impairment (MCI) is the prodromal stage of AD.
Objectives: In this study, we identified subtypes of MCI based on genetic polymorphism and gene expression.
Methods: We utilized the two types of omics data, namely genetic polymorphism and gene expression profiling, derived from 125 MCI patients’ peripheral blood samples from the ADNI-1 dataset. Similarity network fusion (SNF) algorithm was implemented to cluster MCI patient subtypes. And 185 MCI patients in ADNI-2 were utilized to evaluate the effectiveness of this method. Two MCI subtypes were identified by implementing the SNF algorithm.
Results: We used Kaplan-Meier analysis and log-rank testing for the conversion from MCI to AD between two subtypes, and p-value is 4.58×10-3. In addition, we compared patients among two MCI subtypes by the following factors: the changes in Alzheimer’s Disease cognitive scales and MRI image; significantly enriched pathways based on differentially expressed genes. This study proved that MCI is a heterogeneous disease by concluding that AD development in two MCI subtypes is significantly different.
Conclusions: MCI patients with different molecular characteristics have different risks converting to AD. In addition to evaluating statistics, genetic polymorphism and gene expression profiling from MCI patients’ peripheral blood are non-invasiveness and cost-effectiveness markers to identify MCI subtypes for clinical application.

Key words: Alzheimer’s disease, mild cognitive impairment, molecular subtyping, similarity network fusion.


 

Introduction

Alzheimer’s disease (AD) is a chronic degenerative brain disease and the most common cause of dementia in the elderly. According to statistics, about 10% of people older than 65 suffer from AD (1). Due to the lack of understanding of its causes, effective drugs or treatments of AD is yet not invented.
AD is a complex and heterogeneous disease caused by multiple different genetic factors (2). Recently, more and more studies, such as clinicopathologic (3), atrophy patterns on magnetic resonance imaging (MRI) (4) and amyloid-β fibril polymorphism on solid-state nuclear magnetic resonance (ssNMR) (5), have supported the hypothesis on the existence of distinctive AD molecular subtypes. For example, the rapidly progressive form in which neurodegeneration occurs within months and a typical prolonged-duration form are two AD clinical subtypes that been well recognized. Recently, some researchers have found that different AD clinical subtypes were correlated with fibril formations subtypes by researching on 37 brain samples from 18 deceased Alzheimer’s patients obtained by using ssNMR (5). Lately, another research assigned 4,050 people with late-onset AD into six subgroups according to their cognitive functioning at the time of diagnosis and then utilized genetic data to find the biological differences across these subgroups (6). This study supported the biological coherence of cognitively defined subgroups. With more in-depth studies of Alzheimer’s subtypes, new diagnostic criteria, and treatment of AD that target specific kinds of AD subtypes can be expected.
Mild Cognitive Impairment (MCI) is known as the prodromal stage of AD. MCI is a neurological disorder in which an elderly has mild but measurable changes in cognition. It is worth mentioning that not all people with MCI will develop AD. Studies suggest that MCI patients progress to AD at a rate of approximately 10% every year (7). Early identification of high-risk subtypes MCI patients appears to be significant and may enable a more effective, preventive treatment, thereby increasing the possibility of delaying even avoiding conversion from MCI to AD.
For the above reasons, we believe that MCI is a heterogeneous disease. Identifying the subtypes of MCI is critical for implementing precision medicine approaches and for ultimately developing successful subtype-specific drugs for AD. And classifying MCI patients into meaningful subtypes may provide better targeted treatment to delaying or preventing the conversion from MCI to AD. Genetic factors play an important role in MCI and AD (2). However, to our knowledge, the molecular subtyping of MCI based on integrative multi-omic data was not taken into consideration among current studies. Therefore, in this study, we took advantage of the two types of omics data, including genetic polymorphism and gene expression, derived from 125 MCI patients’ peripheral blood samples from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to identify the MCI patient subtypes (8). We used the Similarity Network Fusion (SNF) algorithm to cluster the two types of omics data to determine the subtypes of 125 MCI patients (9). For testing the effectiveness and reliability of the SNF algorithm, 185 MCI patients from ADNI-2 were identified the subtype by the label propagation algorithm (9, 10). The flow chart of our research is illustrated in Figure 1. To prove the biological and clinical significance of subtyping patients based on our method, these different subtypes were compared by the following factors: the time difference of the conversion from MCI to AD; cognitive scales and MRI image; significantly enriched pathways based on differentially expressed genes separately.

Figure 1. Flow chart of our research. (a) The Similarity Network Fusion (SNF) algorithm is used to integrate SNP and gene expression data for subtype identification of MCI patients; (b) The label propagation algorithm is applied to predict the subtype of any new patient from ADNI-GO/2 for testing the effectiveness and reliability of the SNF algorithm

 

Methods

Genomic data and imaging data

Data used in this study were downloaded from ADNI. ADNI was a multi-site study proposed by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and Bioengineering (NIBIB) and the Food and Drug Administration (FDA) in 2003. This organization is holding an ongoing, longitudinal, multicenter study. Its primary goal is to test whether clinical, imaging, genetic, and biochemical biomarkers are effective in clinical trials of MCI and AD. The first stage of ADNI, as known as ADNI-1, was completed in 2010 (8). More up-to-date and detail information is available at http://adni.loni.usc.edu/.
In this article, we used combinations of multi-omics data (genetic polymorphism and gene expression) from the ADNI-1 and ADNI-GO/2 study to identify the MCI molecular subtypes and to predict the conversion from MCI to AD. 125 MCI patients’ SNP and gene expression data were downloaded from ADNI-1 for identification MCI subtypes. Meanwhile, 185 MCI patients were downloaded from ADNI-GO/2 as an independent verification dataset for predicting the subtype of any new patients. The information on MCI patients is listed in Supplementary Excel file 1. Both profiling were collected from peripheral blood samples. ADNI-1 and ADNI-GO/2 subjects were genotyped using the Human 610-Quad BeadChip and Illumina Human Omni Express BeadChip, respectively. Only SNP markers were analyzed for subsequent analysis. Quality control steps were performed on genetic polymorphism using the software package named PLINK (11), release v1.90b.5. SNPs with missing rate >0.05, minor allele frequency < 0.05, and Hardy–Weinberg equilibrium P < 10−3 were excluded from the genetic polymorphism set. Then the SNP data was applied by using the IMPUTE2 program for imputing the missing data with NCBI 1000 Genomes build 37 (UCSC hg19) as the reference panel (12). The Affymetrix Human Genome U219 Array was carried out for expression profiling, which contains 530,467 probes. Thenceforth, we used an R package named RMA for the normalization of gene expression microarray data (13). Finally, 49,293 transcripts were kept in this study.
There are various clinical/cognitive assessment scores from ADNI that are useful to compare clinical information between two subtypes of patients, including Mini Mental State Examination (MMSE), Clinical Dementia Rating Sum of Boxes (CDR-SB) and Activities of Daily Living Score (from the Functional Activities Questionnaire, FAQ). In addition, we downloaded T1 weighted MRI images in NIFTI format from 125 MCI patients’ baseline, 24-month follow-up data set in ADNI, and structural MRI scan applied inversion recovery-fast spoiled gradient recalled (IR-SPGR) for researching two clusters of MCI patients’ differences in areas of brain atrophy. VBM analyses were performed using the SPM12 toolkit (Statistical Parametric Mapping software, http://www.fil.ion.ucl.ac.uk/spm/sofware/ spm12) running under MATLAB 2013a (14).

MCI subtype identification based on similarity network fusion

We applied the similarity network fusion (SNF) algorithm to cluster the MCI patient subtypes (9). SNF is an integrated characterization of genomic profiling at multiple levels for subtype identification. The advantage of using SNF is that it is based on complementarity in multiple genomic data types. First, the SNF algorism uses a similarity measure to constructs a patient-by-patient similarity network for each genomic data type. The nodes of the network for each data type represent patients and the weighted edges are equivalent to pairwise sample similarities. Next, the network fusion step updates every network using a nonlinear method named message-passing theory. Each iteration makes these networks more similar to each other. After many iterations, multiple networks converge to a fusion network. Finally, the fusion network is clustered into several subtypes based on spectral clustering methods. The illustrative example of SNF steps are shown in Figure S1. Some patients (002_S_0729, 010_S_0161 and 011_S_1282 from cluster-1; while 005_S_0546, 027_S_1045 and 037_S_0150 from cluster-2) were used as examples to explain the clustering process of the SNF method (Figure S1 (d)).
More formally speaking, given n MCI patients and M omics (SNP and expression data in this study), the sample×sample similarity graph G=(N, W) is constructed, where node set N represents the samples x1,x2,…,xn and the edges weight W(i, j) represents the weight between xi and xj. W is defined by:


where d(m) (xi,xj) is the Euclidean distance between sample xi,xj for the m-th omic. α is a hyperparameter and α=0.8 in this study. ε is expressed as below:


where Ki is the number of neighbours of xi and Ki=30 in this study, mean (ε(xi, Ki)) is the average distance between xi and each of its neighbors. ε is introduced to eliminate the scaling problem.
A transition probability matrix is constructed between all MCI patients initially by:


Meanwhile, a transition probability matrix between nearest neighbors is defined by:


where Ni represent a set of i’s k nearest neighbors in matrices with measurements from the m-th omic.
Then, the matrix P is updated based on message-passing theory iteratively between the k nearest neighbors by formula:


where Pq(m) is the matrix for omic m at iteration q. The iterative process means that the connection information of different networks is exchanged to achieve the final uniform network.
After completing the network fusion, low-weight edges in each network disappear, and high-weight edges are retained. SNF reduces the noise among these steps, which makes this method robust to noise and the data heterogeneity. Finally, based on spectral clustering methods, namely minimize RatioCut, the fusion network is clustered into several subgroups. Such subgroups are considered as our resulting subtypes. The details of SNF reference (9).

Any new MCI patient’ subtype prediction based on label propagation

We adopted label propagation algorithm which is a simple iterative semi-supervised learning algorithm based on network structure to identify the subtype of the new MCI patient (9, 10). Assume n patients have been determined into y subtypes by the SNF method with a fused network F. To predict the subtype of a new patient, a similarity matrix F=[F s;s’ 1] is constructed, where s is the similarities vector calculated by SNF. Define a (n+1)×(n+1) probabilistic transition matrix T:


where Tij is the probability of jumping from node j to i. Also we define a (n+1)×y label matrix Y, whose i-th row representing the label probabilities of node yi. We iterate the propagation process as follows:
Repeat the following steps:


This process will converge usually in 1000 iterations. And we can predict the subtype of the new patient given by converged Y.

Results

Clustering of MCI patients

We downloaded 138 MCI patients’ gene expression profiling and 361 MCI patients’ genetic polymorphism data from the ADNI-1 dataset. The number of MCI patients with both genetic polymorphism and gene expression was 125. Hence, we used these MCI patients in this article for integrating the two types of omics data to identify MCI patient subtypes. Moreover, 276 MCI patients’ SNP data and 302 gene expression profiling were downloaded from the ADNI-GO/2 dataset. 185 MCI patients who have SNP data, gene expression data, and clinical follow-up data for greater than 36 months were selected as an independent verification set to evaluate the effectiveness of this method. Table S1 shows the characteristics of the MCI patients included in this study.
The subtypes of MCI patients in the ADNI-1 dataset were identification based on SNF method (9). In the beginning, quality control steps were performed on genetic polymorphism using the software package named PLINK (11) and gene expression profiling using an R package named RMA (13) as described in method. Then, we utilized SNF to cluster MCI patients using both SNP and gene expression profiling after quality control. SNFtool R package (v2.3.0) was applied with the parameters K = 30, alpha = 0.8, T = 20 (9). Spectral clustering implemented in the SNFtool package was run on the SNF fused similarity matrix to obtain the groups that each corresponding to k=2 to 5.
After executing the SNF algorithm, we chose the best number of clusters according to two main approaches of the spectral clustering method. One is the connectivity of the network, and the other is to make use of the structure of eigenvectors of the Laplacian L (9). However, the optimal number of clusters based on the connectivity of the network is 2, the best number decided by the other approaches is 3. Therefore, we used the highest average silhouette score as an assistance approach to decide the optimal number of clusters. The silhouette score represents the coherence of clusters to evaluate whether patients are more similar within subtypes. In other words, the silhouette score condenses the cluster quality for each patient’s omics data into a single score that ranges from 1.0 to -1.0. Hence, we had identified two subtypes. The number of patients in cluster-1 is 61, and cluster-2 has 64 patients.
To prove the biological and clinical significance of subtyping patients based on the SNF method, we applied the label propagation algorithm to assign new patients to subtypes in the ADNI-2 datasets (9, 10). Genotype data of MCI patients from the ADNI-GO/2 dataset were downloaded, quality controlled, imputed to the Illumina 610Quad platform and combined. Genotype imputation was conducted to estimate unobserved genotypes. Impute2 software was used with NCBI 1000 Genomes build 37 (UCSC hg19) as the reference panel (12). After executing the label propagation algorithm to 185 patients in ADNI-GO/2, 60 MCI patients were identified in cluster-1, while 125 patients were identified in cluster-2. The detail information on the subtypes of MCI patients in the ADNI-1 and ADNI-GO/2 dataset is listed in Supplementary Excel file 1.

Two MCI subtypes supported by clinical manifestations

We first examined the time difference of the conversion from MCI to AD between two subtypes of patients. Because the exact date of conversion to AD was not known, we used the midpoint between the last follow-up without an AD diagnosis and the first follow-up with an AD diagnosis for analyses. Subjects who did not convert were censored at the time of their last interview. We performed a Kaplan-Meier analysis on MCI of these two clusters. As is shown in Figure 2(a), P-value is 4.58×10-3, demonstrating a significantly different amount of time is consumed for MCI-to-AD conversion between two clusters. Patients that develop the disease more rapidly (red solid line) were cluster-1 MCI patients, and the others (blue dashed line) were cluster-2 MCI patients.

Figure 2. The Kaplan-Meier plot analysis on MCI of the two clusters of clinical data. X axis represents time past after MCI patients participating the study, while Y axis represents estimated percentages of stable MCI patients. The red solid line represents cluster-1 MCI patients in ADNI-1 (a) and ADNI-GO/2 (b), while the blue dashed line represents cluster-2 MCI patients in ADNI-1 (a) and ADNI-GO/2 (b)

 

We also considered the changes in Alzheimer’s Disease cognitive scales. Cognitive function status was measured by the Mini-Mental State Examination (MMSE) (rating 0–30, higher scores indicate good cognitive function), the Clinical Dementia Rating Sum of Boxes (CDR-SB) (rating 0–25, with higher scores representing greater impairmen) and the Functional Assessment Questionnaire (FAQ) (range 0–30, with higher scores representing greater impairment) in two years for two MCI subtypes of patients (8). As is shown in Figure 3(a), cognitive decline in cluster-1 MCI patients tends to be more remarkable than that of cluster-2 over 24 months.
To test the effectiveness and reliability of the SNF algorithm through its application on ADNI-GO/2 patients, we examined the time difference of the conversion from MCI to AD between two subtypes of all MCI patients. As is shown in Figure 2(b), this gives a log-rank P-value of 2.26×10-4. And three AD cognitive scales were also displayed in two years for two MCI subtypes of patients in the ADNI-2 dataset, which is shown in Figure 3(b). The scores change trends of all three cognitive scales in ADNI-GO/2 are similar to the ADNI-1 dataset. Thus, it proved the validity of the SNF method for subtyping MCI patients based on integrative genetic polymorphism and gene expression. Meanwhile, the cluster-1 subtypes having the worse prognosis than the cluster-2 subtypes.

Figure 3. Changes in AD cognitive scales (MMSE, CDR, FAQ) in two years for two MCI subtypes in ADNI-1 (a) and ADNI-GO/2 (b). X axis represents time past after MCI patients participating the study, while Y axis represents Alzheimer’s Disease cognitive scales score. Cognitive decline in cluster-1 MCI patients (red) is tend to be more remarkable than that of cluster-2 (blue) over 24 months

Two MCI subtypes supported by MRI image

We further analyzed the MRI images to illustrate the difference between two clusters of MCI patients’ ADNI baseline and 24-month follow-up MRI dataset using voxel-based morphometry (VBM) analyses in atrophy areas (15). VBM analysis has been developed for characterizing differences in the local composition of brain tissue using MRI and is not restricted to previously called region-of-interest measurements.
Firstly, we normalized images with the voxel sizes of 1.5×1.5×1.5mm3 because it could preserve the total amount of signal in the images. After normalization, T1-weighted images were segmented into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) using default option parameters on SPM12’s unified segmentation procedure. After that, we transformed patients’ images to the Montreal Neurological Institute (MNI) co-ordinate space using a template. Cognitive impairment is related to the MRI of GM decline on longitudinal analysis. Hence, on GM images, the spatial normalization approach was performed with the diffeomorphic anatomical registration using the exponentiated Lie algebra (DARTEL) algorithm (16). Subsequently, the images were smoothed with a 10-mm full-width at half-maximum isotropic Gaussian smoothing kernel. The results of GM images were analyzed with the two-sample t-test. For voxels in GM probability maps between baseline and 24 months, we selected those voxels with P<0.05 corrected by False Discovery Rate (FDR), and only regions of more than 100 contiguous selected voxels were considered in the analysis. To analysis the result of GM atrophy origins, we utilized the predefined anatomical masks obtained from an extension to the SPM package – XjView toolbox (http://www.alivelearn.net/xjview/) and the automated anatomical labeling (AAL, http://www.gin.cnrs.fr/en/tools/aal-aal2/) (17).
Based on the current official anatomical nomenclature proposed by Guilherme et al., the brain structure was divided into six lobes: frontal, parietal, occipital, temporal, insular, and limbic (18). The atrophic number of significantly different voxel regions is shown in Table 1. The result of the above steps was characterized by XjView.The comparison of cluster-1 (a) and cluster-2 (b) MCI patients’ regions of gray matter atrophy between baseline and 24-month follow-up MRI images are shown in Figure 4(a,b).

Figure 4. . Display of voxels with significantly brain areas of decreased gray matter intensity in each cluster. Images are 3D render view of (a) cluster-1 and (b) cluster-2 in sagittal, coronal and transversal. And paired images are MCI patients’ baseline MRI images compared to those of 24-month follow-up using VBM analyses. Colored voxels show regions that were significant in the analyses with p<0.05 corrected by FDR, and regions threshold of 100 contiguous voxels. The color brighter (yellow) indicates the more significant area of brain atrophic voxels in 24 month. (c) The atrophic size of significantly different voxel bunches within six lobes in 24 months

Figure 4(c) reveals that the atrophic size of significantly different voxel bunches of cluster-1 MCI patients in 24 months are apparently larger than that of cluster-2 MCI patients. In addition, the proportion of the atrophic voxels in six lobes accounted for 46.26% of total number of brain atrophic voxels in cluster-1, while in cluster-2 this ratio is 25.00%. This result indicates that not only was the atrophy of voxels in cluster-1 patients significantly more than that of cluster-2 patients, but also the location of atrophy was also concentrated in the functional areas of the brain. Therefore, by comparing the MRI images of cluster-1 and cluster-2 MCI patients collected from two-year data, one can see that AD development of cluster-1 patient is faster than that of cluster-2. Hence, this proves the usefulness of the subtype classification in clinical.

Two MCI subtypes supported by gene annotation

Subsequently, differential expressions of mRNA of MCI cluster-1, cluster-2 compared with the cognitively normal samples were each computed using R package named limma (19). Adjust-P value< 0.05 served as the screening conditions for the significant differences. The significantly different expression gene-set of cluster-1 had 3156 genes, while that of cluster-2 had 178 genes. We applied the functional annotation tool of “Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment” and “Gene ontology (GO)” in Enrichr, which was an integrative web-based software application that included many new gene-set libraries for gene and sequence annotations (20). Enrichr provided an adjustment P-value and combined score to annotate the biological significance of differentially expressed genes. An adjust p-value, 0.1, was chosen as significant thresholds upon filtering the pathway data. Because of too many biological processes in GO, a threshold of the combined score was considered. Common GO analyses were performed with a cut-off of combined score 20 and adjust p-value 0.1. The definition of the combined score in EnrichR is to integrate both p-value and z-score with the formula c = z-score•log(p-value), where c is the combined score, represented by p-value computed using the Fisher exact test, and z-score computed by assessing the deviation from the expected rank. The significant pathways and biological processes of differentially expressed genes between cluster-1, cluster-2, and control are shown in Figure 5. The full list of KEGG pathways and GO enrichment analysis information is in Supplementary Excel file 2.

Figure 5. The enriched significant KEGG pathways and GO biological processes bubble plot of differentially expressed genes (DEG) with FDR<0.05. (a) cluster-1 DEG KEGG enrichment, (b) cluster-2 DEG KEGG enrichment, (c) cluster-1 DEG GO enrichment, (d) cluster-2 DEG GO enrichment. The size of the dots represents the count of DEG in the corresponding pathways or GO terms. Y axis represents the enrichment pathways and biological processes. (a, b) X axis represents the opposite of the logarithm of p-value for each pathway, and (c, d) X axis represents the combine score which is defined by Enrichr for each biological process

 

The most remarkable pathways of cluster-1 are the following: RNA degradation, Amino sugar and nucleotide sugar metabolism, and RNA transport. These pathways are related to a wild range of biological processes. Meanwhile, the significant pathways of cluster-1 were predominated by immune system-related biological fields, such as B cell receptor signaling, TNF signaling and some microbial infection pathway (Epstein-Barr virus infection, Shigellosis and Legionellosis). More research results showed that inflammation is closely related to AD. In the brain, immune system cells called microglia is activated by the presence of toxic amyloid-β and tau proteins (21). Microglia tries to get rid of the remnants of inflammasomes in tiny clumps. However, these remnants continued to spread new amyloid-β clusters as well as aggravating the state of AD. Notably, Epstein-Barr virus infection is also one of the significant pathways. Epstein-Barr virus is known to be the one of herpes viruses. Recent research indicated that herpes viruses abundance was significantly associated with modulators of APP metabolism which revealed viral regulation of AD risk by multiscale networks (22). And insulin signaling pathways have a close relationship with AD. AD has been considered as a metabolic dysfunction disease associated with impaired insulin signaling (23). Proteolytic processes contribute to the amyloid cascade, and proteolysis of tau may be critical to neurofibrillary degeneration, which correlates with AD (24).
The significant biological processes in GO enrichment of cluster-1 are the regulation of transcription and protein catabolic process. These biological processes are closely correlated. Regulation of the transcription, DNA-templated is any process that modulates the frequency, rate or extent of cellular DNA-templated transcription. Regulation of transcription of AD genes might be an important player in the neurodegenerative process. For example, the APP gene is ubiquitously expressed in a variety of tissues, with the highest expression shown in neuronal cells. The abnormally expressed APP will lead to an increased amount and deposition of the amyloid β peptide (Aβ) in the brain triggering AD-related neuronal degeneration (25). Mutant forms of ubiquitin may inhibit proteolysis within neurons, making these cells susceptible to inclusion formation. Therefore, some researchers hold the contention that neurodegenerative diseases collectively referred to as “ubiquitin protein catabolic disorders”. Especially, similar to the KEGG analysis of cluster-1 MCI patients, the significant biological processes are also associated with the immune system. For instance, some biological processes are related to neutrophil and macroautophagy (26). Neutrophils are key components for early innate immunity. Blood samples from AD patients with dementia revealed that the neutrophil hyperactivation was associated with increased reactive oxygen species production as well as the levels of intravascular neutrophil extravascular traps. Moreover, neutrophil phenotype may have a close relationship with the rate of cognitive decline (26).
The cluster-2 significantly enriched pathways mainly consisted of neuronal signaling-related pathways, such as endocytosis and synaptic vesicle cycle. For instance, the synaptic vesicle cycle plays an important role in the biological process of exocytosis and endocytosis. It facilitates a series of events achieving chemical neurotransmission between functionally related neurons. Some study results demonstrated that considerable changes in the expression and functions of presynaptic proteins attributed in parts to direct effects of amyloid-β production and toxicity on the synaptic vesicle cycle (27). In addition, endocytosis is critical for the normal processing of APP, which is central to AD pathogenesis (28).
The most remarkable GO biological process of cluster-2 is the regulation of vascular associated smooth muscle cell migration. The degenerated smooth muscle cells express increased amounts of amyloid β-precursor protein deposition in the medial layer of the cerebral vessel wall and produce Aβ peptide (29). And the low-density lipoprotein particle receptor catabolic process is another important biological process in cluster-2. This biological process results in the breakdown of a low-density lipoprotein particle receptor molecule, a macromolecule that undergoes combination with a neurotransmitter to initiate a change in cell function. The disorder in this biological process could impair the neurotransmitter-triggered signal transduction appearing in AD.

 

Discussion

AD is a neurodegenerative brain disease that yet has no available effective medications or supplemental treatment. Studies have shown that AD is a heterogeneous disease. In this article, we integrated two types of omics data (genetic polymorphism and gene expression profiling) of MCI patients to identify subtypes with biological and clinical significance by the SNF method. We performed SNF, the integrative clustering of multiple genomic data algorithms, to cluster MCI patients. Experimental studies were conducted on subtypes of MCI patients, and we showed that multi-omics data define subtypes characterized by biological and clinical significance.
We utilized the SNF method to identify MCI patient subtypes based on multi-omics characteristics (9). SNF has been used to cluster subtype of specific cancer patients, and satisfactory results have been achieved. After executing the SNF algorithm, we identified two MCI subtypes. By comparing clinical information between two subtypes of patients, we considered the changes in two years on AD cognitive scales (MMSE, CDR, and FAQ) and MRI images in atrophy areas based on VBM. We found that the molecular subtypes of MCI are remarkably different in clinical information. It is necessary to lay the foundation for the precision treatment of MCI patients.
To study the difference in the disease mechanism of cluster-1 and cluster-2, differential expressions of MCI cluster-1, cluster-2 mRNA compared with the cognitively normal samples were computed correspondingly. And the differential expression genes in cluster-1 are significantly more than that of cluster-2. We conjecture that the risk factors of AD in cluster-1 are more complicated. Subsequently, we applied the functional annotation tool of KEGG and GO in Enrichr for enrichment analysis based on these genes. In cluster-1 MCI patients, there are some microorganisms (such as gram-negative bacterium and herpes viruses (22)) that can escape immune responses. These microorganisms activated immune responses, such as microglia, to clear the toxic proteins and widespread remnants from dying cells. Furthermore, these remnants continue to spread new amyloid-β clusters causing inflammatory storms (21). Above is the reason that MCI in cluster-2 patients may have synaptic failure and degeneration conditions. For example, the reduction in synaptic vesicle proteins has been shown to have a strong association with the clinical symptoms of dementia (27). We speculated that it is the storm caused by inflammasomes in the brain that result in cluster-1 MCI patients to develop the disease more rapidly than cluster-2 patients. Also, the perturbations of many other pathways have associated with the cause of AD. For example, Moriguchi et al. proposed that AD may be brain diabetes, and insulin signaling pathway is an important pathway for causing AD (23). And perturbation of pathways such as protein processing in endoplasmic reticulum, inositol phosphate metabolism and fubiquitin mediated proteolysis pathways will contribute to the amyloid cascade, which closely related to senile plaques and thus causing AD (24). The cluster-2 significantly enriched pathways mainly consisted of neuronal signaling-related pathways, and some scholars considered AD as a synaptic dysfunction caused by diffusible oligomeric assemblies of the amyloid-β protein (27). Both cluster-1 and cluster-2 enriched KEGG pathways of significantly differentially expressed genes have the endocytosis pathway. Hence, we speculated that endocytosis is the basic molecular mechanism of AD.
SNP data and mRNA expression profiling collected from patients’ peripheral blood have the characteristics of non-invasiveness and cost-effectiveness markers to identify MCI subtypes for clinical application. Clinical decisions will most likely be dictated by the genetic characteristics of AD patients in the coming years. We believe our method can effectively identify the subtypes of MCI patients, and can be applied in clinical in the future. Tailoring our method based on individual genetic characteristics will help doctors and researchers develop better therapeutic strategies and save many of MCI patients from receiving unnecessary toxic therapy. Further study should take into account the factors that can influence gene expression. For example, some other pathologies, influencing the expression of certain genes, may be present in elderly MCI patients. It may have an impact on the subtyping of MCI patients.
Two experiments can illustrate the clinical relevance of our method. For the first experiment, the expression data of 44 AD patients at baseline from the ADNI dataset were downloaded. We performed a hierarchical clustering analysis of patients with AD and patients of the two subtypes of MCI based on expression data using a similarity measure in SNF. The results are shown in the following Figure S2. This figure clearly shows that most AD patients are clustered with MCI cluster-1 patients. For the other experiment, 27 patients with AD at baseline in the ADNI dataset were downloaded. We applied the label propagation algorithm to assign new patients to subtypes. The subtype labels of these MCI patients were listed in Table S2. To test the effectiveness and reliability of our method, three AD cognitive scales were also displayed in 24-month for two subtypes of AD patients. As is shown in Figure S3, cognitive decline in cluster-1 MCI patients tends to be more remarkable than that of cluster-2 over 24 months, which is similar to the MCI patients in the ADNI dataset.
Hence, we believe our method can effectively identify the subtypes of MCI patients, and can be applied in clinical in the future. We look forward to potential collaborations with doctors and experimental biologists. We hope that the subtyping of MCI patients predicted with our model, will demonstrate its medical and therapeutic meaning. Besides, different types of data share complementary information, which is robust to noise and data heterogeneity (9). In the future, other types of biological data, such as DNA methylation and miRNA expression, can be integrated to explore biological patterns related to identify MCI subtypes. And classifying MCI patients into meaningful subtypes may improve the forecasting performance to proposing a method for predicting the conversion from MCI to AD (30).

Availability of data and material: Data used in this study are available through the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.usc.edu).

Author Contributions: HTL, SXY were involved with conception, design, and interpretation of data. HTL and JSW were involved with data analysis. XS, YL and XZZ provided general overall supervision of the study. XS acquired funding. All authors contributed to drafting and critical revision of the manuscript and have given final approval of the version to be published.

Funding: This research was sponsored by the National Natural Science Foundation of China (61972084, 81830053) and the Key Research and Development Program of Jiangsu province of China (BE2016002-3).

Acknowledgements: Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://www.loni.ucla.edu/ADNI). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. ADNI investigators include (complete listing available at http://www.loni.ucla.edu/ADNI/Collaboration/ ADNI_Authorship_list.pdf).

Conflicts of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Ethical Standards: Data involved in this study came from the Alzheimer’s disease neuroimaging initiative (ADNI) database. All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the Helsinki Declaration of 1975.

SUPPLEMENTARY MATERIAL1

SUPPLEMENTARY MATERIAL2

SUPPLEMENTARY MATERIAL3

References

1. Hebert LE, Weuve J, Scherr PA, Evans DA. Alzheimer disease in the United States (2010–2050) estimated using the 2010 census. Neurology 2013;80:1778–1783
2. Bagyinszky E, Youn YC, An SSA, Kim S. The genetics of Alzheimer’s disease. Clin Interv Aging 2014;9:535–551
3. Murray ME, Graff-Radford NR, Ross OA, Petersen RC, Duara R, Dickson DW. Neuropathologically defined subtypes of Alzheimer’s disease with distinct clinical characteristics: a retrospective study. Lancet Neurol 2011;10:785–796
4. Park J-Y, Na HK, Kim S, et al. Robust Identification of Alzheimer’s Disease subtypes based on cortical atrophy patterns. Sci Rep 2017;7:43270
5. Qiang W, Yau W-M, Lu J-X, Collinge J, Tycko R. Structural variation in amyloid-β fibrils from Alzheimer’s disease clinical subtypes. Nature 2017;541:217–221
6. Mukherjee S, Mez J, Trittschuh EH, et al. Genetic data and cognitively defined late-onset Alzheimer’s disease subgroups. Molecular psychiatry 2018
7. Mitchell AJ, Shiri-Feshki M. Rate of progression of mild cognitive impairment to dementia–meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 2009;119:252–265
8. Weiner MW, Aisen PS, Jack CR, et al. The Alzheimer’s Disease Neuroimaging Initiative: Progress report and future plans. Alzheimers Dement 2010;6:202–11.e7
9. Wang B, Mezlini AM, Demir F, et al. Similarity network fusion for aggregating data types on a genomic scale. Nat Methods 2014;11:333–337
10. Zhu X, Ghahramani Z. Learning from Labeled and Unlabeled Data with Label Propagation. 2002
11. Purcell S, Neale B, Todd-Brown K, et al. PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genetics 2007;81:559–575
12. Howie B, Fuchsberger C, Stephens M, Marchini J, Abecasis GR. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nat Genet 2012;44:955–959
13. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 2003;31:e15
14. Ashburner J, Friston KJ. Voxel-based morphometry—the methods. Neuroimage 2000;11:805–821
15. Wright IC, McGuire PK, Poline JB, et al. A voxel-based method for the statistical analysis of gray and white matter density applied to schizophrenia. Neuroimage 1995;2:244–252
16. Ashburner J. A fast diffeomorphic image registration algorithm. Neuroimage 2007;38:95–113
17. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 2002;15:273–289
18. Ribas GC. The cerebral sulci and gyri. Neurosurg Focus 2010;28:E2
19. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47
20. Kuleshov MV, Jones MR, Rouillard AD, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 2016;44:W90-97
21. Abbott A. Is ‘friendly fire’ in the brain provoking Alzheimer’s disease? Nature 2018;556:426
22. Readhead B, Haure-Mirande J-V, Funk CC, et al. Multiscale Analysis of Independent Alzheimer’s Cohorts Finds Disruption of Molecular, Genetic, and Clinical Networks by Human Herpesvirus. Neuron 2018;99:64-82.e7
23. Moriguchi S, Ishizuka T, Yabuki Y, et al. Blockade of the KATP channel Kir6.2 by memantine represents a novel mechanism relevant to Alzheimer’s disease therapy. Molecular Psychiatry 2018;23:211–221
24. Dickson DW. Apoptotic mechanisms in Alzheimer neurofibrillary degeneration: cause or effect? J Clin Invest 2004;114:23–27
25. Theuns J, Van Broeckhoven C. Transcriptional regulation of Alzheimer’s disease genes: implications for susceptibility. Hum Mol Genet 2000;9:2383–2394
26. Dong Y, Lagarde J, Xicota L, et al. Neutrophil hyperactivation correlates with Alzheimer’s disease progression. Ann Neurol 2018;83:387–405
27. Ovsepian SV, O’Leary VB, Zaborszky L, Ntziachristos V, Dolly JO. Synaptic vesicle cycle and amyloid β: Biting the hand that feeds. Alzheimer’s & Dementia 2018;14:502–513
28. Karch CM, Goate AM. Alzheimer’s Disease Risk Genes and Mechanisms of Disease Pathogenesis. Biological Psychiatry 2015;77:43–51
29. Van Nostrand WE, Melchor J, Wagner M, Davis J. Cerebrovascular smooth muscle cell surface fibrillar A beta. Alteration of the proteolytic environment in the cerebral vessel wall. Ann N Y Acad Sci 2000;903:89–96
30. Vinutha N, Pattar S, Sharma S, Shenoy PD, Venugopal KR. A Machine Learning Framework for Assessment of Cognitive and Functional Impairments in Alzheimer’s Disease: Data Preprocessing and Analysis. J Prev Alzheimers Dis 2020;7:87–94

APPLICATION OF DIGITAL COGNITIVE BIOMARKERS FOR ALZHEIMER’S DISEASE: IDENTIFYING COGNITIVE PROCESS CHANGES AND IMPENDING COGNITIVE DECLINE

J.R. Bock1, J. Hara1,2, D. Fortier1, M.D. Lee3, R.C. Petersen4, W.R. Shankle1-3 The Alzheimer’s Disease Neuroimaging Initiative*

1. Medical Care Corporation, Newport Beach, USA; 2. Pickup Family Neurosciences Institute, Hoag Memorial Hospital Presbyterian, Newport Beach, USA; 3. Dept. of Cognitive Sciences, University of California, Irvine, USA; 4. Department of Neurology, Mayo Clinic, USA; *Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf

Corresponding Author: Junko Hara, Ph.D. Medical Care Corporation, 3900 W. Coast Hwy, Ste 310, Newport Beach, CA 92663, Phone: 949-478-7388, Email: junkoh@mccare.com

J Prev Alz Dis 2020;
Published online November 11, 2020, http://dx.doi.org/10.14283/jpad.2020.63

 


Abstract

Background: Recent Alzheimer’s disease (AD) trials have faced significant challenges to enroll pre-symptomatic or early stage AD subjects with biomarker positivity, minimal or no cognitive impairment, and likelihood to decline cognitively during a short trial period. Our previous study showed that digital cognitive biomarkers (DCB), generated by a hierarchical Bayesian cognitive process (HBCP) model, were able to distinguish groups of cognitively normal individuals with impending cognitive decline from those without. We generated DCBs using only baseline Auditory Verbal Learning Test’s wordlist memory (WLM) item response data from the Mayo Clinic Alzheimer’s Disease Patient Registry.
Objectives: To replicate our previous findings, using baseline ADAS-Cog WLM item response data from the Alzheimer’s Disease Neuroimaging Initiative, and compare DCBs to traditional approaches for scoring word-list memory tests.
Design: Classified decliner subjects (n = 61) as those who developed amnestic MCI or AD dementia within 3 years of normal baseline assessment and non-decliner (n = 442) as those who did not.
Measures: Evaluated the relative value of DCBs compared to traditional measures, using three analytic approaches to group differences: 1) logistic regression of summary scores per ADAS-Cog WLM task; 2) Bayesian modeling of summary scores; and 3) HBCP modeling to generate DCBs from item-level responses.
Results: The HBCP model produced posterior distributions of group differences, of which Bayes factor assessment identified three DCBs with notable group differences: Immediate Retrieval from Durable Storage, (BFds = 11.8, strong evidence); One-Shot Learning, (BFds = 4.5, moderate evidence); and Partial Learning (BFds = 2.9, weak evidence). In contrast, logistic regression of summary scores did not significantly discriminate between groups, and the Bayes factor assessment of modeled summary scores provided moderate evidence that the groups were equivalent (BFsd = 3.4, 3.1, 2.9, and 1.4, respectively).
Conclusions: This study demonstrated DCBs’ ability to distinguish , at baseline, between impending cognitive decline and non-decline groups where individuals in both groups were classified as cognitively normal. This validated findings from our previous study, demonstrating DCBs’ advantages over traditional approaches. This study warrants further refinement of the HBCP DCBs to predict impending cognitive decline in individuals and other factors associated with AD, such as physical biomarker load.

Key words: Wordlist memory test, digital cognitive biomarkers, preclinical Alzheimer’s disease, clinical trial, Bayesian modeling.


 

Introduction

The major socioeconomic and healthcare burdens imposed by Alzheimer’s disease (AD) have pushed the focus of clinical research dramatically toward prevention and treatment in pre-symptomatic stages (1, 2). This shift has been well-aligned with guidance from the FDA in support of earlier stage therapies and new measurement methodologies for establishing clinically meaningful effects of those therapies (3).
However, recent AD trials have faced significant challenges identifying and enrolling subjects who meet thresholds for AD biomarker positivity but who have not yet experienced notable cognitive deficits. Despite great efforts made to enroll subjects in pre-clinical or early stage AD, trial sponsors have seen screen failure rates as high as 80%, primarily driven by required biomarker thresholds in cognitively normal subjects, leading to significantly prolonged enrollment periods and increased costs (4-9).
This underscores an urgent need for better approaches to pragmatically and cost effectively identify subjects who: 1) are cognitive normal; 2) will decline cognitively within 1-3 years; and 3) are likely to have PET scan positive AD biomarkers. Having such predictive capabilities will accelerate enrollment of specific subjects and will also improve study design for potentially shorter trial durations. The present study focuses on predicting impending cognitive decline in cognitively normal subjects. Predicting positivity for AD biomarkers will be discussed in a future publication.
Many efforts have focused on developing more sensitive cognitive assessment tools (e.g., composite scoring), including PACC (10) and ADCOMS (11). While some recently validated assessment tools can outperform less sensitive tools developed to assess dementia severity, as a group they lack the capability to predict impending cognitive decline in cognitively normal subjects.
One such assessment approach could arise from the application of hierarchical Bayesian cognitive process (HBCP) models to item-response data from a wordlist memory (WLM) test. This approach can generate digital cognitive biomarkers (DCB) that correspond to underlying cognitive processes of encoding, storage, and retrieval into and from various states of learning and memory (Figure 1). Such underlying cognitive processes cannot be directly observed or measured, while DCBs can quantify these processes, providing insights into cognitive function that traditional assessment approaches cannot provide. The details of this HBCP model and the generation of DCBs have been previously discussed elsewhere (12).

Figure 1. Hierarchical Bayesian Cognitive Process Model

The HBCP model can quantify underlying cognitive processes that cannot be observed or measured using the traditional approaches, and provide significant insights into how each cognitive process is affected by different conditions. Parameter r corresponds to one-shot learning; a, partial leaning; v, consolidation; b, testing effect; t, immediate retrieval from transient storage; L1, immediate retrieval from durable storage; and L2, delayed retrieval from durable storage.

In our previous work, HBCP-generated DCBs demonstrated the ability to distinguish groups of individuals with impending cognitive decline from those who would remain cognitively normal, using baseline, item-response data from a WLM test. This study was conducted using Auditory-Verbal Learning Test (AVLT) item response data from the Mayo Clinic Alzheimer’s Disease Patient Registry (13). Subjects, including those with normal cognition at baseline who would progress to amnestic MCI and those who would progress to AD dementia, were compared to those who would not decline. Bayes factor assessment identified notable reductions in Immediate Retrieval from Durable Storage, L1 (BFds = 30.4), and Delayed Retrieval from Durable Storage, L2 (BFds > 100). This study also appeared to identify compensatory increases in One-shot Learning, r (BFds = 3.2); Partial Learning, a (BFds = 10.8); and Consolidation, v (BFds = 13.5). However, subsequent work with our HBCP model did not replicate this apparent compensatory effect (12).
The present study, using baseline data from a novel WLM dataset, was designed to replicate our previous findings of deficits in retrieval DCBs for a group of individuals with impending cognitive decline due to AD, and compared this outcome to those generated by traditional scoring approaches. Replicating these results supports the role of DCBs for accelerating the clinical trial recruitment process and could greatly benefit future decisions about clinical trial study designs.

Methods

We used baseline ADAS-Cog WLM item response data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI: www.adni-info.org) database (adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership with the primary goal to test whether serial magnetic resonance imaging, positron emission tomography, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment and early AD.
From the ADNI dataset, we classified non-decliner subjects (n = 442) as those whose diagnosis remained normal for 3 or more years after normal baseline assessment and decliner subjects (n = 61) as those who developed amnestic MCI or AD dementia within 3 years of normal baseline assessment. Table 1 shows sample characteristics.
Three analytic approaches were compared to demonstrate the relative value of DCBs.

Table 1. ADNI Sample Characteristics

 

Logistic Regression

Traditional summary scores per ADAS-Cog task were assessed for group differences. Logistic regression modeling was performed with individual subjects’ summary scores for immediate free recall tasks 1 through 3 and for the delayed free recall task, each included as predictors of impending cognitive decline as the outcome.

Bayesian Modeling

A Bayesian model of summary scores was assessed. Gaussian distributions were fitted to individuals’ number of items recalled, from 1 to 10, and uniform distributions for probability of 0 items recalled, on each free recall task.

HBCP Model

The HBCP model was applied to non-decliner and decliner group item response data, aggregated across subjects within each group. The model estimated DCBs across WLM items with the Batchelder multinomial processing tree model of memory for each item’s recall pattern across the four free recall tasks (12).

 

Results

Logistic regression of summary scores generated β coefficients (Table 2) that did not significantly discriminate between groups, either individually or across the test as a whole.

Table 2. Summary Score Logistic Regression Analysis

Note. IFR = Immediate Free Recall; DFR = Delayed Free Recall; OR = Odds Ratio. χ2(498) = 2.18, pseudo-R2 = -.005, p = .702.

Bayes Factor assessment of fitted Gaussian distributions to each free recall task by summary score measurement provided moderate evidence that the groups were measurably equivalent (BFsd = 3.4, 3.1, 2.9, and 1.4, respectively; Figure 2).

Posterior distributions of Bayesian-modeled summary scores and posterior distributions of mean differences across ADAS-Cog tasks are presented, with Savage-Dickey density ratio Bayes factors calculated for mean differences against prior distributions of no change.

The HBCP model produced posterior distributions of group differences (Figure 3). Bayes Factor assessment identified three DCBs with notable group differences: Immediate Retrieval from Durable Storage, L1 (BFds = 11.8, strong evidence), One-shot Learning, r (BFds = 4.5, moderate), and Partial Learning, a (BFds = 2.9, weak).

Figure 3. Bayes Factors for HBCP DCB Notable Group Difference Parameters

Posterior distributions of DCB mean differences are presented against prior distributions of no change for three parameters with notable group differences, along with Savage-Dickey density ratio Bayes factors.

 

Discussion

The present study validated our previous findings by demonstrating the HBCP DCBs’ ability to distinguish a group of cognitively normal individuals with impending cognitive decline from a group that would remain cognitively normal. This study also showed DCBs’ advantages over the traditional approach of summary score assessments and their applicability for detection of impending cognitive decline in asymptomatic AD patients.
The HBCP DCBs have an advantage over composite or summary score approaches because of their ability to measure and quantify underlying cognitive processes (Figure 1). Among these processes, only some are affected in the cognitively normal or pre-clinical stages of AD (14), and each is affected differently as the disease progresses (15). The HBCP model can also be applied to any existing, well-validated, WLM test protocol (e.g., AVLT, ADAS-Cog, MCI Screen), so DCBs can be generated on WLM data from past academic studies and clinical trials to examine which processes were improved by particular AD therapies (16) and which were not, even when traditional outcome measures identified no overall differences. This will provide novel insights into efficacy and trial design, potentially targeting different cognitive or disease processes.
This study warrants further development of the HBCP DCBs to predict impending cognitive decline at the individual level and to predict other factors associated with AD, such as the identification of stage progression, the accumulation of biomarkers, and the presence of other cognition-impairing conditions.

 

Funding: This study was supported by the National Institute on Aging of the National Institutes of Health under Award Number R44AG065126. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Acknowledgement: Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

Conflicts of interest: JRB, JH, DF, and WRS are employees of Medical Care Corporation.

Ethical standards: IRB exemption status was obtained for this study from WIRB.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

References

1. Cummings J, Lee G, Mortsdorf T, Ritter A, Zhong K. Alzheimer’s disease drug development pipeline: 2017. Alzheimers Dement (N Y). 2017;3(3):367-384.
2. Cummings J, Ritter A, Zhong K. Clinical Trials for Disease-Modifying Therapies in Alzheimer’s Disease: A Primer, Lessons Learned, and a Blueprint for the Future. J Alzheimers Dis. 2018;64(s1):S3-S22.
3. Food and Drug Administration. Early Alzheimer’s Disease: Developing Drugs for Treatment Guidance for Industry: DRAFT GUIDANCE. February 2018. https://www.fda.gov/media/110903/download
4. Grill JD. Recruiting to preclinical Alzheimer’s disease clinical trials through registries. Alzheimers Dement (N Y). 2017;3(2):205-212. PMID: 28439532.
5. Jansen WJ, Ossenkoppele R, Tijms BM, Fagan AM, Hansson O, Klunk WE, et al. Association of Cerebral Amyloid-β Aggregation With Cognitive Functioning in Persons Without Dementia. JAMA Psychiatry. 2018;75(1):84-95. PMID: 29188296.
6. Langbaum JB, Karlawish J, Roberts JS, Wood EM, Bradbury A, High N, Walsh TL, Gordon D, Aggarwal R, Davis P, Stowell C, Trisko L, Langlois CM, Reiman EM, Tariot PN. GeneMatch: A novel recruitment registry using at-home APOE genotyping to enhance referrals to Alzheimer’s prevention studies. Alzheimers Dement. 2019;15(4):515-524. PMID: 30772251.
7. Rafii MS, Aisen PS. Alzheimer’s Disease Clinical Trials: Moving Toward Successful Prevention. CNS Drugs. 2019;33(2):99–106. PMID: 30560544.
8. Wolz R, Schwarz AJ, Gray KR, Yu P, Hill DL; Alzheimer’s Disease Neuroimaging Initiative. Enrichment of clinical trials in MCI due to AD using markers of amyloid and neurodegeneration. Neurology. 2016;87(12):1235–1241. PMID: 27558378.
9. Vermunt L, Muniz-Terrera G, Ter Meulen L, Veal C, Blennow K, Campbell A, Carrié I, Delrieu J, Fauria K, Huesa Rodríguez G, Ingala S, Jenkins N, Molinuevo JL, Ousset PJ, Porteous D, Prins ND, Solomon A, Tom BD, Zetterberg H, Zwan M, Ritchie CW, Scheltens P, Luscan G, Brookes AJ, Visser PJ, IMI-EPAD collaborators. Prescreening for European Prevention of Alzheimer Dementia (EPAD) trial-ready cohort: impact of AD risk factors and recruitment settings. Alzheimers Res Ther. 2020;12(1):8. PMID: 31907067.
10. Donohue M.C., Sperling R.A., Salmon D.P., Rentz D.M., Raman R., Thomas R.G. AIBL, ADNI, ADCS, the preclinical Alzheimer cognitive composite: measuring amyloid-related decline. JAMA Neurol. 2014;71:961–970.
11. Wang J, Logovinsky V, Hendrix SB, Stanworth SH, Perdomo C, Xu L, et al. ADCOMS: a composite clinical outcome for prodromal Alzheimer’s disease trials. J Neurol Neurosurg Psychiatry. 2016;87(9):993-9.
12. Lee MD, Bock JR, Cushman I, Shankle WR. An application of multinomial processing tree models and Bayesian methods to understanding memory impairment. J. Math. Psych. 2020;95:102328.
13. Shankle WR, Hara J, Bock JR, Fortier D, Mangrola T, Lee MD, Alexander GE, Batchelder WH, Petersen RC, Kremers W. Using Graphical Hierarchical Bayesian Cognitive Process Models Applied to Common Memory Tests to Predict AD Pathology within Normal Subjects. 2018. Clinical Trial in Alzheimer’s Disease. Poster Presentation. Barcelona, November 2018.
14. Braak H, Feldengut S, Del Tredici K. [Pathogenesis and prevention of Alzheimer’s disease: when and in what way does the pathological process begin?]. Nervenarzt. 2013;84(4):477-482. Review. German. PMID: 23508204.
15. Halliday GM, McCann H. Human-based studies on alpha-synuclein deposition and relationship to Parkinson’s disease symptoms. Exp Neurol. 2008;209(1):12-21. PMID: 17706644.
16. Roy DS, Arons A, Mitchell TI, Pignatelli M, Ryan TJ, Tonegawa S. Memory Retrieval by Activating Engram Cells in Mouse Models of Early Alzheimer’s Disease. Nature. 2016;531(7595):508-12. doi: 10.1038/nature17172.

A SYSTEMATIC REVIEW ON THE FEASIBILITY OF SALIVARY BIOMARKERS FOR ALZHEIMER’S DISEASE

M. Bouftas

Corresponding Author: Mohamed Bouftas, 610 Purdue Mall, West Lafayette, IN 47907, USA, mbouf7@gmail.com

J Prev Alz Dis 2021;1(8):84-91
Published online October 25, 2020, http://dx.doi.org/10.14283/jpad.2020.57


Abstract

Early AD diagnosis is critical for ameliorating prognosis and treatment. The analysis of CSF biomarkers yields accurate results, but it necessitates a lumbar puncture procedure. Screening for peripheral biomarkers in saliva is advantageous since this medium is noninvasive and inexpensive to obtain. The objective of this systematic review is to analyze saliva biomarker studies which aim to diagnose AD. Titles, abstracts, and reference lists for publications from January 2004 to February 2020 were screened for by searching Google Scholar and PubMed. The inclusion criteria involved published studies that consisted of both AD and control groups. 88 studies were screened, and 20 publications fulfilled the inclusion criteria. These selected publications were scrutinized and included in this review. Aβ42, tau, certain metabolites, and oral microbiota might serve as reliable biomarkers for AD diagnosis. These results showcase the legitimate feasibility of proteomic, metabolomic, and microbiotic compounds in saliva for AD diagnostics in the near future. Supplemental studies must consider standardizing the analytical methods of measuring salivary biomarkers to establish coherence for the selection of valid AD biomarkers. Validation studies will require a large sample size of biomarker-diagnosed individuals for independent populations. This ensures accuracy and rigidity for receiver operating characteristic (ROC) curves that can be set for the most optimal salivary biomarkers in future clinical settings.

Key words: Alzheimer, biomarker, dementia, saliva.

Abbreviations: Aβ: Amyloid beta; Aβ40: Amyloid beta-40 protein; Aβ42: Amyloid beta-42 protein; AchE: Acetylcholinesterase; AD: Alzheimer’s Disease; aMCI: Amnestic mild cognitive impairment; ANS: Autonomic nervous system; APP: Amyloid Precursor Protein; CSF: Cerebrospinal fluid; EG-ISFET: Extended gate ion-sensitive field-effect transistor; ELISA: Enzyme-linked immunosorbent-type assay; FTD: Frontotemporal Dementia; LC-MS: Liquid chromatography-mass spectrometry; PD: Parkinson’s Disease; P-tau: Phosphorylated tau; ROC: Receiver Operating Characteristic; SIMOA: Single molecule array; T-tau: Total tau; UPLC-MS: Ultra performance liquid chromatography-mass spectrometry


 

Background

Alzheimer’s Disease (AD) accounts for roughly 70% of all cases pertaining to dementia (1). The characteristic hallmarks of AD include the presence of amyloid plaques and tau tangles, but the direct cause of AD is unclear. These compounds can be present many years before clinical symptoms become visible. The amyloid hypothesis suggests that the accumulation of amyloid aggregates serves as the primary catalyzer for AD pathogenesis, but a multitude of failed clinical trials attempting to “plaque-bust” have posed serious questions concerning the legitimacy of the amyloid hypothesis.
Eli Lilly developed Solanezumab, a monoclonal antibody that was investigated for its potency against soluble Aβ oligomers (2). 400 mg of this drug was given to patients with mild Alzheimer’s every four weeks. 1,057 patients were administered Solanezumab and 1,072 patients were administered a placebo. After undergoing phase III trials, Solanezumab was not significantly effective at slowing cognitive decline. Roche developed Gantenerumab, another monoclonal antibody investigated for its ability to clear Aβ clumps in people with familial AD. Fewer than 1% of the AD demographic consists of individuals with familial AD (3). The potency of Gantenerumab was also assessed along with Solanezumab as a phase 2/3 clinical trial by the Dominantly Inherited Alzheimer Network Trials Unit (DIAN-TU). A total of 194 participants participated in this DIAN-TU clinical trial, of whom 52 participants were administered Gantenerumab. Study subjects participated for an average of five years. After undergoing phase 2/3 trials, Gantenerumab did not prove to be significantly potent against cognitive decline (4).
The observed results of Solanezumab and Gantenerumab in these clinical trials is probable justification for revisiting the authenticity of the amyloid hypothesis, but a more fundamental observation must be made. Treatment may yield no significant effect if it is administered during the symptomatic phase of AD. Therefore, there is a pressing need to develop a simple and noninvasive test which can pinpoint the presence of AD in its presymptomatic phase so that treatment can more successfully mitigate its neurodegenerative effect.
Biomarkers in saliva are being explored as an alternative diagnostic approach. The autonomic nervous system (ANS) regulates the innervation of both CN VII (facial nerve) and CN IX (glossopharyngeal nerve), which triggers saliva secretion by salivary glands. AD compromises ANS function, which may be related to a modified salivary composition for individuals with AD (5). The ability to conduct an early diagnosis in a noninvasive manner remains elusive. Although an array of clinical approaches are implemented to determine the presence of AD, utilizing a biomarker-driven test can aid in establishing an early diagnosis. Viable biomarkers in CSF are used to detect AD, but this method is both costly and invasive.
Significantly different levels of amyloid and tau in blood have been reported for AD subjects in comparison to controls (6-8). Saliva also contains a plethora of biomarkers, as about 40% of diagnostic blood proteins are also found in saliva (9). It features many advantages for diagnostic purposes compared to other bodily fluids. Saliva is both cheap and easy to obtain, as it can be collected in a noninvasive manner. These advantages facilitate the sampling process of this medium, which is useful for rapid disease screening. The objective of this review is to provide an overview of the literature pertaining to the utility of saliva as a medium for analyzing biomarkers that are specifically associated with AD.

 

Methods

A standard protocol was implemented for selection of publications in this review. PubMed and Google Scholar were used to conduct a literature search for publications from January 2004 to February 2020. Searches were conducted with the following keywords: Alzheimer, biomarker, dementia, saliva. Titles, abstracts, and reference lists were used to select pertinent publications. Studies included in this systematic review are original publications that analyze potential salivary AD biomarker candidates. Each study consisted of saliva samples from both AD subjects and control subjects. Subject metrics such as age, gender, and sample size for both AD and control groups were considered, along with biomarker type, technique of biomarker quantification, and statistical analysis.

 

Results

Of the 88 screened studies, 20 were selected, scrutinized, and included in this review. Several studies determined the potential utility of Aβ42 and tau as salivary biomarkers, but other compounds including acetylcholinesterase, metabolites, and oral microbiota were investigated as well.

Acetylcholinesterase

Salivary acetylcholinesterase (AchE) levels were analyzed in two studies. Ellman’s colorimetric method was implemented for both of the studies. Bakhtiari et al. tested saliva samples from 15 AD subjects and 15 control subjects. Higher levels of AchE were reported, but statistical significance was not established. Sayer et al. tested saliva samples from 47 volunteers (22 AD cases, 14 AD nonresponder cases, and 11 control cases). They found an overall negative correlation between age and AchE levels, as the r-value was -0.768 (with p<0.001). It was also reported that AD subjects had 73% lower levels of AchE in comparison to control subjects (with p<0.005).

Figure 1. Publication Search and Selection Flowchart

 

Aβ42

Salivary Aβ42 levels were analyzed in six studies, and all of them utilized an enzyme-linked immunosorbent-type assay (ELISA) with the exception of two. Bermejo-Pareja et al. tested 126 saliva samples from both AD and control cases, in addition to 51 saliva samples from Parkinson’s patients. They concluded that salivary Aβ42 levels were significantly greater in patients suffering from mild to moderate AD, but not for patients with severe AD. Their results did not conclude a significant difference between Parkinson’s patients and controls. Lee et al. analyzed the expression of Aβ42 in both saliva and other tissues. 37 volunteers participated in the study, including 27 non-AD and 7 AD cases. They reported a mean of 22.06±0.41 pg/mL of salivary Aβ42 for the non-AD cases and a mean of 59.07±6.33 pg/mL for the AD cases. Furthermore, McGeer et al. analyzed Aβ42 levels in another study which included 30 AD cases, 89 high normal (at-risk) cases, and 148 low control cases. They reported a mean of 21.26±1.73 pg/mL for low control cases, 37.96±8.13 pg/mL for high control cases (with p<0.01), and 51.70±10.50 pg/mL for AD cases (with p<0.05). Kim et al. utilized an immunoassay containing nanobeads to detect salivary Aβ42 levels for 45 individuals (28 AD cases and 17 normal controls). Their results concluded higher levels of salivary Aβ42 for the AD cases vs. control cases, but their study did not have a p-value. Sabbagh et al. analyzed salivary Aβ42 levels from 15 AD patients and 7 normal controls. They reported a mean of 21.1 ±0.3 pg/mL for the normal controls and a mean of 51.7±1.6 pg/mL for AD cases, with p<0.001. Tvarijonaviciute et al. tested 152 saliva samples from both AD and control cases by using a multiplex assay. They reported 12% lower salivary Aβ42 levels for AD patients in comparison to control cases. A mean of 3.15±0.72 pg/mL was determined for AD subjects and a mean of 3.57±0.93 pg/mL was determined for control subjects.

Table 1. Summary of results and subjects involved in analyzing salivary AD biomarker levels

 

Metabolites

AD progression damages the autonomic nervous system, which maintains saliva secretion. Compromising saliva secretion may affect salivary metabolite composition. Levels of various salivary metabolites were analyzed in three studies, which totaled 310 AD subjects, 60 MCI subjects, and 288 healthy control subjects. Huan et al. used liquid chromatography mass spectrometry (LC-MS) to assess the following metabolites: alanylphenylalanine, aminobytyric acid + H2, amino-dihydroxybenzene, choline-cytidine, glucosyl-galactosyl-hydroxylysine * (H2O), histidylphenylalanine, methylguanosine, phenylalanylphenylalanine, phenylalanylproline, and urocanic acid. Their work featured two clinical studies to further confirm their findings. Between AD and control subjects, there was a significant difference for the following metabolites (with p<0.01): choline-cytidine, histidylphenylalanine, methylguanosine, phenylalanylphenylalanine, phenylalanylproline, and urocanic acid. Between AD and aMCI subjects, there was a significant difference for the following metabolites (with p<0.01): alanylphenylalanine, aminobytyric acid + H2, amino-dihydroxybenzene, glucosyl-galactosyl-hydroxylysine* (H2O), and phenylalanylproline. Liang et al. implemented ultraperformance liquid chromatography mass spectrometry (UPLC-MS) to assess the following metabolites: inosine, ornithine, phenyllactic acid, and spinganine-1-phosphate. They concluded significantly elevated levels of spinganine-1-phosphate and ornithine for AD subjects in comparison to control subjects and significantly lower levels of inosine for AD subjects in comparison to control subjects (with p<0.01). Marksteiner et al. utilized a mass spectrometry kit (AbsoluteIDQ® p150) to analyze endogenous metabolites. ANOVA was used to conduct statistical analysis, along with the Dunnett post hoc test (with p<0.05). Between AD and control subjects, they concluded a significant decrease of the following lipids: PCae C34:1-2, PCae C36:1-2-3, PCae C38:1c3, and PCae C40:2-3. For MCI subjects, there was also a reported decrease in PCae C36:1-2-3. However, among all groups in this study, there were no significant differences for the following lipids: diacyl-phosphatidylcholines, lyso-acyl-phosphatidylcholines, and sphingomyelins.

Salivary Microbiome

Miklossy observed the presence of spirochetes in 451 of 495 AD brains (10). Her research notes the association between these bacteria and AD. Spirochetes play an etiological role in the manifestation of diseases like Lyme disease and syphilis. Spirochetes are a part of the oral microbiome, and further studies suggest that the oral microbiome is involved in AD pathogenesis. There may be species within the oral microbiome that provide insight into dementia progression.
Species within the oral microbiome were investigated in two studies. Liu et al. assessed the abundance of oral microbiota with 39 AD subjects and 39 control subjects. This study also took into consideration the presence or absence of APOEε4 for each subject. Two techniques were implemented: 16S rRNA sequencing was used to examine oral microbiota and Sanger sequencing was conducted to genotype subjects as either APOEε4(+) or APOEε4(-). None of the bacterial species were found to accelerate AD progression in this study. AD subjects had significantly higher levels of Moraxella, Leptotrichia, and Sphaerochaeta in comparison to control subjects. However, AD subjects had significantly lower levels of Rothia compared to control subjects. APOEε4(+) subjects had significantly lower levels of Actinobacillus and Actinomyces, and APOEε4(-) subjects had significantly higher levels of Abiotrophia and Desulfomicrobium.
Bathini et al. used a bead-based immunoassay to measure the following bacteria: P. gingivalis, F. villosus, L. wadei, F. alocis, C. valvarum, and P. tannerae. Subjects were categorized in four groups: cognitively normal healthy (n=27), cognitively normal at risk (n=15), MCI (n=21), and AD (n=15). The linear discriminant analysis (LDA) conducted on the oral microbiome indicated an accuracy of 0.94 (95% CI: 0.92,0.95 w/p<0.001). Logistic regression analysis (LRA) showed that F. alocis and F. villosus best differentiate among cognitive normal healthy and cognitive normal at risk, F. villosus and L. wadei best differentiate among cognitive normal healthy and MCI, while F. villosus and P. tannerae best differentiate among cognitive normal healthy and AD. Cumulatively, all of the bacterial species within the oral microbiome provide the most optimal discriminating capacity with respect to the cognitive healthy normal group. Interestingly, the analysis of L. wadei coupled with the UPSIT smell identification test yields a strong discriminating capacity with respect to the cognitive healthy normal group.

Tau

Salivary tau levels were analyzed in three studies. Ashton et al. tested 213 saliva samples from both AD and control cases (53 AD and 160 healthy older controls), in addition to 68 saliva samples from individuals with aMCI. T-tau levels were analyzed in duplicate using the human total tau assay on the HD-1 SIMOA device. There was not a statistically significant relationship between salivary total-tau levels and age (p=0.190, r=0.080) or gender (female median: 9.6 ng/L, male median: 12.3 ng/L, p=0.872). Increased median t-tau levels in AD patients were observed between healthy older controls, aMCI subjects, and AD subjects (9.6 ng/L, 9.8 ng/L, and 12.2 ng/L respectively), but statistical significance was not established. Shi et al. utilized both mass spectrometry and ELISA to measure salivary tau. Mass spectrometry of whole samples was initially done to detect the presence of tau protein. Luminex ELISA assays were utilized to analyze p-tau and t-tau levels. Mann-Whitney U-tests were applied to assess statistical differences in tau levels between the AD and healthy control groups (21 AD cases and 38 control cases). Control subjects were consenting volunteers who also scored above 27 on the Mini-Mental State Examination (MMSE). In comparison to healthy controls, AD subjects had lower t-tau levels, as well as higher p-tau and p-tau/t-tau levels (p<0.05). However, these differences were determined to be statistically insignificant, and there was minimal difference of total protein levels of tau between the AD and control groups (p<0.05). Pekeles et al. obtained unstimulated saliva in order to analyze the p-tau/t-tau ratio at different phosphorylation sites. Tau-4 antibody measured t-tau levels, and antibodies binding to Thr181, Ser396, and Ser404 were utilized to quantify phosphorylation for sites T181, S396, and S404 respectively. The combined antibody Ser400/Thr403/Ser404 was used for the combined phosphorylation site S400/T403/T404. The Western Blot technique was then implemented to analyze p-tau and t-tau levels in saliva. 337 volunteers participated throughout the two clinical studies conducted by Pekeles et al., including 87 AD subjects and 167 control subjects (of which there were 91 normal elderly control subjects and 76 young normal controls). Their first study included 55 aMCI subjects as well, and their second study included 16 FTD subjects and an additional 12 neurological patients that did not suffer from dementia. Nonparametric tests such as the Shapiro-Wilk test and the Mann-Whitney U test were used to assess the expression of salivary tau at each respective phosphorylation site for the first round of the study. Their findings indicated a significantly higher p-tau/t-tau ratio at the S396 and S404 sites, as well as the combined S400/S404/T404 site for AD patients in comparison to the elderly control individuals. The second round of study (using the two-tailed Kruskal-Wallis statistical test) reported higher median p-tau/t-tau levels at site S396 for AD subjects versus those of normal elderly controls. However, Pekeles et al. reported no correlation between elevated salivary tau levels and both CSF tau and hippocampal volume. There was also significant variation for salivary tau levels in AD subjects, which may pose a limitation towards implementation of tau as a legitimate AD biomarker.

Trehalose and Lactoferrin

Both trehalose and lactoferrin levels in saliva were analyzed in two studies. Lau et al. utilized an extended gate ion-sensitive field-effect transistor (EG-ISFET) biosensor to analyze salivary trehalose. Trehalose is a salivary sugar which has shown to alter the metabolism of the Amyloid Precursor Protein (APP), meanwhile reducing the aggregation rate of amyloid (11). 60 saliva samples were tested, including 20 AD subjects, 20 PD subjects, and 20 control subjects. Higher salivary trehalose levels were found in the AD subjects, but statistical significance was not established. Carro et al. used an ELISA to detect salivary lactoferrin levels. It has been observed that pathogenic microbes could contribute to the development of AD (12). The presence and effect of antimicrobial peptides to counteract microbes involved in the pathophysiology of AD remain an underexplored area of research. Lactoferrin is a non-enzymatic antimicrobial peptide which is present in various bodily fluids, including saliva. The objective was to determine if decreased lactoferrin levels could serve as an indicator of AD. 365 individuals participated throughout the two clinical studies conducted by Carro et al., including 116 AD subjects, 59 aMCI subjects, and 131 control subjects. Their first study also included 59 aMCI subjects. Mass spectrometry was implemented to confirm that this protein could be detected in saliva before further experimentation. This study also analyzed salivary lactoferrin levels for aMCI and PD subjects. Carro et al. concluded (with p<0.001) significantly lower levels for both AD and aMCI subjects in comparison with the control subjects, but PD subjects had significantly higher levels in comparison with control subjects. 7.43 μg/mL was the established cutoff value between AD/MCI subjects and controls in this study.

Supplementary Biomarkers

Supplementary biomarkers were analyzed in two studies. Manni et al. tested 38 saliva samples from both AD and control cases to assess dim light melatonin onset (DLMO) and salivary melatonin levels in order to determine the circadian phase of subjects experiencing early AD. An in-home melatonin salivary test was implemented for this study. It was concluded (with p=0.028) that DLMO ensued later for AD subjects compared to control subjects. Consequently, melatonin levels in AD subjects were significantly lower for control subjects. Ralbovsky et al. utilized a genetic algorithm coupled with an artificial neural network to conduct Hyper-Raman spectroscopy on samples from 39 volunteers (11 AD, 18 MCI, 10 control). They were able to identify several regions within the Raman spectrum which successfully distinguished between the three groups investigated.

 

Discussion

This systematic review aims at providing a proper assessment on the literature addressing salivary AD biomarker candidates. The studies observing salivary AchE suggest that it may not serve as a reliable biomarker, despite overall decreased AchE with age (13). Many other biological factors play a role in affecting overall AchE levels in both the brain and saliva, but significantly lower salivary AchE levels might prove to serve as a potential method of determining a compromised cholinergic system (14). Salivary Aβ42 seems to be a reliable biomarker, as five studies in this review analyzing salivary Aβ42 detected significant differences between AD subjects and control subjects. One study reported no significant differences when analyzing other isoforms of Aβ42, including Aβ40 (15). Another study conducted multivariate analysis and noted that although there were differences in Aβ42 concentrations between AD and control subjects, the data was neither statistically significant nor correlated to the severity of AD (16).
The disaccharide trehalose was analyzed as well as a multitude of metabolites. There seems to be a correlation between the expression of trehalose and metabolism of the Amyloid Precursor Protein (APP) (11). There were no significant differences in levels of trehalose, but there were significant differences in levels of various metabolites between AD subjects and controls (17-19). Research observing the relationship between AD and the salivary microbiome has been recent. Both studies reported different salivary microbiome compositions among AD individuals and controls. One study noted that the microbiotic composition begins to change for those who are classified as cognitively normal but are at-risk for acquiring AD (20). Two studies concluded significantly higher levels of p-tau/t-tau for AD subjects (21, 22), but one of these studies reported great variance in their data (22). Statistical insignificance in salivary tau levels was determined in one study (23). It was also reported that salivary tau expression was well characterized at the S396 phosphorylation site (22).
Carro et al.’s results show some validation of lactoferrin as a potential biomarker. Lactoferrin is present in several biological fluids and serves as part of the innate immune system. Some studies have suggested that certain pathogens may play a role in AD by compromising the function of the blood-brain barrier, thus enabling accelerated Aβ42 growth. This may justify the reason for lower lactoferrin levels for individuals with AD, but further studies are needed to confirm this. Manni et al. noted that individuals with mild to moderate AD experienced delayed melatonin secretion, thus explaining significantly lower salivary melatonin levels. The machine learning model incorporated by Ralbovsky et al. sustained very high accuracy, sensitivity, and specificity averages. It was able to successfully identify a multitude of strong differentiating salivary biomarkers among AD, MCI, and control groups.
Several analysis techniques were implemented throughout these studies, so a standardization by which to investigate salivary biomarkers would provide a more coherent method of selecting future AD biomarkers. Many of these clinical studies featured a small sample size, so a large sample needs to be incorporated for future studies in order to establish reliable reference ranges for biomarker expression levels. Saliva production, circadian rhythms, and oral health are important factors which affect saliva composition. This necessitates further research into how these factors may affect the accuracy of saliva as a medium for AD diagnosis. The precise mechanisms by which these biomarkers become secreted in saliva is not understood. There is still a need to acquire insightful knowledge to explain the presence of these biomarkers in saliva. Advancing the understanding of the pathophysiology of AD requires a thorough comprehension of the association between saliva and AD.

 

Conclusions

This systematic review intends to determine the feasibility of various salivary biomarkers in order to achieve an early diagnosis of AD. Subject metrics, biomarker type, and methods of biomarker analysis were examined to establish a solid answer on the viability of a saliva test. The reported data indicates that certain salivary compounds may serve as valid AD biomarkers, but a large sample size and a standardization of biomarker analysis techniques must be implemented to further assess the reproducibility of the studies included in this systematic review. More studies featuring large biomarker-diagnosed populations can further validate the use of other potential salivary biomarkers in independent populations. Establishing accurate ROC curves for each respective salivary AD biomarker is necessary to shift towards a biomarker-based AD diagnosis.

Ethics Approval and Consent to Participate: Not applicable.

Consent for Publication: Not applicable.

Availability of Data and Materials: All analyzed and generated data in this study, as well as supplemental information are included in this manuscript.

Competing Interests: The author declares no potential conflicts of interest.

Funding: Not applicable.

Authors’ Contributions: MB conducted the literature search, extracted the data, selected the studies for inclusion, analyzed the data, and wrote the manuscript.
Acknowledgements: Not applicable.

References

1. Kametani, F., & Hasegawa, M. Reconsideration of Amyloid Hypothesis and Tau Hypothesis in Alzheimer’s Disease. Frontiers in neuroscience, 2018;12, 25. https://doi.org/10.3389/fnins.2018.00025
2. Honig, L. S., Vellas, B., Woodward, M., Boada, M., Bullock, R., Borrie, M., Siemers, E. Trial of Solanezumab for Mild Dementia Due to Alzheimer’s Disease. New England Journal of Medicine, 2018;378(4), 321–330. doi: 10.1056/nejmoa1705971
3. National Institute on Aging. What Causes Alzheimer’s Disease. Available from: https://www.nia.nih.gov/health/what-causes-alzheimers-disease.
4. “Roche provides topline results from investigator-led Phase II/III trial with gantenerumab in rare inherited form of Alzheimer’s disease,” February 2020, https://www.roche.com/media/releases/med-cor-2020-02-10.htm
5. Farah, R., Haraty, H., Salame, Z., Fares, Y., Ojcius, D. M., & Said Sadier, N. Salivary biomarkers for the diagnosis and monitoring of neurological diseases. Biomedical journal, 2018;41(2), 63–87. doi:10.1016/j.bj.2018.03.004
6. Nakamura A, Kaneko N, Villemagne VL, et al. High performance plasma amyloid-β biomarkers for Alzheimer’s disease. Nature. 2018;554(7691):249-254. doi:10.1038/nature25456
7. Schindler SE, Bollinger JG, Ovod V, et al. High-precision plasma β-amyloid 42/40 predicts current and future brain amyloidosis. Neurology. 2019;93(17):e1647-e1659. doi:10.1212/WNL.0000000000008081
8. Thijssen EH, La Joie R, Wolf A, et al. Diagnostic value of plasma phosphorylated tau181 in Alzheimer’s disease and frontotemporal lobar degeneration. Nat Med. 2020;26(3):387-397. doi:10.1038/s41591-020-0762-2
9. Loo, J. A., Yan, W., Ramachandran, P., & Wong, D. T. Comparative human salivary and plasma proteomes. Journal of dental research, 2010;89(10), 1016–1023. doi:10.1177/0022034510380414
10. Miklossy, J. Alzheimer’s disease – a neurospirochetosis. Analysis of the evidence following Koch’s and Hill’s criteria. J Neuroinflammation 2011;8, 90. https://doi.org/10.1186/1742-2094-8-90
11. Tien, N. T., Karaca, I., Tamboli, I. Y., & Walter, J. Trehalose Alters Subcellular Trafficking and the Metabolism of the Alzheimer-associated Amyloid Precursor Protein. The Journal of biological chemistry, 2016;291(20), 10528–10540. doi:10.1074/jbc.M116.719286
12. Hill J.M., Clement C., Pogue A.I., Bhattacharjee S., Zhao Y., Lukiw W.J. Pathogenic microbes, the microbiome, and Alzheimer’s disease (AD) Front Aging Neurosci. 2014;6:127.
13. Sayer, R., Law, E., Connelly, P. J., & Breen, K. C. Association of a salivary acetylcholinesterase with Alzheimer’s disease and response to cholinesterase inhibitors. Clinical Biochemistry, 2004;37(2), 98–104. doi: 10.1016/j.clinbiochem.2003.10.007
14. Bakhtiari, S., Moghadam, N. B., Ehsani, M., Mortazavi, H., Sabour, S., & Bakhshi, M. Can Salivary Acetylcholinesterase be a Diagnostic Biomarker for Alzheimer?. Journal of clinical and diagnostic research : JCDR, 2017;11(1), ZC58–ZC60. doi:10.7860/JCDR/2017/21715.9192
15. Bermejo-Pareja, F., Antequera, D., Vargas, T., Molina, J. A., & Carro, E. Saliva levels of Abeta1-42 as potential biomarker of Alzheimer’s disease: a pilot study. BMC neurology, 2010;10, 108. doi:10.1186/1471-2377-10-108
16. Tvarijonaviciute, A., Zamora, C., Cerón, J.J., Bravo-Cantero, A.F., Pardo-Marín, L., Valverde, S., & López-Jornet, P. Salivary biomarkers in Alzheimer’s disease. Clinical Oral Investigations, 2020;1 – 8.
17. T. Huan, T. Tran, J. Zheng et al., “Metabolomics analyses of saliva detect novel biomarkers of Alzheimer’s disease,” Journal of Alzheimer’s Disease, vol. 65, no. 4, pp. 1401–1416, 2018.
18. Q. Liang, H. Liu, T. Zhang, Y. Jiang, H. Xing, and A. H. Zhang, “Metabolomics-based screening of salivary biomarkers for early diagnosis of Alzheimer’s disease,” RSC Advances, vol. 5,no. 116, pp. 96074–96079, 2015.
19. Marksteiner, J., Oberacher, H., & Humpel, C. Acyl-Alkyl-Phosphatidlycholines are Decreased in Saliva of Patients with Alzheimer’s Disease as Identified by Targeted Metabolomics. Journal of Alzheimers Disease, 2019;68(2), 583–589. doi: 10.3233/jad-181278
20. Bathini, P, Foucras, S, Dupanloup, I, et al. Classifying dementia progression using microbial profiling of saliva. Alzheimer’s Dement. 2020; 12:e12000. https://doi.org/10.1002/dad2.12000
21. M. Shi, Y. T. Sui, E. R. Peskind et al., “Salivary tau species are potential biomarkers of Alzheimer’s disease,” Journal of Alzheimer’s Disease, vol. 27, no. 2, pp. 299–305, 2011.
22. H. Pekeles, H. Y. Qureshi, H. K. Paudel, H. M. Schipper, M. Gornistky, and H. Chertkow, “Development and validation of a salivary tau biomarker in Alzheimer’s disease,” Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring, vol. 11, pp. 53–60, 2018.
23. N. J. Ashton, M. Ide, M. Schöll et al., “No association of salivary total tau concentration with Alzheimer’s disease,” Neurobiology of Aging, vol. 70, pp. 125–127, 2018.
24. M. Lee, J. P. Guo, K. Kennedy, E. G. McGeer, and P. L. McGeer, “A method for diagnosing Alzheimer’s disease based on salivary amyloid-β protein 42 levels,” Journal of Alzheimer’s Disease, vol. 55, no. 3, pp. 1175–1182, 2017.
25. McGeer, P. L., Lee, M., Kennedy, K., & McGeer, E. G. Saliva Diagnosis as a Disease Predictor. Journal of clinical medicine, 2020;9(2), 377. https://doi.org/10.3390/jcm9020377
26. C.-B. Kim, Y. Y. Choi, W. K. Song, and K. B. Song, “Antibody based magnetic nanoparticle immunoassay for quantification of Alzheimer’s disease pathogenic factor,” Journal of Biomedical Optics, vol. 19, no. 5, article 051205, 2014.
27. M. N. Sabbagh, J. Shi, M. Lee et al., “Salivary beta amyloid protein levels are detectable and differentiate patients with Alzheimer’s disease dementia from normal controls: preliminary findings,” BMC Neurology, vol. 18, no. 1, p. 155, 2018.
28. Liu XX, Jiao B, Liao XX, et al. Analysis of Salivary Microbiome in Patients with Alzheimer’s Disease. J Alzheimers Dis. 2019;72(2):633-640. doi:10.3233/JAD-190587
29. H.-C. Lau, I. K. Lee, P. W. Ko et al., “Non-invasive screening for Alzheimer’s disease by sensing salivary sugar using Drosophila cells expressing gustatory receptor (Gr5a) immobilized on an extended gate ion-sensitive field-effect transistor (EGISFET) biosensor,” PLoS One, vol. 10, no. 2, article e0117810, 2015.
30. Hill J.M., Clement C., Pogue A.I., Bhattacharjee S., Zhao Y., Lukiw W.J. Pathogenic microbes, the microbiome, and Alzheimer’s disease (AD) Front Aging Neurosci. 2014;6:127.
31. E. Carro, F. Bartolomé, F. Bermejo-Pareja et al., “Early diagnosis of mild cognitive impairment and Alzheimer’s disease based on salivary lactoferrin,” Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring, vol. 8, pp. 131–138, 2017.
32. Manni R, Cremascoli R, Perretti C, et al. Evening melatonin timing secretion in real life conditions in patients with Alzheimer disease of mild to moderate severity. Sleep Med. 2019;63:122-126. doi:10.1016/j.sleep.2019.04.018
33. Ralbovsky NM, Halámková L, Wall K, Anderson-Hanley C, Lednev IK. Screening for Alzheimer’s Disease Using Saliva: A New Approach Based on Machine Learning and Raman Hyperspectroscopy. J Alzheimers Dis. 2019;71(4):1351-1359. doi:10.3233/JAD-190675
34. “Biogen/Eisai Halt Phase 3 Aducanumab Trials | ALZFORUM,” April 2019, https://www.alzforum.org/news/ research-news/biogeneisai-halt-phase-3-aducanumab-trials.
35. Blennow K. A Review of Fluid Biomarkers for Alzheimer’s Disease: Moving from CSF to Blood. Neurology and therapy, 2017;6(Suppl 1), 15–24. doi:10.1007/s40120-017-0073-9
36. Dementia. (n.d.). Retrieved from https://www.who.int/news-room/fact-sheets/detail/dementia.
37. Prvulovic D, Hampel H. Amyloid beta (Abeta) and phospho-tau (ptau) as diagnostic biomarkers in Alzheimer’s disease. Clin Chem Lab Med 2011;49(3): 367-374.
38. Reale, M., Gonzales-Portillo, I., & Borlongan, C. V. Saliva, an easily accessible fluid as diagnostic tool and potent stem cell source for Alzheimer’s Disease: present and future applications. Brain Research, 2019;146535. doi: 10.1016/j.brainres.2019.146535
39. Serrano-Pozo, A., Frosch, M. P., Masliah, E., & Hyman, B. T. Neuropathological alterations in Alzheimer disease. Cold Spring Harbor perspectives in medicine, 2011;1(1), a006189. doi:10.1101/cshperspect.a006189
40. Yilmaz, A., Geddes, T., Han, B., Bahado-Singh, R. O., Wilson, G. D., Imam, K., … Graham, S. F. Diagnostic Biomarkers of Alzheimer’s Disease as Identified in Saliva using 1H NMR-Based Metabolomics. Journal of Alzheimers Disease, 2017;58(2), 355–359. doi: 10.3233/jad-161226
41. Zhang CZ, Cheng XQ, Li JY, Zhang P, Yi P, Xu X, et al. Saliva in the diagnosis of diseases. Int J Oral Sci 2016;8(3): 133-37.

ESTABLISHING A TRIAL READY COHORT TO ACCELERATE ALZHEIMER’S CLINICAL TRIAL ENROLLMENT AND TREATMENTS

E.A. Meyers, M.C. Carrillo

Alzheimer’s Association, Chicago, IL USA

Corresponding Author: Maria C. Carrillo PhD, Medical & Scientific Relations, Alzheimer’s Association, 225 N. Michigan Ave. Suite 1800, Chicago, IL 60601, P:312.335.5722, mcarrillo@alz.org

J Prev Alz Dis 2020;4(7):202-203
Published online August 11, 2020, http://dx.doi.org/10.14283/jpad.2020.42


Early detection is critical in our fight to stop or slow Alzheimer’s dementia, and even more so to prevent Alzheimer’s disease (AD). Current diagnosis of Alzheimer’s dementia relies largely on documenting mental decline, at which point, severe cognitive and functional damage has occurred. According to the National Institute on Aging and Alzheimer’s Association Research Framework, Alzheimer’s disease is defined by its underlying pathologic processes that can be documented by postmortem examination or in vivo by biomarkers. The diagnosis is not based on the clinical consequences of the disease (i.e., symptoms/signs) in this research framework, which shifts the definition of AD in living people from a syndromal to a biological construct (1). The “Framework” is based on research that confirms Alzheimer’s disease pathologic changes in the brain begin 15-20 years before the development of symptoms (2). The neuropathologic hallmarks of AD include: amyloid plaques, neurofibrillary tangles (NFTs), Glial responses, and synaptic and neuronal loss. This approach enables a more precise approach to interventional trials where specific pathways can be targeted in the disease process and in the appropriate people. It is hypothesized that during the preclinical period, 10-15 years prior to severe symptoms where fibrillar brain amyloid increases with minimal impact on cognition, that disease-modifying therapy can be most effective (3).
The Anti-Amyloid treatment in Asymptomatic Alzheimer’s (A4) Study is examining the effectiveness of solanezumab, a drug targeting beta-amyloid, in over 1,000 symptom-free volunteers whose positron emission tomography (PET) scans show abnormally high levels of beta-amyloid in the brain (4). Enrollment to the A4 study has a 71% screen fail rate for participants, highlighting the difficulties of high cost and patient burden in recruiting eligible participants. Nevertheless, the A4 study demonstrates feasibility as well as the challenges of an early, large, intervention trial to treat AD.
Recruiting individuals in the preclinical stage is especially challenging because the mild nature of symptoms means that individuals do not seek medical care for memory decline. To address this issue, a large number of cognitively normal individuals must be screened with a lengthy and expensive process (including education, behavioral assessment prior to scanning, then scanning and disclosure) in order to fully enroll a prevention trial.
The Trial-Ready Cohort for Preclinical/Prodromal Alzheimer’s disease (TRC-PAD) was established to accelerate enrollment of high risk individuals into early stage AD clinical trials (5). TRC-PAD consists of three main elements: the Alzheimer’s Prevention Trial (APT) Webstudy, the Site Referral System (SRS) and the Trial Ready Cohort (TRC).
It is projected that 25,000-50,000 participants are needed in the APT Webstudy in order to identify enough eligible participants for the TRC. In addition to local recruitment efforts (media and outreach), TRC-PAD is the first program to leverage existing national registries in order to invite individuals to participate in the APT Webstudy. A referral strategy has been established through partnerships with four national registries, the Alzheimer’s Prevention Registry (APR), Alzheimer’s Association TrialMatch, the Cleveland Clinic Healthy Brains Registry and the UCI Consent to Contact (C2C) Registry. Currently, these registries are the primary source of participants into the APT Webstudy, on an average of 1,514 per month (6). Successes in recruitment through national registries is in part because participants in registries have already demonstrated an interest in research.
The APT Webstudy obtains demographic, medical and lifestyle information in addition to tracking cognitive performance on a quarterly basis using remote cognitive and functional assessments. To address issues of retention, the APT Webstudy was developed as a user-friendly interface, requiring minimal time commitment, and users are well supported to quickly address any problems that surface. Participant engagement is optimized by allowing users to access results from assessments, receive reminder emails to complete tasks and subscribe to a quarterly newsletter.
Data obtained by the APT Webstudy is used to assess an individual’s risk for amyloid elevation in the brain. An adaptive algorithm predicts amyloid positivity in participants and identifies individuals that meet criteria to attend in-clinic visits for additional screening for the TRC (7). Machine learning techniques were refined using pre-randomized data of n=4,486 from the A4 study to derive the first cross-sectional predictive models. Variables used in this analysis include demographics, cognitive and functional assessments, and apolipoprotein E (APOE) genotype. This algorithm dramatically improves the accuracy in predicting amyloid positivity, with the greatest improvement when APOE genotype was known. This is intended to reduce burden to trial participants and high cost of screening.
Participants identified by the algorithm as having relatively high risk for amyloid elevation in the brain and are geographically located near TRC-PAD clinical sites are presented to clinical site teams through the SRS. A list of potential participants is provided to TRC sites on a monthly basis and final selections are reviewed manually, with the expectation that this will become increasingly automated. At the in-person visit, additional cognitive assessments are performed, as well as performing APOE genetic testing. With this additional data, participants’ risk assessment is updated prior to screening for amyloid burden, either by PET or Cerebrospinal Fluid (CSF) collection. Participants in the APT Webstudy that are not geographically located near a TRC site are provided with the opportunity to obtain a report containing their performance on various assessments as well as an explanation of the assessments that they can review with their healthcare provider.
Design of the informatics architecture for TRC-PAD will allow for longitudinal data from the TRC to inform therapeutic trials and will integrate with the Alzheimer’s Treatment Research Institute/Alzheimer’s Clinical Trial Consortium (ATRI/ACTC).
The APT Webstudy was officially launched in December 2017 and has consented 30,554 participants with 25 individuals enrolled in the TRC. While exceeding original expectations, the APT Webstudy has not been successful in attracting a diverse group of participants that is representative of the US population (6). Recruitment strategies are being implemented to address this deficiency by creating Spanish language study materials and other community-based approaches.
Continued refinement of the risk assessment will focus on utilizing longitudinal cognitive and functional changes as well as the use of blood-based biomarkers to improve performance of these predictive models (8).
A lot of work remains to be done towards the prevention of Alzheimer’s disease and ultimately Alzheimer’s dementia. Focusing on the preclinical stage of the disease, where there is low amyloid burden in the brain, may give potential treatments an advantage over the disease. TRC-PAD has developed a framework to enroll a large number of individuals at the early stages of disease, to provide longitudinal cognitive assessments and to predict the elevation of amyloid in the brain. Establishing a cohort of high-risk participants for enrollment into interventional clinical trials ought to fast track the critical treatments needed most. The proof will ultimately be in the ability for TRC-PAD to create a more efficient path towards enrollment into clinical trials, of a population that is in the earliest stages of Alzheimer’s disease, and representative of the population in the United States.

 

Conflict of interests: The author declares there are no conflicts.

 

References

1. Jack CR, Jr, Bennett DA, Blennow K, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018;14(4):535-562.
2. Villemagne VL, Burnham S, Bourgeat P, et al. Amyloid β deposition, neurodegeneration, and cognitive decline in sporadic Alzheimer’s disease: a prospective cohort study. Lancet Neurol. 2013; 12(4):357-367.
3. Jack CR, Jr., Albert MS, Knopman DS, et al. Introduction to the recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7(3):257-262.
4. Sperling RA, Donohue MC, Raman R, et al. Association of Factors With Elevated Amyloid Burden in Clinically Normal Older Individuals. JAMA Neurol. 2020;77(6):1-11
5. Walter S, Langford O, Clanton T, et al. The Trial-Ready Cohort for Preclinical/Prodromal Alzheimer’s disease (TRC-PAD): Experience from the first 3 years. J Prev Alzheimers Dis. 2020; DOI: 10.14283/jpad.2020.47.
6. Walter S, Clanton T, Langford O, et al. Recruitment into the Alzheimer Prevention Trials (APT) Webstudy for a Trial-Ready Cohort for Preclinical and Prodromal Alzheimer’s Disease (TRC-PAD). J Prev Alzheimers Dis. 2020; DOI: 10.14283/jpad.2020.46.
7. Langford O, Raman R, Sperling RA, et al. Predicting amyloid burden to accelerate recruitment of secondary prevention clinical trials. J Prev Alzheimers Dis. 2020; DOI: 10.14283/jpad.2020.44.
8. PS Aisen, RA Sperling, J Cummings, et al. The Trial-Ready Cohort for Preclinical/prodromal Alzheimer’s Disease (TRC-PAD) Project: An Overview. J Prev Alzheimers Dis. DOI: 10.14283/jpad.2020.45.

What do you want to do ?

New mail

NEUROPSYCHOLOGICAL, PSYCHIATRIC, AND FUNCTIONAL CORRELATES OF CLINICAL TRIAL ENROLLMENT

 

D.B. Hammers1, N.L. Foster1, J.M. Hoffman2, T.H. Greene3, K. Duff1

 

1. Center for Alzheimer’s Care, Imaging, and Research, Department of Neurology, University of Utah, Salt Lake City, UT, USA; 2. Center for Quantitative Cancer Imaging, Huntsman Cancer Institute, University of Utah, Salt Lake City, UT, USA; 3. Study Design and Biostatistics Center, University of Utah, Salt Lake City, UT, USA

Corresponding Author: Dustin B. Hammers, PhD, ABPP(CN), Center for Alzheimer’s Care, Imaging and Research, University of Utah, Department of Neurology, 650 Komas Drive #106-A, Salt Lake City, UT 84108, Tel: 801-585-6546. Fax: 801-581-2483. E-mail: dustin.hammers@hsc.utah.edu

J Prev Alz Dis 2019;
Published online September 19, 2019, http://dx.doi.org/10.14283/jpad.2019.38

 


Abstract

Screen failure rates in Alzheimer’s disease (AD) clinical trial research are unsustainable, with participant recruitment being a top barrier to AD research progress. The purpose of this project was to understand the neuropsychological, psychiatric, and functional features of individuals who failed screening measures for AD trials. Previously collected clinical data from 38 patients (aged 50-83) screened for a specific industry-sponsored clinical trial of MCI/early AD (Biogen 221AD302, [EMERGE]) were analyzed to identify predictors of AD trial screen pass/fail status. Worse performance on non-memory cognitive domains like crystalized knowledge, executive functioning, and attention, and higher self-reported anxiety, was associated with failing the screening visit for the EMERGE AD clinical trial, whereas we were not able to detect a relationship between screening status and memory performance, self-reported depression, or self-reported daily functioning. By identifying predictors of AD trial screen passing/failure, this research may influence decision-making about which patients are most likely to successfully enroll in a trial, thereby potentially lowering participant burden, maximizing study resources, and reducing costs.

Key words: Cognition, Alzheimer’s disease, mild cognitive impairment, clinical trial.


 

Difficulty with participant recruitment is considered one of the top barriers to Alzheimer’s disease (AD) clinical research progress (1). Many barriers exist to successful recruitment, including patient comorbidity, limited availability of studies and logistical issues, and eligibility criteria (2, 3). While most clinical drug trials targeting AD currently require recruitment of patients with mild dementia severity or Mild Cognitive Impairment (MCI), only 20 to 25% of patients diagnosed with AD are eligible for AD clinical trials (2), due to factors like medical comorbidities and medications (3), or the lack of an adequate study partner (4). More specifically, screen failure rates are strikingly high, as roughly ten patients may need to be screened to enroll one participant (5). These screen failure issues appear to be widespread across the industry (2), and are problematic for drug study sponsors, clinical trial sites, and participants themselves. Failed screening visits represent wasted time and lost revenues for both sponsors and sites, add to existing logistical and scheduling challenges, and extend the timelines to reach recruitment quotas (6). Additionally, high screen failure rates amplify participants’ perception of AD drug trial inaccessibility and dampen participant interest (7). Consequently, the current trial recruitment strategy is not optimally prepared to take on the National Plan to Address Alzheimer’s Disease’s ambitious goal of preventing and treating AD by 2025.
To begin to address potential solutions to the ever-present recruitment shortage, AD-related programs and task forces have focused on patient registries, raising participant awareness, site performance and funding, and reducing barriers to participation (5). Unfortunately, there is little to no emphasis on reducing screen failure rates based on study inclusion criteria for the current sources of participants already being recruited. Disease severity is a common cause of participant screen failure (8), which is typically measured by cognitive test performance and/or informant/participant rating scales of functioning. However, there tends to be minimal overlap between measures of disease severity used in clinical trials and those used in typical clinic settings, and judgment about severity in a clinical trial is often based upon multiple metrics that may be discrepant. As a result, for patients recruited through referral from clinic, physicians and study teams have limited capacity to predict who will meet inclusion criteria for disease severity prior to the AD trial screening visit, which contributes to higher screen failure rates.
To help address the current limitations of recruitment strategies in AD clinical trials, the current study seeks to better understand the neuropsychological, psychiatric, and functional features of individuals who pass/fail screening measures for AD trials, using previously collected data from patients enrolled both in a cognitive specialty clinic and an AD clinical trial. We hypothesized that worse performance on memory, other cognitive functioning domains, and psychiatric measures during the clinical evaluation would be associated with lower participant screen failure into an AD clinical trial, and that younger participants would be more likely to fail AD clinical trial screening measures. Similarly, we expected worse self-reported daily functioning would be associated with lower screen failure rates.

 

Methods

Sample and Study Design

The current study is a retrospective, cross-sectional analysis of the neuropsychological, psychiatric, and functional predictors of AD clinical trial enrollment. A database in the Division of Cognitive Neurology at a university in the western United States was searched for participants having (1) previously received a clinical diagnostic workup (including dementia-expert cognitive evaluation and diagnostic neuropsychological assessment) at the university’s transdisciplinary cognitive specialty clinic and subsequently diagnosed with either MCI or early AD and (2) previously screened for a specific industry-sponsored clinical trial of MCI/early AD (Biogen 221AD302 (9), Phase 3 Study of Aducanumab in Early Alzheimer’s Disease [EMERGE], which will be referred to as “the EMERGE trial” for the remainder of this manuscript). Thirty-eight participants met the inclusion criteria for this retrospective study, and no other inclusion/exclusion criteria were applied to the current study. Please see Figure 1 for a flow diagram of participants recruited for the EMERGE trial, with the 38 participants who “Screened” for the trial representing our current study’s sample population. As a result of the inclusion age for the EMERGE trial being between 50 and 85, only those aged participants were included in the current study. Fourteen participants screen passed the EMERGE trial, and 24 screen failed. Causes for screen failure included medical comorbidity (n = 3), negative amyloid status (n = 1), and inappropriate disease severity (n = 20) based on the participant-based Mini-Mental Status Examination (10) (MMSE) and Repeatable Battery for the Assessment of Neuropsychological Status (11) (RBANS), and the informant/participant-based Clinical Dementia Rating Scale (CDR) (12). Specifically, for the EMERGE trial the participant needed to score between 24-30 on the MMSE, at or below a demographically-normed standard score of 85 on the Delayed Memory Index from the RBANS, and at the level of 0.5 on CDR. Of the 18 participants with too severe of impairment on cognitive/informant examination at EMERGE screening, 17 participants performed below the cutoff for the MMSE, and one participant performed worse than permissible on the CDR. Both participants who were too intact on the cognitive/informant examination performed above the cutoff on the RBANS Delayed Memory Index. All procedures for the current study received approval by the university’s Institutional Review Board.
All participants underwent a standard clinical neuropsychological evaluation during the diagnostic neuropsychological assessment prior to their screening for the EMERGE trial, which included the following commonly administered neuropsychological, psychiatric, and functional tests. Readers are referred to Lezak and colleagues (13) and respective test manuals for test descriptions and psychometric properties.
•    Neuropsychological measures: Digit Span, Arithmetic, Information, and Matrix Reasoning subtests from the Wechsler Adult Intelligence Scale-IV, which measure attention, crystalized intelligence, and executive functioning, respectively; Brief Visual Memory Test-Revised (BVMT-R), which measures visual learning and memory; Hopkins Verbal Learning Test-Revised (HVLT-R), which measures verbal list-learning and memory; Trail Making Test Part B (TMT-B), which measures executive functioning; Montreal Cognitive Assessment (MOCA), which measures mental status; and Controlled Oral Word Association Test (COWA), which measures language. All individual subtests utilized raw scores, with higher scores indicating better performance for all tasks except TMT-B.
•    Psychiatric measures: Self-reported depression was assessed using the 30-item Geriatric Depression Scale (GDS), and self-reported anxiety was examined via the Zung Anxiety Self-Assessment Scale. Higher scores reflect greater symptoms of depression or anxiety.
•    Functional measures: Self-reported instrumental activities of daily living were assessed using the 10-item Functional Activities Questionnaire (FAQ). Higher scores indicate lower functioning.

Statistical Analysis

Group status of screen pass/fail was based on the EMERGE trial criteria described above. Evaluation of normality was undertaken for all continuous variables (14, 15), and all measures were determined to have a normal distribution except the FAQ. Independent samples t-tests were used to compare normal continuous data from neuropsychological, psychiatric, and functional performances with screen pass/fail group status, and independent samples Mann-Whitney U tests were used to compared non-normal continuous data (i.e., FAQ). For the categorical analysis of gender, Fisher’s exact test analysis was calculated based on screen pass/fail group categorization as the independent variable. Measures of effect size were expressed as Cohen’s d values for continuous data and Phi coefficients for categorical data. Two-tailed alpha levels were set using Holm’s Sequentially Rejective Bonferroni Test in order to control for multiple comparisons.

 

Results

Of the 38 participants in the current study, 14 participants screen passed this AD clinical trial, and 24 screen failed. The mean age was 72.5 years old (+/- 7.1 years) and the mean level of education was 16.4 years (+/- 2.7 years). All participants were non-Hispanic/Caucasian. No significant differences in age nor education were observed between screen pass/fail groups, t(36) = -0.64, p = .53, d = -0.21, for age and, t(36) = 1.54, p = .13, d = 0.51, for education (see Table 1). Conversely, higher screen failure rates were significantly related to female gender (p = .02, Fisher’s exact test, Phi = -0.40), with 83% of female participants screen failing this AD trial versus 45% of male participants.

Figure 1. CONSORT-like flow diagram of participants evaluated for the EMERGE trial, with those “Screened” representing the current study’s sample population

Figure 1. CONSORT-like flow diagram of participants evaluated for the EMERGE trial, with those “Screened” representing the current study’s sample population

 

There was no difference in performance on visual memory, t(35) = -0.38, p = .71, d = -0.13, or verbal memory tasks, t(34) = -0.47, p = .64, d = -0.16, between screen pass/fail groups, nor on a composite screen of mental status (MOCA), t(35) = 1.59, p = .12, d = 0.54. In contrast, performance differences were observed between screen pass/fail groups on several non-memory cognitive domains. Specifically, the screen fail group for this AD clinical trial tended to perform worse on Information, t(8) = 6.56, p = .001, d = 4.63, TMT-B, t(32.84) = -3.09, p = .004, d = -1.08, and Arithmetic, t(36) = 2.95, p = .006, d = 0.98. While trends were observed, no group differences were evident for Matrix Reasoning, t(35) = 2.25, p = .03, d = 0.76, Digit Span, t(35) = 2.14, p = .04, d = 0.72, or COWA, t(34) = 1.77, p = .09, d = 0.61 after controlling for multiple comparisons. Additionally, the screen fail group reported greater levels of anxiety, t(6) = -9.38, p < .001, d = 7.66, but not depression, t(32) = -0.16, p = .89, d = -0.06. An independent samples Mann-Whitney U test indicated that there was no difference in endorsements on the FAQ between the screen pass (Median = 14.69) and the screen fail (Median = 16.94) groups, U = 134.00, p = .49.

Table 1. Demographics and neuropsychological, psychiatric, and functional performance based on screening status

Table 1. Demographics and neuropsychological, psychiatric, and functional performance based on screening status

Note: 95% CI = 95% Confidence Interval of the Difference, MOCA = Montreal Cognitive Assessment, BVMT-R = Brief Visual Memory Test-Revised, HVLT-R = Hopkins Verbal Learning Test-Revised, COWA = Controlled Oral Word Association Test, TMT-B = Trail Making Test, Part B, Zung = Zung Anxiety Inventory, GDS = Geriatric Depression Scale, FAQ = Functional Assessment Questionnaire. Values listed as Mean (Standard Deviation).

 

Discussion

The current study analyzes neuropsychological, psychiatric, and functional data from clinical neuropsychological and neurological evaluations that were collected prior to the EMERGE trial screening visits in order to predict trial appropriateness and subsequently reduce AD trial screen fail rates. All results should be considered within the context of the small sample size of this exploratory study. Our results revealed that worse performance on non-memory neuropsychological domains was related to screen failure status for the EMERGE AD clinical trial. Specifically, participants performing worse on domains related to crystallized intelligence (d = 4.63), executive functioning (d = 1.08), and attention (d = 0.98) tended to screen fail this trial, and while not remaining significant after controlling for multiple comparisons, additional measures of executive functioning and attention possessed moderate to large effect sizes (d = 0.72 – 0.76) . The directionality of our findings—that worse performance on non-memory domains is associated with screen failing an AD trial—is somewhat unexpected. Upon further consideration, this result may be explained by the typical recruitment pathway from clinic to trials, which requires a diagnosis of interest (e.g., MCI or AD), but is otherwise up to the discretion of the physician to predict if the patient will “fit” into a trial. Physicians may erroneously view more globally-impaired patients as being better fits into clinical trials, resulting in greater recruitment of those patients and subsequently higher screen failure rates for those patients whose disease severity is too advanced for a particular trial. Alternatively, it is possible that participants who screen fail AD trials may have deficits that are atypical for MCI/early AD, and that their non-memory impairments may be at least partly due to non-AD pathology. These results suggest that recruiting patients into clinical trials earlier in their disease course, when their disease severity is less, may result in reduced screen failure rates in AD trials.
Conversely, we were not able to detect a relationship between memory-related tasks and screen fail/pass status. This finding was opposite of our hypothesis and in contrast with several large-scale studies suggesting that conversion to AD is associated with memory impairment (Alzheimer’s Disease Neuroimaging Initiative [ADNI] (16)). One explanation may be that the measures used in the EMERGE trial to gauge memory severity are not as sensitive to subtle changes in memory as neuropsychological memory measures that approximate a normal distribution of test performances. Specifically, only 3 points out of a total of 30 on the MMSE pertain to memory 10, and the CDR incorporates an ordinal scale of memory performance (0 – No Impairment, 0.5 – Questionable Impairment, 1 – Mild Impairment, 2 – Moderate Impairment, and 3 – Severe Impairment) with few participants in outpatient settings scoring at the highest levels (e.g., CDR levels 2 and 3 require “severe memory loss” with “new material either rapidly lost” or “only fragments remain” (12)). An alternative explanation may be that memory dysfunction is so common in AD and for patients considered for an AD trial that it is not necessarily surprising that memory performance does not distinguish who will be successfully screened into an AD clinical trial. As such, these results suggest that such memory dysfunction may be necessary but not sufficient to screen pass into an AD clinical trial, and that performances on other non-memory cognitive domains possess higher discriminative value.
In addition, our results showed that greater endorsements of anxiety are associated with higher screen failure rates (d = 7.66). This finding is congruous with research consistently observing higher levels of self-reported anxiety in more severe presentations of AD (17), and is similar to our other results suggesting that participants who screen failed the EMERGE trial displayed worse disease severity. Together, these results further support the notion that recruitment of patients earlier in the disease course may reduce screen failure rates in AD clinical trials. In contrast, a subjective measure of functional skills was not significantly associated with screen failure status in our study, which was unexpected given other findings in the literature that greater endorsements on functional scales were associated with greater conversion to AD (18). It is possible that the non-normal distribution of the sample of FAQ scores (skewness value of 1.62 [Standard Error (SE) = .42] and skew/SE ratio of 3.84, kurtosis value of 2.55 [SE = .82] and kurtosis/SE ratio of 3.10) may have limited our ability to find significance, though like memory dysfunction, functional loss may be necessary but not sufficient to discriminate screen pass/fail status.
Further, the current study examined demographic variables that were hypothesized to influence screen failure rates in this AD clinical trial. Our study observed that women displayed greater screen failure rates than men. This finding seems counter to research suggesting that women tend to worry more about health-related factors and men tend to minimize health-related risks (19), though this result potentially sheds light on the importance of spousal and care partner involvement (4) in patients with MCI or AD. Specifically, in this preliminary study, 78% of female participants were accompanied by their male spouse as care partner (14 of 18), and 90% of male participants were accompanied by their female spouse as care partner (18 of 20). As the majority of participants who screen failed the EMERGE trial did so due to below-cutoff performance on the MMSE (71% of overall screen failures, and 94% of participants failing due to performing too severely on screening measures), these differential results based on gender may suggest that male care partners may not identify the need for their partner to be involved in an AD trial until later in the disease, at which point the partner may have advanced to more severe disease states that would exclude them from successful trial enrollment. Finally, the lack of a significant difference between screen pass/fail groups for factors like education and age was contrary to our hypotheses, and to research showing that reduced education level and advanced age are both associated with worse cognitive performance (13).

Study Limitations

The proposed study is not without limitations. As alluded to above, our sample size likely hindered our ability to find statistical significance for some analyses. Additionally, our study is only attempting to examine clinical predictors of screen failure for patients that have already been diagnosed with the condition of interest (MCI or AD) and pre-screened for easily identifiable exclusionary medical comorbidities. Also, our sample is not representative of all patients seeking care from a cognitive specialty clinic or identified through advertisement without a prior clinical evaluation, and our predictor and outcome variables are also specific to those measures administered in our particular clinical evaluations and for this particular trial, respectively. Finally, this study is the first step in developing a rigorous model to investigate further ways to reduce AD trial screen failure rates, and does not address all barriers to AD trial recruitment or initiatives being undertaken elsewhere to improve recruitment (such as creating registries of trial-ready participants). Although this study only addresses pharmacological AD trials, one would assume that results would relate to non-pharmacological AD trials as well. This would be a future direction to examine, along with consideration of issues with enrolling a wider demographic of participants into AD trials associated with homogeneity of trial samples related to education (mostly highly educated), ethnicity (Caucasian), language (English-speaking), care-partner status (mostly opposite-gender spouse), and health status (without sensory impairments that would exclude from cognitive testing). Of importance, these preliminary findings do not suggest that other innovations described briefly above should not be undertaken, but propose a method to optimize successful recruitment of participants from current recruitment sources.

Future Directions

This current study is an exploratory examination of potential cognitive and psychiatric factors that may influence screen failure rates in AD clinical trials. By identifying predictors of AD trial screen failure that are already available to AD clinical trial teams, we hope to influence decision-making about which participants are most likely to be successfully enrolled in a trial with minimal additional effort required by the AD trial team. For example, if faced with limited screening resources, a clinical trials team member might review existing neuropsychological test results to identify a male patient with AD and memory dysfunction but otherwise largely preserved cognition rather than a female patient with AD and global cognitive dysfunction and anxiety, as the latter individual is more likely to screen fail the trial. Consequently, by building upon these initial findings, this research has potential to reduce screen fail rates in AD clinical trials, which will lower participant burden, maximize study resources, and cut costs. Future examination of 1) a collection of industry-sponsored trials and 2) large-scale databases from multi-site studies such as ADNI may further refine the process and potentially examine predictors not evaluated in the current study. Future studies could also apply this methodology to patients attending Annual Wellness Visits to streamline the pathway of participation from the Primary Care Clinic to AD intervention trials. Overall, these findings have the potential to advance the field by helping to enhance trial-recruitment infrastructure and to encourage greater engagement of older adults in AD research.

 

Funding: Funding for this project was provided by University of Utah Center for Alzheimer’s Care, Imaging and Research.

Acknowledgement: None.

Ethical standards: This study was conducted according to the University of Utah’s standards for Ethical Research. All procedures for the current study received approval by the University’s Institutional Review Board.

 

References

1.    Fargo KN, Carrillo MC, Weiner MW, Potter WZ, Khachaturian Z. The crisis in recruitment for clinical trials in Alzheimer’s and dementia: An action plan for solutions. Alzheimer’s & dementia 2016;12:1113-1115.
2.    Cooper C, Ketley D, Livingston G. Systematic review and meta-analysis to estimate potential recruitment to dementia intervention studies. International Journal of Geriatric Psychiatry 2014;29:515-525.
3.    Treves TA, Verchovsky R, Klimovitsky S, Korczyn AD. Recruitment rate to drug trials for dementia of the Alzheimer type. Alzheimer Disease and Associated Disorders 2000;14:209-211.
4.    Grill JD, Karlawish J. Study partners should be required in preclinical Alzheimer’s disease trials. Alzheimer’s Research & Therapy 2017;9:93.
5.    Watson JL, Ryan L, Silverberg N, Cahan V, Bernard MA. Obstacles and opportunities in Alzheimer’s clinical trial recruitment. Health Aff (Millwood) 2014;33:574-579.
6.    Aisen P, Touchon J, Amariglio R, et al. EU/US/CTAD Task Force: Lessons Learned from Recent and Current Alzheimer’s Prevention Trials. The Journal of Prevention of Alzheimer’s Disease 2017;4:116-124.
7.    Fitzpatrick AL, Fried LP, Williamson J, et al. Recruitment of the elderly into a pharmacologic prevention trial: the Ginkgo Evaluation of Memory Study experience. Contemporary Clinical Trials 2006;27:541-553.
8.    Grill JD, Monsell SE. Choosing Alzheimer’s disease prevention clinical trial populations. Neurobiology of Aging 2014;35:466-471.
9.    US National Library of Medicine: 221AD302 Phase 3 Study of Aducanumab (BIIB037) in Early Alzheimer’s Disease (EMERGE) [online]. Available at: https://clinicaltrials.gov/ct2/show/NCT02484547.
10.    Folstein MF, Folstein SE, McHugh PR. Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research 1975;12:189-198.
11.    Randolph C, Tierney MC, Mohr E, Chase TN. The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS): preliminary clinical validity. Journal of Clinical and Experimental Neuropsychology 1998;20:310-319.
12.    Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology 1993;43:2412-2414.
13.    Lezak M, Howieson D, Loring D. Neuropsychological Assessment, 4th ed. New York, NY: Oxford University Press, 2004.
14.    George, D, Mallery, P. SPSS for Windows Step by Step: A Simple Guide and Reference 17.0 Update. Boston, MA: Pearson, 2010.
15.    Tabachnick B, Fidell L. Using Multivariate Statistics, 6th ed. Boston: Pearson Education, 2013.
16.    Callahan BL, Ramirez J, Berezuk C, Duchesne S, Black SE. Predicting Alzheimer’s disease development: a comparison of cognitive criteria and associated neuroimaging biomarkers. Alzheimer’s Research & Therapy 2015;7:68.
17.    Burke SL, Cadet T, Alcide A, O’Driscoll J, Maramaldi P. Psychosocial risk factors and Alzheimer’s disease: the associative effect of depression, sleep disturbance, and anxiety. Aging Ment Health 2017:1-8.
18.    Gomar JJ, Bobes-Bascaran MT, Conejero-Goldberg C, Davies P, Goldberg TE, Alzheimer’s Disease Neuroimaging I. Utility of combinations of biomarkers, cognitive markers, and risk factors to predict conversion from mild cognitive impairment to Alzheimer disease in patients in the Alzheimer’s disease neuroimaging initiative. Arch Gen Psychiatry 2011;68:961-969.
19.    Gafvels C, Lithner F, Borjeson B. Living with diabetes: relationship to gender, duration and complications. A survey in northern Sweden. Diabet Med 1993;10:768-773.

MED I T ERRANEAN-DASH INTERVENTION FOR NEURODEGENERATIVE DELAY (MIND) DIET SLOWS COGNITIVE DECLINE AFTER STROKE

 

L. Cherian, Y. Wang, K. Fakuda, S. Leurgans, N. Aggarwal, M. Morris

 

Rush University Medical Center, Chicago, IL USA

Corresponding Author: Laurel Cherian, Rush University Medical Center, Chicago, IL USA,  laurel_j_cherian@rush.edu

J Prev Alz Dis
Published online June  12, 2019, http://dx.doi.org/10.14283/jpad.2019.28

 


Abstract

Objective: This study sought to determine if the MIND diet (a hybrid of the Mediterranean and Dash diets, with modifications based on the science of nutrition and the brain), is effective in preventing cognitive decline after stroke.
Design: We analyzed 106 participants of a community cohort study who had completed a diet assessment and two or more annual cognitive assessments and who also had a clinical history of stroke. Cognition in five cognitive domains was assessed using structured clinical evaluations that included a battery of 19 cognitive tests. MIND diet scores were computed using a valid food frequency questionnaire (FFQ). Dietary components of the MIND diet included whole grains, leafy greens and other vegetables, berries, beans, nuts, lean meats, fish, poultry, and olive oil and reduced consumption of cheese, butter, fried foods, and sweets. MIND diet scores were modeled in tertiles. The influence of baseline MIND score on change in a global cognitive function measure and in the five cognitive domains was assessed using linear mixed models adjusted for age and other potential confounders.
Results: With adjustment for age, sex, education, APOE-ε4, caloric intake, smoking, and participation in cognitive and physical activities, the top vs lowest tertiles of MIND diet scores had a slower rate of global cognitive decline (β = .08; CI = 0.0074, 0.156) over an average of 5.9 years of follow-up.
Conclusions: High adherence to the MIND diet was associated with a slower rate of cognitive decline after stroke.

Key words: Stroke, cognitive decline, diet, nutrition, prevention.


 

 

Cognitive decline is a common and devastating clinical sequela of stroke (1). Compared to the normal rate of neuron loss with aging, ischemic stroke causes 3.6 years’ worth of aging for every hour of untreated symptoms (2). With the average duration of a non-lacunar stroke lasting 10 hours, a brain may experience a magnitude of aging equivalent to several decades in just one day. Perhaps not surprisingly, stroke survivors have nearly double the risk of developing dementia compared to those who have not suffered a stroke (3). This results in a significant burden on our healthcare system, both in terms of the direct and indirect costs of stroke and dementia, as well as the emotional toll on patients and their caregivers. Therefore, lifestyle factors that may protect against these cognitive changes in stroke survivors are of great public health importance.
One lifestyle approach that may be effective for preventing post-stroke cognitive decline is diet. A number of studies have found protective associations between cognitive decline and greater adherence to the Mediterranean, Dietary Approaches to Stop Hypertension (DASH), and Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diets (4-7). There is limited data, however, on whether these dietary patterns might be effective in slowing the cognitive decline that can occur after stroke. In this study, we examined the associations among these healthy diet patterns and cognitive change in a community study of older adults with a clinical history of stroke.

 

Methods

Study Population

This study was conducted using data from the Rush Memory and Aging Project (MAP), a study of volunteers living in retirement communities and senior public housing units in the Chicago area. The ongoing open cohort study began in 1997 and includes annual clinical neurological examinations, as previously described (8). Beginning in 2004, MAP study participants began to complete comprehensive food frequency questionnaires (FFQ). Of the 1911 older persons enrolled in the MAP study, 1068 had at least one valid FFQ that served as the baseline for these analyses, of which 970 also had two or more annual cognitive assessments for the measurement of cognitive change. Among these, 106 participants had a clinical history of stroke. Average study follow-up time was 5.9 years (Figure 1).  The Institutional Review Board of Rush University Medical Center approved the study, and all participants gave written informed consent.

Figure 1. Analysis cohort

Figure 1. Analysis cohort

 

Cognitive Evaluations

Cognition was assessed in 5 domains (episodic memory, semantic memory, working memory, perceptual orientation, and perceptual speed), using annual structured clinical evaluations that included a battery of cognitive tests, administered by technicians trained and certified in standardized neuropsychological testing methods (9). Episodic memory was assessed with the following tests: word list, word list recall, word list recognition, East Boston immediate recall, East Boston delayed recall, logical memory 1 (immediate), and logical memory II (delayed). Semantic memory was assessed with the following tests: Boston naming (15 items), category fluency, and reading test (10 items). Working memory was assessed with the following tests: digits forward, digits backward, digit ordering. Perceptual orientation was assessed with the following tests: line orientation, progressive matrices (16 items). Finally, perceptual speed was assessed with the following tests: symbol digits modality-oral, number comparison, stroop color naming, and stroop word reading. Standardized scores were computed for each test, using the mean and standard deviation from the baseline tests, and the standardized scores were averaged over each cognitive domain and over all tests to create a global cognitive score. Out of all MAP participants, 93.4% complete annual cognitive evaluations. Of the participants in this study, 52.0% had 5 or more annual cognitive assessments, with a range of 2 to 10 years.

Diet Pattern Scoring

Diet pattern scores were based on responses to a modified Harvard semi-quantitative FFQ, that was validated for use in older Chicago community residents.(10). Typical frequency of intake of 144 food items was reported by participants over the prior 12 months. The caloric content and nutrient levels for each food item were based on age- and sex-specific portion sizes from national dietary surveys, or by a logical portion size (e.g. a slice of bread). Details of the dietary components and maximum scores for the MIND, DASH, and Mediterranean diets have been previously reported (4, 11, 12). Briefly, the MIND diet score is based on a combination of 10 healthy food groups (leafy green vegetables, other vegetables, nuts, berries, beans, whole grains, fish, poultry, olive oil, and wine) and 5 unhealthy food groups (red meats, butter and stick margarine, cheese, pastries and sweets, fried food, and fast food). If olive oil was reported as the primary oil used at home, it was scored 1. Otherwise, olive oil consumption was scored 0. For the remaining components, the frequency of consumption of each food item for a given score component was summed and then given a concordance score of 0, 0.5, or 1, where 1 represented the highest concordance (4). The final MIND diet score was the sum of the 15 component scores.
Scoring for the DASH diet was determined based on consumption of 3 dietary components (total fat, saturated fat, and sodium) and 7 food groups (grains, fruits, vegetables, nuts, seeds and legumes, dairy, and meat) (12). Scores of 0, 0.5, and 1 were assigned to each food group based on the frequency of consumption. Total possible scores ranged from 0 (lowest) to 10 (highest) diet concordance.
The Mediterranean diet pattern was based on the MedDiet score as described by Panagiotakos and colleagues (11) that uses serving quantities of the traditional Greek Mediterranean diet as the comparison metric. Eleven dietary components (non-refined cereals, potatoes, fruits, vegetables, legumes, fish, red meat and products, poultry, full fat dairy products, the use of olive oil in cooking, and alcohol) are each scored from 0 to 5 and then summed for a total score ranging from 0 to 55 (highest concordance).

Covariates

Non-dietary variables in the analysis were obtained at the participant’s baseline clinical evaluation through a combination of clinical evaluation, self-report, medication inspection, and measurements.  The process is identical to that performed in the Religious Orders Study, and was designed to reduce costs and enhance uniformity of diagnostic decisions over time and space (13). Participants self-reported their birth date and years of education. A 5 point scale was used to assess the frequency of cognitively stimulating activities (such as writing letters, visiting the library, reading, and playing games).[14] Physical activity was determined by participants self-reported minutes spent over the previous 2 weeks on 5 activities (walking for exercise, yard work, calisthenics, biking, and water exercise) (15). A modified 10-item version of the Center for Epidemiological Studies-Depression (CESD) scale was used to evaluate depressive symptoms (16). High throughput sequencing was used to determine APOE- genotyping as previously described (17). Height and weight were measured to determine body mass index (BMI=weight in kg/height in m2) and modeled as two indicator variables, BMI ≤20 and BMI ≥30. Hypertension was defined by an average of 2 blood pressure measurements ≥ 160 mmHg systolic or ≥ 90 mmHg diastolic, or if the patient reported a clinical history of hypertension or was currently taking antihypertensive medications. Myocardial infarction history was based on the current use of cardiac glycosides (e.g. lanoxin or digoxin) or by self-reported history. Clinical history of diabetes was obtained by self-reported medical diagnosis or by current use of diabetic medications. Diagnosis of stroke was obtained through a combination of clinical evaluation and self-report to the question “has a doctor, nurse, or therapist ever told you that you have had a stroke?” (18). Medication use was based on interviewer inspection.

Statistical Analysis

The data were summarized using median and quartiles, mean and SD or number (relative frequency) as appropriate.  Baseline characteristics were compared across MIND diet tertiles using Kruskal-Wallis, ANOVA, chi-squared tests or Fisher’s exact tests, as appropriate.  Linear mixed models were used to model the longitudinal global cognitive scores and the 5 cognitive domains on diet scores for the MIND, DASH, and Mediterranean diets to describe the relationships among dietary patterns and cognitive decline over time in stroke survivors. The 3 dietary patterns were examined in separate models: an age-adjusted model and a basic-adjusted model that included potential confounders previously associated with Alzheimer disease: age, sex, education, participation in cognitively stimulating activities, physical activity, smoking, and APOE-ε4.  Total energy intake, which is closely related to diet, was also included as a potential confounder. The dietary scores were modeled as both continuous variables and as indicators of the top two tertiles in each of these models.

 

Results

Of the 106 MAP participants with a clinical history of stroke, the mean age was 82.8 years (SD=7.1) and 29 (27%) were male. The mean years of education was 14.4 (SD=2.7) Overall, 16% had APOE-ε4 alleles. Participants who had high MIND diet scores were less likely to be male, more likely to have never been smokers, and more likely to participate frequently in cognitive and physical activities (Table 1).

Table 1. Baseline Characteristics of Memory and Aging Project Subjects with History of Stroke

Table 1. Baseline Characteristics of Memory and Aging Project Subjects with History of Stroke

 

In separate models adjusted for age, sex, education, APOE-ε4, late-life cognitive activity, caloric intake, physical activity, and smoking, with diet scores modeled in tertiles, the top versus the lowest tertile of MIND diet scores were associated with a slower rate of global cognitive decline (β=0.08, 95% confidence interval (CI): 0.01, 0.16), as well as with a slower decline in semantic memory (β=0.07,  95% CI: 0.00, 0.14) and perceptual speed (β=0.07, 95% CI: 0.00, 0.14), (Figure 2). Those with moderate adherence (tertile 2) to the MIND diet showed a non-significant trend toward slower rates of cognitive decline.  In continuous models, the MIND diet was associated with slower rates of decline in cognitive function over time for both global cognition (p=0.034) and semantic memory (p=0.04). The DASH and Mediterranean diets were not associated with slower rates of global cognitive decline over time (p =0.26 and p= 0.11, respectively) or slower decline in any of the 5 cognitive domains (Table 2).

Figure 2. Cognitive Decline Over Time by Adherence to the MIND Diet

Figure 2. Cognitive Decline Over Time by Adherence to the MIND Diet

A graphical representation of the decrease in cognitive decline over time based on adherence to the MIND diet for 106 participants found to have had a stroke at baseline. The highest adherence (represented by the green line) versus lowest adherence (represented by the red line) to the MIND diet showed a significant decrease in cognitive decline (ß=0.08 CI= 0.00, 0.16). The decrease in cognitive decline for moderate adherence (represented by the blue line) versus lowest adherence approached significance (ß=0.06 CI= -0.01, 0.13).

Table 2. Cognitive Function by Dietary Pattern

Table 2. Cognitive Function by Dietary Pattern

Adjustments – age, sex, education, APO-E4, late life cog act, caloric intake, physical activity, & smoking; Italicized and bold – statistically significant: Italicized – approaching significance

 

Discussion

Although an extensive body of literature exists on the role of diet in stroke prevention, relatively few studies have examined the role of diet on cognitive decline post-stroke, even though stroke nearly doubles the risk of dementia (3). In the present study, we observed a community cohort of older persons with a clinical history of stroke but no diagnosis of dementia at their baseline enrollment to determine the role that diet may play in preventing post-stroke cognitive decline. In this observational study, we found that the MIND diet significantly slowed the rate of decline in global cognition, as well as in the individual cognitive domains of semantic memory and perceptual speed. The Mediterranean and DASH diets were not associated with slowing global cognitive decline or slowing decline in any of the 5 cognitive domains. This suggests that while the Mediterranean and DASH diets may be useful in preventing stroke and other cardiovascular conditions, the MIND diet, which is specifically tailored for brain health, may be more effective in preventing post-stroke cognitive decline.
Large, prospective cohort studies that established the role of diet in the prevention of cardiovascular disease include the Nurses Health Study, the Reasons for Geographic and Racial Differences in Stroke study, The Northern Manhattan Study, and The Framingham Heart Study (19-21). A smaller number of randomized controlled trials, such as PREDIMED[22], have also found diet to be effective in the prevention of cardiovascular outcomes including stroke. Fewer data exist on the role of diet in secondary stroke prevention, although several studies such as ONTARGET, TRANSCEND (23) and the Lyon Heart Study (24) have shown that diet may be a valuable target in secondary stroke prevention as well, with some studies suggesting that diet may provide an effect size similar to that of statins (25).
Despite separate studies advocating the role of diet both in stroke prevention and the prevention of cognitive decline, most of these studies did not examine the role of diet in preventing cognitive decline in subjects with a history of stroke specifically, a population that is at higher risk for dementia than the general population. In fact, many of the existing large observational cohort studies have excluded subjects with a clinical history of stroke at baseline (20).
The MIND diet, which is a hybrid of the Mediterranean and DASH diets, was designed to emphasize nutrients that have been associated with dementia prevention and to discourage elements, such as saturated/hydrogenated fats, that have been associated with dementia (4). The MIND diet recommends greater than or equal to 3 servings of whole grain per day (26), greater than or equal to 6 servings of leafy green vegetables per week (in addition to one or more daily servings of other vegetables) (27), greater than or equal to 2 servings of berries per week (28), greater than or equal to one serving of fish per week (29), greater than or equal to 2 servings of poultry per week, greater than 3 servings of beans per week, and greater than or equal to 5 servings of nuts per week (30). The MIND diet recommends that olive oil be used as the primary source of fat (31, 32) and allows one serving of alcohol/wine per day (33). The following food items are discouraged by the MIND diet: red meat and products, less than 4 servings per week; fast food and fried food, less than one serving per week; butter/margarine, less than 1tsp per day; cheese, less than once per week; and pastries/sweets, less than 5 servings per week.
The MIND diet is a rich source of many different dietary components that have been linked to brain health, including vitamin E, folate, n-3 fatty acids, carotenoids, and flavonoids. Multiple prospective cohort studies have shown that avoiding saturated and trans-unsaturated (hydrogenated) fats and increasing the consumption of antioxidant nutrients and B-vitamins are associated with slower rates of cognitive decline (34-36). The emphasis on the consumption of berries vs. fruit in general was based on findings from multiple epidemiological studies of cognition, showing that, whereas overall fruit consumption does not appear to impart a protective effect (27, 37-39), the subtype of fruit, berries, does appear to slow cognitive decline (28). Vegetables, and leafy green vegetables in particular, have also been shown in several large prospective studies to reduce cognitive decline (27, 37).
The Mediterranean diet has been widely studied (40, 41) and recommends greater than or equal to 4 tablespoons of olive oil per day, 3 or more servings of tree nuts and peanuts per week, 3 or more servings of fruit per day, 2 or more servings of vegetables per day, 3 or more servings of fish (particularly fatty fish) per week, 3 or more servings of legumes per week, using white meat as a substitute for red meat, and drinking 1 or more glasses of wine with meals, 7 or more times per week. The Mediterranean diet limits soda to less than one per day, consumption of commercial baked goods, sweets, and pastries to less than 3 per week; spreadable fats to less than 1 per day; and red and processed meats to less than once per day.
The Mediterranean diet was associated with higher cognitive scores in a sub-study of PREDIMED[31], a randomized trial designed to test diet effects on cardiovascular outcomes among Spaniards at high cardiovascular risk. In our study, although the Mediterranean diet was associated with slower rates of global cognitive decline in the age-adjusted model, this association became nonsignificant when basic adjustments for sex, education, APOE-ε4, late-life cognitive activity, caloric intake, physical activity, and smoking were applied.
The DASH diet was not associated with slower rates of cognitive decline in our study, although prior studies have shown this diet to be effective for prevention of both cognitive decline (26, 42, 43) and stroke prevention (44).
This study has several limitations, the most important of which is that it is observational in nature; as such, it cannot claim a cause and effect relationship. While replication in other observational cohort studies would be useful to confirm the associations seen in this study, a diet intervention trial in stroke survivors is needed to establish a causal role between diet and post-stroke cognition. Another limitation of this study is its small sample size resulting in low power to observe associations. It may be possible to observe protective associations of the DASH and Mediterranean diets on cognitive decline in larger stroke populations. Nonetheless, many larger observational cohort studies examining the role of nutrition on cognitive decline have excluded subjects with a clinical history of stroke. Therefore, we believe that this is an important and under-studied population that may be disproportionately prone to developing dementia, and preliminary data are important to guide future studies.
Clinical history of stroke was determined by self-report or by diagnosis during an annual clinical neurologic examination, but the lack of MRI or CT to confirm this diagnosis or to differentiate between stroke sub-type is a limitation. Subjects with a clinical history of mild stroke or a radiographic infarct may have been excluded from our sample, but the Framingham Offspring Study found that individuals with silent cerebral infarcts have similar risk profiles to those with a clinical history of stroke (45). We suspect that the inclusion of these individuals in our analysis would have been more likely to strengthen our findings than to invalidate them. Other large prospective observational cohort studies, such as the Nurses Health Study (46) have employed questionnaires and clinical evaluations to identify cardiovascular outcomes, and suggested that self-reported stroke is a valid approach to assessing the prevalence of stroke in a population (47-49). In the Tromso Study, researchers followed up with 213 individuals who had self-reported histories of stroke at a community health fair and found that upon more intensive evaluation (physician examination and review of medical records, including neuroimaging) 79.2% of self-reported strokes were confirmed (47). Self-reported stroke was found to have a similar prognostic value for predicting recurrent stroke in the Health in Men Study, and the authors concluded that self-reported stroke may be useful in further epidemiological studies (49).
The MAP cohort is an older, predominantly non-Hispanic white population, so findings should not be generalized to other ethnic groups or younger cohorts. The dietary questionnaires had limited questions regarding some of the dietary components and information on frequency of consumption. For example, a single item each provided information on the consumption of nuts, berries, beans, and olive oil. This study’s strengths include the use of a validated food questionnaire for comprehensive dietary assessment, the measurement of cognitive change with a large battery of standardized tests annually for up to 10 years, and statistical control of the important confounding factors.
The MIND diet is a hybrid of the Mediterranean and DASH diets, with additional emphasis on the nutritional components that have been shown to optimize brain health. The MIND diet has previously been shown to slow cognitive decline in the general population in an observational cohort study,[4] but it was unclear whether this association would remain strong for subjects with a clinical history of stroke. This observational study suggests that not only is the MIND diet strongly associated with slowing cognitive decline post-stroke, its estimated effect was twice the size of that observed in the overall MAP cohort (41). Additionally, the MIND diet appeared superior to the Mediterranean and DASH diets in slowing cognitive decline in stroke survivors. Given the projected burden of stroke and dementia in an aging population, further studies are warranted to explore the role of the MIND diet in preventing cognitive decline in stroke survivors. High adherence to the MIND diet was associated with slower rates of cognitive decline in an observational study of older adults with a clinical history of stroke.

 

Funding: Supported by grants from the NIA (R01 AG054476 and R01AG17917). The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; in the preparation of the manuscript; or in the review or approval of the manuscript.

Conflict of Interest and Financial Disclosures: Laurel Cherian: Study concept and design, Interpretation of data, and writing of manuscript. Dr. Cherian reports no disclosures. Yamin Wang: Analysis and interpretation. Dr. Wang reports no disclosures. Keiko Fakuda: Background research and initial draft of introduction and discussion. Ms. Fakuda reports no disclosures. Sue Leurgans: Critical revision of the manuscript for important intellectual content. Dr. Leurgans reports no disclosures. Neelum Aggarwal: Study concept and design, critical revision of the manuscript for important intellectual content. Dr. Aggarwal reports no disclosures. Martha Clare Morris: Study concept and design, critical revision of the manuscript for important intellectual content. Dr. Morris reports no disclosures.

Ethical standards: The authors attest that they have provided an accurate account of the work performed as well as an objective discussion of its significance. Relevant raw data has been accurately provided.  The authors attest that the work is original and has not been published elsewhere. Pertinent work from other sources has been appropriately cited. The Institutional Review Board of Rush University Medical Center approved the study, and all participants gave written informed consent.

 

References

1.    Schellinger, P.D., et al., Assessment of additional endpoints for trials in acute stroke – what, when, where, in who? Int J Stroke, 2012. 7(3): p. 227-30.
2.    Saver, J.L., Time is brain–quantified. Stroke, 2006. 37(1): p. 263-6.
3.    Portegies, M.L., et al., Prestroke Vascular Pathology and the Risk of Recurrent Stroke and Poststroke Dementia. Stroke, 2016. 47(8): p. 2119-22.
4.    Morris, M.C., et al., MIND diet slows cognitive decline with aging. Alzheimers Dement. 11(9): p. 1015-22.
5.    Morris, M.C., et al., MIND diet associated with reduced incidence of Alzheimer’s disease. Alzheimers Dement. 11(9): p. 1007-14.
6.    Scarmeas, N., et al., Physical activity, diet, and risk of Alzheimer disease. JAMA, 2009. 302(6): p. 627-37.
7.    Tangney, C.C., et al., Adherence to a Mediterranean-type dietary pattern and cognitive decline in a community population. Am J Clin Nutr, 2011. 93(3): p. 601-7.
8.    Bennett, D.A., et al., Overview and findings from the rush Memory and Aging Project. Curr Alzheimer Res, 2012. 9(6): p. 646-63.
9.    Wilson, R.S., et al., Odor identification and decline in different cognitive domains in old age. Neuroepidemiology, 2006. 26(2): p. 61-7.
10.    Morris, M.C., et al., Validity and reproducibility of a food frequency questionnaire by cognition in an older biracial sample. Am J Epidemiol, 2003. 158(12): p. 1213-7.
11.    Panagiotakos, D.B., et al., Adherence to the Mediterranean food pattern predicts the prevalence of hypertension, hypercholesterolemia, diabetes and obesity, among healthy adults; the accuracy of the MedDietScore. Prev Med, 2007. 44(4): p. 335-40.
12.    Epstein, D.E., et al., Determinants and consequences of adherence to the dietary approaches to stop hypertension diet in African-American and white adults with high blood pressure: results from the ENCORE trial. J Acad Nutr Diet, 2012. 112(11): p. 1763-73.
13.    Bennett, D.A., et al., Religious Orders Study and Rush Memory and Aging Project. J Alzheimers Dis, 2018. 64(s1): p. S161-S189.
14.    Wilson, R.S., et al., Early and late life cognitive activity and cognitive systems in old age. J Int Neuropsychol Soc, 2005. 11(4): p. 400-7.
15.    Buchman, A.S., et al., Physical activity and motor decline in older persons. Muscle Nerve, 2007. 35(3): p. 354-62.
16.    Kohout, F.J., et al., Two shorter forms of the CES-D (Center for Epidemiological Studies Depression) depression symptoms index. J Aging Health, 1993. 5(2): p. 179-93.
17.    Buchman, A.S., et al., Apolipoprotein E e4 allele is associated with more rapid motor decline in older persons. Alzheimer Dis Assoc Disord, 2009. 23(1): p. 63-9.
18.    Bennett, D.A., Secular trends in stroke incidence and survival, and the occurrence of dementia. Stroke, 2006. 37(5): p. 1144-5.
19.    Fung, T.T., et al., Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation, 2009. 119(8): p. 1093-100.
20.    Tsivgoulis, G., et al., Adherence to a Mediterranean diet and prediction of incident stroke. Stroke. 46(3): p. 780-5.
21.    Willey, J.Z., et al., Lipid profile components and risk of ischemic stroke: the Northern Manhattan Study (NOMAS). Arch Neurol, 2009. 66(11): p. 1400-6.
22.    Estruch, R., et al., Primary prevention of cardiovascular disease with a Mediterranean diet. N Engl J Med. 368(14): p. 1279-90.
23.    Dehghan, M., et al., Relationship between healthy diet and risk of cardiovascular disease among patients on drug therapies for secondary prevention: a prospective cohort study of 31 546 high-risk individuals from 40 countries. Circulation, 2012. 126(23): p. 2705-12.
24.    de Lorgeril, M., et al., Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study. Circulation, 1999. 99(6): p. 779-85.
25.    Dearborn, J.L., V.C. Urrutia, and W.N. Kernan, The case for diet: a safe and efficacious strategy for secondary stroke prevention. Front Neurol, 2015. 6: p. 1.
26.    Wengreen, H., et al., Prospective study of Dietary Approaches to Stop Hypertension- and Mediterranean-style dietary patterns and age-related cognitive change: the Cache County Study on Memory, Health and Aging. Am J Clin Nutr, 2013. 98(5): p. 1263-71.
27.    Morris, M.C., et al., Associations of vegetable and fruit consumption with age-related cognitive change. Neurology, 2006. 67(8): p. 1370-6.
28.    Devore, E.E., et al., Dietary intakes of berries and flavonoids in relation to cognitive decline. Ann Neurol, 2012. 72(1): p. 135-43.
29.    Morris, M.C., et al., Fish consumption and cognitive decline with age in a large community study. Arch Neurol, 2005. 62(12): p. 1849-53.
30.    Urpi-Sarda, M., et al., Virgin olive oil and nuts as key foods of the Mediterranean diet effects on inflammatory biomakers related to atherosclerosis. Pharmacol Res. 65(6): p. 577-83.
31.    Martinez-Lapiscina, E.H., et al., Mediterranean diet improves cognition: the PREDIMED-NAVARRA randomised trial. J Neurol Neurosurg Psychiatry, 2013. 84(12): p. 1318-25.
32.    Razquin, C., et al., A 3 years follow-up of a Mediterranean diet rich in virgin olive oil is associated with high plasma antioxidant capacity and reduced body weight gain. Eur J Clin Nutr, 2009. 63(12): p. 1387-93.
33.    Larrieu, S., et al., Nutritional factors and risk of incident dementia in the PAQUID longitudinal cohort. J Nutr Health Aging, 2004. 8(3): p. 150-4.
34.    Morris, M.C., Nutritional determinants of cognitive aging and dementia. Proc Nutr Soc, 2012. 71(1): p. 1-13.
35.    Gillette Guyonnet, S., et al., IANA task force on nutrition and cognitive decline with aging. J Nutr Health Aging, 2007. 11(2): p. 132-52.
36.    Morris, M.C. and C.C. Tangney, Dietary fat composition and dementia risk. Neurobiol Aging, 2014. 35 Suppl 2: p. S59-64.
37.    Kang, J.H., A. Ascherio, and F. Grodstein, Fruit and vegetable consumption and cognitive decline in aging women. Ann Neurol, 2005. 57(5): p. 713-20.
38.    Nooyens, A.C., et al., Fruit and vegetable intake and cognitive decline in middle-aged men and women: the Doetinchem Cohort Study. Br J Nutr, 2011. 106(5): p. 752-61.
39.    Chen, X., Y. Huang, and H.G. Cheng, Lower intake of vegetables and legumes associated with cognitive decline among illiterate elderly Chinese: a 3-year cohort study. J Nutr Health Aging, 2012. 16(6): p. 549-52.
40.    Mitjavila, M.T., et al., The Mediterranean diet improves the systemic lipid and DNA oxidative damage in metabolic syndrome individuals. A randomized, controlled, trial. Clin Nutr. 32(2): p. 172-8.
41.    Gadgil, M.D., et al., The effects of carbohydrate, unsaturated fat, and protein intake on measures of insulin sensitivity: results from the OmniHeart trial. Diabetes Care. 36(5): p. 1132-7.
42.    Solfrizzi, V., et al., Relationships of Dietary Patterns, Foods, and Micro- and Macronutrients with Alzheimer’s Disease and Late-Life Cognitive Disorders: A Systematic Review. J Alzheimers Dis, 2017. 59(3): p. 815-849.
43.    Tangney, C.C., et al., Relation of DASH- and Mediterranean-like dietary patterns to cognitive decline in older persons. Neurology, 2014. 83(16): p. 1410-6.
44.    Folsom, A.R., E.D. Parker, and L.J. Harnack, Degree of concordance with DASH diet guidelines and incidence of hypertension and fatal cardiovascular disease. Am J Hypertens, 2007. 20(3): p. 225-32.
45.    Das, R.R., et al., Prevalence and correlates of silent cerebral infarcts in the Framingham offspring study. Stroke, 2008. 39(11): p. 2929-35.
46.    Fung, T.T., et al., Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med, 2008. 168(7): p. 713-20.
47.    Engstad, T., K.H. Bonaa, and M. Viitanen, Validity of self-reported stroke : The Tromso Study. Stroke, 2000. 31(7): p. 1602-7.
48.    Jackson, C.A., et al., Moderate agreement between self-reported stroke and hospital-recorded stroke in two cohorts of Australian women: a validation study. BMC Med Res Methodol, 2015. 15: p. 7.
49.    Jamrozik, E., et al., Validity of self-reported versus hospital-coded diagnosis of stroke: a cross-sectional and longitudinal study. Cerebrovasc Dis, 2014. 37(4): p. 256-62.

TREATABLE VASCULAR RISK AND COGNITIVE PERFORMANCE IN PERSONS AGED 35 YEARS OR OLDER: LONGITUDINAL STUDY OF SIX YEARS

 

M.E.A. van Eersel1, H. Joosten2, R.T. Gansevoort3, J.P.J. Slaets1, G.J. Izaks1

 

1. University of Groningen, University Medical Center Groningen, University Center for Geriatric Medicine, Groningen, The Netherlands; 2. Department of Internal Medicine, Maastricht University Medical Center, Maastricht, The Netherlands; 3. University of Groningen, University Medical Center Groningen, Department of Nephrology, Groningen, The Netherlands

Corresponding Author: Marlise E.A. van Eersel (MEAE), University Center for Geriatric Medicine, University Medical Center Groningen, Internal Postcode: AA41, PO Box 30.001, 9700 RB Groningen, The Netherlands, Phone: 0031(0)50 361 39 21, Fax: 0031(0)50 361 90 69, Email: mea.eersel@umcg.nl

J Prev Alz Dis 2019;(6) in press
Published online December 14, 2018, http://dx.doi.org/10.14283/jpad.2018.47

 


Abstract

Background: Poor cognitive performance is associated with high vascular risk. However, this association is only investigated in elderly. As neuropathological changes precede clinical symptoms of cognitive impairment by several decades, it is likely that cognitive performance is already associated with vascular risk at middle-age.
OBJECTIVES: To investigate the association of cognitive performance with treatable vascular risk in middle-aged and old persons.
DESIGN: Longitudinal study with three measurements during follow-up period of 5.5 years.
SETTING: City of Groningen, the Netherlands.
PARTICIPANTS: Cohort of 3,572 participants (age range, 35-82 years; mean age, 54 years; men, 52%).
EXPOSURE: Treatable vascular risk as defined by treatable components of the Framingham Risk Score for Cardiovascular Disease at the first measurement (diabetes mellitus, smoking, hypercholesterolemia and hypertension).
MEASUREMENTS: Change in cognitive performance during follow-up. Cognitive performance was measured with Ruff Figural Fluency Test (RFFT) and Visual Association Test (VAT), and calculated as the average of the standardized RFFT and VAT score per participant.
RESULTS: The mean (SD) cognitive performance changed from 0.00 (0.79) at the first measurement to 0.15 (0.83) at second measurement and to 0.39 (0.82) at the third measurement (Ptrend<0.001). This change was negatively associated with treatable vascular risk: the change in cognitive performance between two measurements decreased with 0.004 per one-point increment of treatable vascular risk (95%CI, -0.008 to 0.000; P=0.05) and with 0.006 per one-year increment of age (95%CI,  -0.008 to -0.004; P<0.001).
CONCLUSIONS: Change in cognitive performance was associated with treatable vascular risk in persons aged 35 years or older.

Key words: Cognitive performance, treatable vascular risk, longitudinal analysis, cardiovascular disease, preventing cognitive impairment.


 

 

Introduction

Several studies have shown that poor cognitive performance is associated with vascular risk factors in persons aged 60 years or older (1). As neuropathological changes start several decades prior to the clinical expression of cognitive impairment (2), it is likely that cognitive performance is already associated with vascular risk factors at middle-age. However, vascular risk factors in middle-aged persons often are only marginally elevated if considered separately. Nevertheless they result in a clearly increased vascular risk if considered together (3-6), which may contribute to the onset of neurodegenerative changes in the brain (7). Therefore, it is essential to know whether cognitive performance is associated with a high vascular risk in middle-aged persons.
Vascular risk is usually estimated with multicomponent risk scores that predict an individual’s risk of a vascular event within the next years (3-6). These vascular risk scores are largely based on age. However, although age is a major vascular risk factor, it is not amenable to treatment. For effective prevention of cognitive impairment it is essential to know whether cognitive performance is associated with treatable vascular risk based on treatable components like, for example, diabetes mellitus, hypertension and hypercholesterolemia. Up till now, three longitudinal studies have found a negative association of cognitive performance with treatable vascular risk independent of age (8-10). However, one study included a relatively small sample of 235 men aged 60 years or older (8), whereas the two other studies mainly included even older persons from the same source population (the Alzheimer’s Disease Centers) (9,10). Furthermore, in these three studies, the treatable vascular risk was based on a stroke-specific risk score and did not include the risk of cardiac or peripheral vascular events (4). Therefore, it is still unclear whether cognitive performance is associated with general treatable vascular risk, and not only with stroke-specific risk. Finally, the association between cognitive performance and treatable vascular risk is not yet investigated in middle-aged persons since current data are only available for elderly (8-10).

Therefore, this longitudinal study aims to investigate the association of cognitive performance with treatable (general) vascular risk independent of age over a follow-up period of six years in both middle-aged as old persons.

 

Methods

Study design

This study was part of the Prevention of REnal and Vascular ENd-stage Disease (PREVEND) cohort. The PREVEND study is a prospective cohort study investigating the natural course of microalbuminuria and its association with renal and cardiovascular disease. Details of the PREVEND study have been described elsewhere (11, 12). Briefly, at baseline 8,592 participants aged 28-75 years were selected from inhabitants of the city of Groningen (Netherlands) based on their urinary albumin excretion. These participants completed the baseline survey in 1997-1998 and were followed over time. Surveys included assessment of demographic and vascular risk factors, and measurements of haematological and biochemical parameters. Cognitive function tests were introduced at the third survey (2003-2006) and repeated at the fourth survey (2006-2008) and fifth survey (2008-2012). A total of 3,601 participants completed two to three measurements of cognitive performance.
The PREVEND study was approved by the medical ethics committee (METc) of University Medical Center Groningen, Groningen, the Netherlands, and conducted in accordance with the guidelines of the Helsinki declaration. All participants gave written informed consent.

Cognitive performance

Cognitive performance was measured as a composite score of two tests: the Ruff Figural Fluency Test (RFFT) and the Visual Association Test (VAT). The RFFT is generally seen as a measure of executive function but provides also information regarding planning, divergent thinking and the ability to shift between different cognitive tasks. The RFFT requires the participants to draw as many designs as possible within a set time limit while avoiding repetitions of designs. The main outcome of the RFFT is the total number of unique designs, which range from 0 points (worst score) to 175 points (best score) (13). The RFFT is sensitive to changes in cognitive performance in both young and old persons (13, 14).
The VAT is a brief learning task that is designed to detect memory impairment including anterograde amnesia. The test consists of six drawings of pairs of interacting objects. The participant is asked to name each object and, later, is presented with one object from the pair and asked to name the other object. The lowest (worst) score is 0 points, the highest (best) score is 12 points (15).
To create a composite cognitive score, the raw RFFT and VAT scores at each measurement were standardized to z-scores (based on the mean and standard deviation of each test at the first measurement) and subsequently averaged.

Treatable vascular risk

Treatable vascular risk was based on the components of the Framingham Risk Score for Cardiovascular Disease (FRS-CD) that are amenable to treatment: diabetes mellitus (yes/no), current smoker status (yes/no), systolic blood pressure (mmHg), total cholesterol (mmol/l), HDL cholesterol (mmol/l) and use of blood pressure lowering drugs (yes/no). The FRS-CD is designed to predict the risk of a new cardiovascular, cerebrovascular or peripheral vascular event within the next ten years. This model was validated for persons aged 30-74 years without vascular history (3).
A higher treatable vascular risk score is associated with a higher risk of a new vascular event: the lowest score is -5 (10-year risk <1%), and the highest score is 21 (10-year risk >30%) (3).

Measurements of treatable vascular risk components

Data on the treatable vascular risk were obtained in the third survey of the PREVEND study at the same visit at which the first measurement of cognitive performance was done: total cholesterol, HDL-cholesterol and glucose were measured with fasting blood tests. Diabetes mellitus was defined as a fasting glucose ≥7.0 mmol/L (126 mg/dl) or a non-fasting glucose ≥11.0 mmol/L (200 mg/dl) or the use of glucose-lowering drugs. Smoking was defined as current smoker based on self-report. Systolic blood pressure was automatically measured (Dinamap) in a supine position during ten minutes and reported as the average of the two last measurements. Data on actual drug use were obtained from the InterAction DataBase that comprised pharmacy-dispending data from regional community pharmacies (16).

Covariates

Demographic factors were measured at the first measurement. Data on age, gender and educational level were obtained from a questionnaire. Educational level was divided into four groups: primary school (0 to 8 years of education), lower secondary education (9 to 12 years of education), higher secondary education (13 to 15 years of education), and university (≥16 years of education). Because the effect of vascular risk on cognitive function is possibly modified by APOE ε4 carriership (17), APOE ε4 genotype was included as a covariate. Participants were categorized as APOE ε4 carriers (allele combinations e2/e4 or e3/e4 or e4/e4) or noncarriers (allele e2/e2 or e2/e3 or e3/e3).

Statistical analysis

Parametric data are presented as mean and standard deviation (SD) and nonparametric data as median and interquartile range (IQR). Differences were tested by independent-samples t test or, if appropriate, Mann-Whitney U test. Differences between paired observations were tested by paired-samples t test or, if appropriate, Wilcoxon signed-rank test. Differences in proportion were tested by Chi-Square test. Trends across measurements were analyzed by ANOVA for parametric data and by Kruskal-Wallis H test for nonparametric data.
The longitudinal association of cognitive performance with the treatable vascular risk was investigated by linear multilevel analysis (linear mixed model analysis). Cognitive performance was the dependent variable. Treatable vascular risk at the first measurement was the independent variable. The analysis included the data of all participants who completed the cognitive tests on at least two measurements. Consecutive measurement (1, 2, or 3) was the lowest level and participant the highest level. Interaction between the treatable vascular risk and consecutive measurement was investigated by entering the product term treatable vascular risk x consecutive measurement into the regression model. Interaction between the treatable vascular risk and APOE ε4 carriership was tested by entering treatable vascular risk x APOE ε4 carriership into the model. Adjustment was made for age, educational level, consecutive measurement and interaction age x consecutive measurement. To study the effect of the separate components of treatable vascular risk, a similar regression model was built with all separate components (Supplement). In all models, the variables cognitive performance, consecutive measurement, age (years) and treatable vascular risk (points) were entered as continuous variables. Educational level and APOE ε4 carriership were entered as categorical variables. The level of statistical significance was set at 0.05. The linear multilevel analyses were performed using MLwiN Version 2.29 (Centre for Multilevel Modelling, University of Bristol, Bristol, UK) (18), the other analyses were performed using IBM SPSS Statistics 22.0 (IBM, Amonk, NY).

Sensitivity analyses

Various a priori-defined analyses were performed. First, the analyses were limited to persons aged 35-74 years without vascular history, because the FRS-CD was only validated in this age group (3). Second, to investigate the generalizability of our findings, analyses were repeated with two other risk scores based on the treatable components of the Framingham Risk Score for Coronary Heart Disease (FRS-CHD) and the SCORE risk system (5,6). Third, the analyses were repeated after exclusion of all APOE ε2 carriers (allele combinations ε2/ε2, ε2/ε3 and ε2/ε4) because the APOE ε2 allele appears to reduce the risk of Alzheimer’s disease (19). Finally, the analyses were repeated in a subset of the PREVEND cohort, the Groningen Random Sample, which had a similar prevalence of microalbuminuria (8%) and other cardiovascular risk factors as the general population (20).

 

Results

Study population

Overall, 3,601 participants completed the cognitive tests at multiple measurements: 2,431 (68%) participants at three measurements and 1,170 (32%) participants at two measurements. Eighteen (0.5%) participants were excluded because their educational level was not known and three (0.1%) participants because their age was younger than 35 years and their number too small to form a separate age group. Eight (0.2%) persons were excluded because of missing data on treatable vascular risk. Thus, the total study population included 3,572 persons with a mean (SD) age of 54 (11) years, 52% were men and 96% of Western-European descent (Table 1).

Longitudinal course of cognitive performance and treatable vascular risk

The mean (SD) total follow-up time was 5.5 (0.7) years. The mean (SD) cognitive performance of the total study population changed per consecutive measurement from 0.00 (0.79) at the first measurement to 0.15 (0.83) at second measurement and to 0.39 (0.82) at third measurement (Ptrend<.001). The change in cognitive performance per consecutive measurement was most clear in the age groups 35 to 44 years, 45 to 54 years, and 55 to 64 years (Table 2). Treatable vascular risk ranged from -5 to +17 points with a mean (SD) of 2 (4) points at the first measurement. Except for the age group 35 to 44 years, treatable vascular risk did not change statistically significantly per consecutive measurement (Table 2).

Table 1. Characteristics of the study population at the first measurement (baseline)

Table 1. Characteristics of the study population at the first measurement (baseline)

Abbreviations: HDL, high-density lipoprotein; SD, standard deviation; * APOE ε4 carriership included the allele combinations ε2/ε4, ε3/ε4 and ε4/ε4.

Table 2. Change in cognitive performance* and treatable vascular risk† across measurements per age group

Table 2. Change in cognitive performance* and treatable vascular risk† across measurements per age group

All values are noted as mean (SD). Abbreviations: SD, standard deviation; * Cognitive performance was measured as a composite score of two tests (z-score): the Ruff Figural Fluency Test (RFFT) and the Visual Association Test (VAT) (13,15); † Treatable vascular risk is based on the components of Framingham Risk Score for Cardiovascular Disease that are amenable to treatment and included diabetes mellitus, current smoker status, total cholesterol, HDL-cholesterol, systolic blood pressure and use of blood pressure lowering drugs (3).

 

Longitudinal change in cognitive performance and treatable vascular risk

Longitudinal change in cognitive performance was dependent on treatable vascular risk: the change in cognitive performance was negatively associated with treatable vascular risk (Figure 1). The mean change in cognitive performance between the first and third measurement was 0.46 (95%CI, 0.37 to 0.55; P<.001) in persons with the lowest treatable vascular risk whereas it was 0.28 (95%CI, 0.08 to 0.47; P=.006) in persons with the highest treatable vascular risk. The association between cognitive performance and treatable vascular risk was confirmed by multilevel analysis. Adjusted for age, educational level, consecutive measurement and interaction age x consecutive measurement, the multilevel regression model did not only show a statistically significant effect for treatable vascular risk (B-coefficient, -0.011; 95%CI, -0.019 to -0.003; P=.01) but also for the interaction between treatable vascular risk and consecutive measurement (Table 3). The change in cognitive performance between two measurements decreased with 0.004 per one-point increment of treatable vascular risk (B-coefficient, -0.004; 95%CI, -0.008 to 0.000; P=.05). This is comparable to the decrease in change in cognitive performance between two measurements per one-year increment of age (B coefficient, -0.006; 95%CI,  -0.008 to -0.004; P<.001) (Table 3).

Figure 1. Mean cognitive performance per measurement dependent on the treatable vascular risk at first measurement. Bars represent 95% confidence intervals. Cognitive performance was measured as a composite score of two tests (z-score): the Ruff Figural Fluency Test (RFFT) and the Visual Association Test (VAT) (13,15). Treatable vascular risk is based on the components of Framingham Risk Score for Cardiovascular Disease that are amenable to treatment and included diabetes mellitus, current smoker status, total cholesterol, HDL-cholesterol, systolic blood pressure and use of blood pressure lowering drugs (3).

Figure 1. Mean cognitive performance per measurement dependent on the treatable vascular risk at first measurement. Bars represent 95% confidence intervals. Cognitive performance was measured as a composite score of two tests (z-score): the Ruff Figural Fluency Test (RFFT) and the Visual Association Test (VAT) (13,15). Treatable vascular risk is based on the components of Framingham Risk Score for Cardiovascular Disease that are amenable to treatment and included diabetes mellitus, current smoker status, total cholesterol, HDL-cholesterol, systolic blood pressure and use of blood pressure lowering drugs (3).

Table 3. Longitudinal association of cognitive performance* on the treatable vascular risk†: multilevel linear analysis

Table 3. Longitudinal association of cognitive performance* on the treatable vascular risk†: multilevel linear analysis

Abbreviations: B, unstandardized B-coefficient; CI, confidence interval; * Cognitive performance was measured as a composite score of two tests (z-score): the Ruff Figural Fluency Test (RFFT) and the Visual Association Test (VAT) (13,15); † Treatable vascular risk is based on the components of Framingham Risk Score for Cardiovascular Disease that are amenable to treatment and included diabetes mellitus, current smoker status, total cholesterol, HDL-cholesterol, systolic blood pressure and use of blood pressure lowering drugs (3); ‡ Consecutive measurement; § For model 1: -2*log likelihood 16864.08; || For model 2: -2*log likelihood 16822.30; { For model 3: -2*log likelihood 16817.67.

 

Effect of APOE ε4 carriership

The effect of treatable vascular risk on cognitive performance was not modified by APOE ε4 carriership as there was no statistically significant interaction between treatable vascular risk and APOE ε4 carriership: B-coefficient for treatable vascular risk, -0.009 (95%CI,  -0.019 to 0.001; P=.07), for APOE ε4 carriership, 0.003 (95%CI, -0.048 to 0.054; P=.91), and for the interaction treatable vascular risk x APOE ε4 carriership, -0.003 (95%CI, -0.015 to 0.009; P=.62).

Association with separate components of treatable vascular risk

Cognitive performance was not only associated with treatable vascular risk but also with different components of treatable vascular risk. Adjusted for age, gender, educational level, consecutive measurement and interaction age x consecutive measurement, the full multilevel regression model showed that cognitive performance was negatively associated with diabetes mellitus (B coefficient, -0.11; 95%CI, -0.19 to -0.02; P=.01), current smoker (B-coefficient, -0.08; 95%CI, -0.13 to -0.04; P<.001) and hypertension (B-coefficient, -0.05; 95%CI,  -0.10 to 0.00; P=.03), and positively associated with HDL-cholesterol (B-coefficient, 0.08; 95%CI, 0.02 to 0.13; P=.005). However, the longitudinal change in cognitive performance was not dependent on any of the separate components of treatable vascular risk because there was no statistically significant interaction between separate components and consecutive measurement (P=.20) (Supplement).

Sensitivity analyses

Essentially similar results were found if the analyses of the association of cognitive performance with the treatable vascular risk were limited to persons aged 35 to 74 years without vascular history. If the analyses were repeated with treatable vascular risks based on treatable components of FRS-CHD or SCORE as independent variables, the negatively association between cognitive performance and treatable vascular risk was also found. If the analyses were repeated after exclusion of all APOE ε2 carriers, there was no interaction between treatable vascular risk and APOE ε4 carriership. Finally, the association of cognitive performance with treatable vascular risk was also found in the Groningen Random Sample (Supplement).

 

Discussion

In this large community-based study, cognitive performance was negatively associated with treatable vascular risk over a follow-up period of almost six years in persons aged 35 to 82 years old. As reported previously (12), cognitive performance increased across the measurements probably due to the repeated exposure to the cognitive tests. However, the change in cognitive performance was dependent on treatable vascular risk and was lower if treatable vascular risk was higher. In addition, our data suggested that the effect of treatable vascular risk on cognitive performance was comparable to the effect of age.
Our results were comparable to the findings of the National Aging Study (NAS) and the two studies from the National Alzheimer Coordinating Center (NACC) cohort (8-10). In all studies, poor cognitive performance was associated with high treatable vascular risk independent of age. However, our study differs from these studies in study population, duration of follow-up, APOE ε4 carriership and type of treatable vascular risk score. Whereas the other studies included specific populations of elderly people, our study showed this association in population that comprised both middle-aged and old persons. Furthermore, in the NAS and NACC studies the negative association of cognitive performance with treatable vascular risk was found over an average follow-up period of three years (8-10). Notably, our study adds that this association persisted after a period of almost six years. Comparable to one NACC study (10), our study also showed that the effect of treatable vascular risk factors on cognitive performance was not changed by APOE ε4 carriership whereas the two other studies did not evaluate the interaction of APOE ε4 carriership with treatable vascular risk (8,9). Moreover, in our study the treatable vascular risk was based on a general vascular risk score and not on a stroke-specific risk score which was used in the NAS and NACC studies (8-10). Therefore, vascular risk management programmes based on general vascular risk may not only prevent cardiac, cerebrovascular and peripheral vascular events but possibly also cognitive impairment. In addition, our findings from a study population of middle-aged and old persons support the hypothesis that the start of vascular risk management at late-life may be too late for effective prevention of cognitive impairment and dementia (21, 22).
Interestingly, our data suggested that the effect of treatable vascular risk on cognitive performance was comparable to the effect of age. This is in agreement with the finding of the NAS study that the association between cognitive performance and treatable vascular risk was almost as strong as that between cognitive performance and age (8). As a result, it may be estimated that one-point decrement of treatable vascular risk per year can probably gain one-year in cognitive age. One-point decrement of treatable vascular risk can be achieved by 10 mmHg reduction in systolic blood pressure or 1 mmol/L reduction in total cholesterol (3). These target values are usually achieved in clinical practice and randomized controlled trials (RCTs) (23, 24). Smoking cessation even results in three-points decrement of treatable vascular risk (3). Several studies did not only show that smoking is a risk factor for cognitive impairment, but also that smoking cessation decreased the risk of cognitive impairment to the risk of persons who have never smoked (25). So, smoking cessation is a good preventive measure and may compensate cognitive decline that occurs in three-years increment of age. Thus, a relevant decrease in vascular risk is probably feasible and is like to gain several years in cognitive age.
Recently, it was observed that over the past decades management of vascular risk factors has coincided with a decline in the prevalence of dementia (26). Our findings supported that vascular risk management may not only result in a lower incidence of cardiovascular disease but possibly also in a lower incidence of cognitive impairment and dementia. However, up till now, various RCTs have found inconsistent results about the effect of treatment of vascular risk factors on cognitive performance (27). Only the Syst-Eur trial suggested a protective effect of antihypertensive treatment on dementia in contrast to other trials (27,28). Similarly, intensified treatment of diabetes mellitus or cholesterol lowering treatment had no effect on cognitive performance in other large trials such as the ADVANCE study and the PROSPER trial (24, 27, 29). It is generally acknowledged that these negative findings may be explained by the use of a relatively insensitive cognitive test or short follow-up period (27). The FINGER, preDIVA and MAPT trials did not have these shortcomings (30-32). The FINGER trial showed that a multidomain intervention including treatment of vascular risk factors during two years could improve or maintain cognitive performance. However, the effect of treatment of vascular risk factors on cognitive performance per se was unclear as the multidomain intervention also included cognitive training (30). On the other hand, the preDIVA trial did not show a positive effect of the multidomain vascular intervention on cognitive performance, possibly because there was a similar reduction in cardiovascular risk in the intervention and control group (31). Similarly, the MAPT study did not found a difference in 3-year cognitive decline between control group and the multidomain intervention including physical activity, cognitive training and nutritional advice (32). Moreover, the trials included only old persons who were at risk for cognitive impairment (30-32). Considering our findings, starting vascular risk management in old age or risk groups may be too late for effective prevention of cognitive impairment and dementia (21, 22).

Some limitations of this study have to be noted. First, our study had an observational design whereas it is generally acknowledged that observational studies may give results that differ from subsequent RCTs on the same questions, and may overestimate treatment effects (33). However, RCTs with duration of four years or even longer seem hardly feasible due to high costs and the ethics of not treating vascular risk factors for a long time in placebo group (34, 35). Therefore, observational studies with a long follow-up period are still essential to gain more insight in the consequence of increased vascular risk in middle-age. Second, cognitive performance was measured with two cognitive tests in this study which may not evaluate all cognitive domains. However, the RFFT measures a wide range of different cognitive abilities such as initiation, planning, divergent reasoning, and the ability to switch between different tasks (13). In addition, because of its wide score range, the RFFT is not limited by a ceiling or floor effect and, thereby, sensitive to subtle changes in cognitive performance in young and old persons (13, 14). Furthermore, the VAT was added as a measure of memory (15). Although both tests are dependent on language and relatively specific measures of frontal network functions, semantic and episodic memory, these two tests combined reflect the cognitive domains commonly affected by Alzheimer’s disease and vascular dementia. Finally, in our study the cognitive performance increased across the measurements probably due to repeated exposure to the tests resulting in a practice effect (12). Practice effects appear in most, if not all, cognitive tests which assess various cognitive domains like memory, attention and executive functions (36). Practice effects can be ascribed to different factors such as memory of previous responses and learning test strategies, and could explain that people improve or maintain their cognitive performance despite a cognitive decline (36). However, in our study the association of cognitive performance with treatable vascular risk was adjusted for repeated consecutive measurement by entering the variable consecutive measurement and the interaction age x consecutive measurement in the model.
Despite these limitations, the present study also has several strengths. Our study was based on a large community-based cohort and included a large number of both middle-aged and elderly people whereas other longitudinal studies used selected populations of elderly (8-10). In addition, by using a (general) vascular risk score we explored the synergistic effects of vascular risk factors instead of focusing on a single risk factor. Risk scores have the advantage that multiple separate risk factors are weighted to generate optimal overall risk estimation for individual patients. Additionally, they yield a single variable that is the optimal estimate for overall cardiovascular burden, which limits the number of variables in small studies or multivariate analyses (3-6). Moreover, vascular risk scores are particularly valuable to identify increased vascular risk in middle-aged people because in this age group vascular risk factors often are only marginally elevated if considered separately but result in a clearly increased vascular risk if considered together (3-6).
In conclusion, in this large community-based cohort change in cognitive performance was associated with treatable vascular risk in both middle-aged and old people. Our data support the hypothesis that the start of vascular risk management at late-life may be too late for effective prevention of cognitive impairment and dementia.

 

Additional Contributions: The authors thank the PREVEND study group and the supporting staff of PREVEND for their role in the logistics of the study and the acquisition of the data used in this article.

Author Disclosures: All authors reported no disclosures.

Conflict of interest: There is no conflict of interest.

Ethical standards: The PREVEND study was approved by the medical ethics committee (METc) of Universitair Medical Center Groningen, Groningen, the Netherlands, and conducted in accordance with the guidelines of the Helsinki declaration.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

References

1.    Harrison SL, Ding J, Tang EY, et al. Cardiovascular disease risk models and longitudinal changes in cognition: A systematic review. PLoS One. 2014;9(12):e114431.
2.    Sperling RA, Aisen PS, Beckett LA, et al. Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the national institute on aging-Alzheimer’s association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7(3):280-292.
3.    D’Agostino RB S, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: The Framingham Heart study. Circulation. 2008;117(6):743-753.
4.    Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: A risk profile from the Framingham study. Stroke. 1991;22(3):312-318.
5.    Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97(18):1837-1847.
6.    Conroy RM, Pyorala K, Fitzgerald AP, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: The SCORE project. Eur Heart J. 2003;24(11):987-1003.
7.    Rodrigue KM, Rieck JR, Kennedy KM, et al. Risk factors for beta-amyloid deposition in healthy aging: vascular and genetic effects. JAMA Neurol. 2013;70(5):600-606.
8.    Brady CB, Spiro A,3rd, McGlinchey-Berroth R, Milberg W, Gaziano JM. Stroke risk predicts verbal fluency decline in healthy older men: Evidence from the normative aging study. J Gerontol B Psychol Sci Soc Sci. 2001;56(6):P340-6.
9.    Jefferson AL, Hohman TJ, Liu D, et al. Adverse vascular risk is related to cognitive decline in older adults. J Alzheimers Dis. 2015;44(4):1361-1373.
10.    Viswanathan A, Macklin EA, Betensky R, Hyman B, Smith EE, Blacker D. The influence of vascular risk factors and stroke on cognition in late life: Analysis of the NACC cohort. Alzheimer Dis Assoc Disord. 2015;29(4):287-293.
11.    Mahmoodi BK, Gansevoort RT, Veeger NJ, et al. Microalbuminuria and risk of venous thromboembolism. JAMA. 2009;301(17):1790-1797.
12.    van Eersel ME, Joosten H, Koerts J, Gansevoort RT, Slaets JP, Izaks GJ. Longitudinal study of performance on the Ruff Figural Fluency Test in persons aged 35 years or older. PLoS One. 2015;10(3):e0121411.
13.    Ruff R, Light R, Evans R. The Ruff Figural Fluency Test: A normative study with adults. Dev Neuropsychol. 1987;3:37-51.
14.    Izaks GJ, Joosten H, Koerts J, Gansevoort RT, Slaets JP. Reference data for the Ruff Figural Fluency Test stratified by age and educational level. PLoS One. 2011;6(2):e17045.
15.    Lindeboom J, Schmand B. Visual association test. Manual. PITS B V Leiden, The Netherlands. 2003.
16.    Visser ST, Schuiling-Veninga CC, Bos JH, de Jong-van den Berg LT, Postma MJ. The population-based prescription database IADB.nl: Its development, usefulness in outcomes research and challenges. Expert Rev Pharmacoecon Outcomes Res. 2013;13(3):285-292.
17.    Caselli RJ, Dueck AC, Locke DE, et al. Cerebrovascular risk factors and preclinical memory decline in healthy APOE epsilon4 homozygotes. Neurology. 2011;76(12):1078-1084.
18.    Rasbash J, Charlton C, Browne WJ, Healy M, Cameron B. MLwiN version 2.1. Centre for multilevel modelling, University of Bristol. 2009:http://www.bristol.ac.uk/cmm/software/mlwin/. Accessed on 14 August 2018.
19.    Bertram L, McQueen MB, Mullin K, Blacker D, Tanzi RE. Systematic meta-analyses of Alzheimer disease genetic association studies: The AlzGene database. Nat Genet. 2007;39(1):17-23.
20.    Lambers Heerspink HJ, Brantsma AH, de Zeeuw D, et al. Albuminuria assessed from first-morning-void urine samples versus 24-hour urine collections as a predictor of cardiovascular morbidity and mortality. Am J Epidemiol. 2008;168(8):897-905.
21.    Alzheimer’s Disease International. World Alzheimer repport 2014. Dementia and risk reduction: An analysis of protective and modifiable factors. London. 2014:1-104.
22.    Dorresteijn JA, Kaasenbrood L, Cook NR, et al. How to translate clinical trial results into gain in healthy life expectancy for individual patients. BMJ. 2016;352:i1548.
23.    SPRINT Research Group, Wright JT,Jr, Williamson JD, et al. A randomized trial of intensive versus standard blood-pressure control. N Engl J Med. 2015;373(22):2103-2116.
24.    Shepherd J, Blauw GJ, Murphy MB, et al. Pravastatin in elderly individuals at risk of vascular disease (PROSPER): A randomised controlled trial. Lancet. 2002;360(9346):1623-1630.
25.    Zhong G, Wang Y, Zhang Y, Guo JJ, Zhao Y. Smoking is associated with an increased risk of dementia: A meta-analysis of prospective cohort studies with investigation of potential effect modifiers. PLoS One. 2015;10(3):e0118333.
26.    Satizabal CL, Beiser AS, Chouraki V, Chene G, Dufouil C, Seshadri S. Incidence of dementia over three decades in the Framingham heart study. N Engl J Med. 2016;374(6):523-532.
27.    Ligthart SA, Moll van Charante EP, Van Gool WA, Richard E. Treatment of cardiovascular risk factors to prevent cognitive decline and dementia: A systematic review. Vasc Health Risk Manag. 2010;6:775-785.
28.    Forette F, Seux ML, Staessen JA, et al. The prevention of dementia with antihypertensive treatment: New evidence from the systolic hypertension in Europe (syst-eur) study. Arch Intern Med. 2002;162(18):2046-2052.
29.    ADVANCE Collaborative Group, Patel A, MacMahon S, et al. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med. 2008;358(24):2560-2572.
30.    Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial. Lancet. 2015;385(9984):2255-2263.
31.    Moll van Charante EP, Richard E, Eurelings LS, et al. Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA): A cluster-randomised controlled trial. Lancet. 2016;388(10046):797-805.
32.    Andrieu S, Guyonnet S, Coley N, et al. Effect of long-term omega 3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function in elderly adults with memory complaints (MAPT): a randomised, placebo-controlled trial. Lancet Neurol. 2017;16:377-389.
33.    Hemkens LG, Contopoulos-Ioannidis DG, Ioannidis JP. Agreement of treatment effects for mortality from routinely collected data and subsequent randomized trials: Meta-epidemiological survey. BMJ. 2016;352:i493.
34.    Richard E, Andrieu S, Solomon A, et al. Methodological challenges in designing dementia prevention trials – the European dementia prevention initiative (EDPI). J Neurol Sci. 2012;322(1-2):64-70.
35.    Richard F, Pasquier F. Can the treatment of vascular risk factors slow cognitive decline in Alzheimer’s disease patients? J Alzheimers Dis. 2012;32(3):765-772.
36.    Calamia M, Markon K, Tranel D. Scoring higher the second time around: Meta-analyses of practice effects in neuropsychological assessment. Clin Neuropsychol. 2012;26(4):543-570.