A Longitudinal Study of Alzheimer's disease Treatment Efficacy and Predictors of Cognitive Decline

The inspiration of this project came as a result of our experience at a senior center, where we conducted activities to improve the quality of life of elderly dementia patients. We saw firsthand how dementia can impact memory, causing immense difficulties maintaining personal relationships with loved ones.

Background

Alzheimer's Disease (AD) is a neurodegenerative disease and the leading cause of dementia, accounting for 60%-80% of all of the 59 million dementia cases (World Health Organization, 2025) and is expected to rise, per Figure 1. The leading hypothesis states that the decline is caused by the aggregation of Amyloid Beta plaques, which can cause inflammation and result in neuron death (Hardy et al, 2016).

Objective

Compare 3 distinct treatment mechanisms via meta analysis

Model cognitive decline of AD patients and develop app to predict cognitive decline over time.

The first phase compares how treatments that target the amyloid plaques, treatments that attack the symptoms, and cognitive stimulation treatments, impact cognitive decline over time via a meta analysis, per Table 1. There is no cure for AD , and these treatments mitigate disease progression and effect (Petrella, 2013).

The second phase implements machine learning in order to compare factors impacting disease progression, as well as predicting the future cognitive deterioration of Alzheimer’s patients and providing a treatment recommendation. The main problem that these two components solve is to give more information for more informed clinical decisions.

The Meta-Analysis and ML models together can help medical professionals and patients make informed decisions.

Meta-Analysis

Efficacy was determined by graphing improvement in the 70pt Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) score and the 30pt Mini-Mental State Examination(MMSE), which primarily assess recall, language, and praxis (Balsis et. al, 2015).

Data was sourced from clinical trials (PubMed, ClinicalTrials.gov, NIH) with the following inclusion criteria: only one drug used, diagnosis of mild AD, provision of longitudinal MMSE/ADAS-Cog data, mean age >60, and sample size >28. Trials were excluded if participants used more than one drug, or were not in English.

Analysis of the data was done by first comparing the drugs within a certain mechanism to determine internal similarity of performance. Afterwards, the individual drugs were grouped by mechanism and compared against other mechanisms. Performance was quantified by using linear regression, giving the Pearson-Coefficient value and plotting the slope and its 95% CI to determine longitudinal effectiveness and variance.

Machine Learning

ADNI data, led by Michael Weiner, measures the progression of dementia in patients across North America(Weiner et. al, 2024) and was utilized to train the ML models. After the data request was accepted, necessary data was isolated and cleaned using python’s pandas package.

KMeans clustering was utilized to analyze the effect of baseline characteristics and demographics on cognitive deterioration. The elbow method graph (Figure 2), shows how many clusters minimizes dispersion and overfitting, found to be 4. ADAS-Cog rate of change, an untrained variable, was then compared across the 4 resulting clusters to determine which clusters could be compared to understand feature importances.

The prediction website’s client-side was programmed using Streamlit. A Random-Forest-Regression predicts a patient's cognitive impairment trajectory, while a Random-Forest-Classifier predicts their decline-rate category. For treatment recommendation, propensity-score-matching, which matches a patient to similar training-data patients from different treatment groups, was utilized to determine which treatment group minimizes cognitive decline.

Meta-Analysis

As per figure 3, 392 randomized-controlled trials were obtained via searching the treatment name, and filtered through the inclusion criteria to give 25 total trials.

Figure 4 shows that the treatments within each mechanism have very similar performance, meaning they can be grouped and compared against other mechanisms.

Figure 5 shows that cognitive therapies have the most variation in performance, evidenced by large error bars in the scatterplot and the largest confidence interval in slope of decline. Conversely, anti-amyloid has the most linear performance, with the highest R2 value of 0.723, and the smallest error bars. Figure 6 shows that Anti-Amyloids have the lowest CI range, and that symptomatic treatments show statistically significant improvements over them, due to no overlap between error bars. Moreover, Symptomatic treatments have a slope of -0.0881pts/week and a CI range of 0.056pts/week, meaning that they have relatively little variation.

Figure 7 validates that cognitive therapies have large variation and anti-amyloid treatments have the most linear performance over time.

Overall, the analysis showed symptomatic treatments were associated with greater reduction of cognitive decline over time compared to anti-amyloid treatments, and cognitive therapies have large amounts of variation compared to the other treatments.

ML Model

Per table 2, the fastest and slowest declining clusters were found to be significantly different with a p-value of 0.0073 after an independent t-test. This way, these two clusters could be compared in order to compare feature importance. Holistic biological measures such as total brain volume had higher Cohen's d values (difference in standard deviation units), between the two clusters. Cohen's d for total brain volume was 4.1, and compared to baseline score (0.6) or hippocampal volume (1.9), this was a major difference. Next, education in years, despite having a moderately large effect size (1.2), was separated by only 2 years between the fastest and slowest declining groups. Lastly, sex composition had a consistent trend as decline rate increased, showcasing how females are much more likely to be faster decliners than males, which is consistent with current studies (Emrani & Sundermann, 2025).

For the predictive model, a random forest regression model was fitted on a training set and evaluated on the testing set. The R2 value of the model was found to be 0.7, meaning that the model explained 70% of the variance in the unseen training set. However, this high R2 was being influenced strongly by baseline score, and the other features were simply fine tuning it. In order to get a clearer picture of the predictive power of the training features, change in ADAS-Cog scores was predicted, leading to an R2 of 0.3, which is typical for clinical data (Gupta et. al, 2024). Because of this, a random forest classifier was fitted on the same data in order to predict which category of decliners a new patient would belong to, which had a much higher accuracy (R2=0.7).

Meta-Analysis

The Meta Analysis helps patients weigh the risk between targeting disease progression and managing the symptoms of the disease, due to anti amyloid treatments having a tendency for severe side effects (Schindler et al, 2025). These results also support other pathologies for Alzheimer’s, suggesting that the cognitive decline associated with Alzheimer's might be less correlated with the concentration and accumulation of amyloid plaques. This supports research finding that the major cause of AD associated cognitive decline is not the amyloid plaques, with other hypotheses ranging from lack of specific neurotransmitters (Francis et al, 1999) to the hyperphosphorylation of tau (Macconi et al, 2010). This analysis is limited as there is relatively little regarding anti-amyloid treatments due to their clinical recency, with the first therapy in this class receiving approval only in 2021.

Machine Learning Model

The clustering analysis showcases how baseline brain volume as opposed to baseline cognitive score can be tracked by clinicians to gain a better understanding of a patient’s future cognitive deterioration. Also, it establishes that even 2 more years of education can separate the fast from slow decliners, showcasing the importance of cognitive stimulation, giving individuals guidance on how to prevent and/or decrease the severity of dementia. Lastly, it suggests association of sex with the severity of Alzheimer-associated cognitive decline. The publicly accessible website is intended to allow doctors and caregivers to make informed, personalized decisions regarding patients, giving them an idea of what to expect and how to prepare appropriately.

The results for the meta analysis could be strengthened by verifying results through an analysis of more cognitive tests, such as the Montreal Cognitive Assessment. Future research could analyze the efficacy of different clinical stages of AD (Pinto et al, 2018), which would offer insights into how treatment efficacy changes across disease progression. To build upon the machine learning analysis, analyzing the extent to which biological changes such as increases in biomarker concentrations correlate with rate of cognitive decline could provide insights into their association with Alzheimer’s disease progression. Additionally, the predictive website can be implemented in clinical practice.

The project would not have been possible in its current state without the help of the following people: Dr. Iaci Soares for keeping the reviewers accountable and on track, Mx. Dallas Mythril for her help in editing and reviewing the work, Mr. Merrick Fanning for his help for all statistics related queries, and advice on statistical analysis, the Team Calgary coordinators for their continued support and feedback.

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

Bibliography for Meta Analysis

Citations

Alzheimer’s: Medicines help manage symptoms and slow decline [Internet]. Mayo Foundation for Medical Education and Research; 2024 [cited 2026 Feb 15]. Available from: https://www.mayoclinic.org/diseases-conditions/alzheimers-disease/in-depth/alzheimers/art-20048103

Benge JF, Balsis S, Geraci L, Massman PJ, Doody RS. How well do the Adas-cog and its subscales measure cognitive dysfunction in alzheimer’s disease? [Internet]. U.S. National Library of Medicine; 2009 [cited 2026 Feb 5]. Available from: https://www.researchgate.net/profile/Lisa-Geraci/publication/26704106_How_Well_Do_the_ADAS-cog_and_its_Subscales_Measure_Cognitive_Dysfunction_in_Alzheimer’s_Disease/links/554115d90cf2b790436bc644/How-Well-Do-the-ADAS-cog-and-its-Subscales-Measure-Cognitive-Dysfunction-in-Alzheimers-Disease.pdf

Birks JS, Harvey RJ. Donepezil for Dementia due to Alzheimer's Disease [Internet]. U.S. National Library of Medicine; 2018 [cited 2026 Feb 5]. Available from: https://pubmed.ncbi.nlm.nih.gov/29923184/

Comas-Herrera A, Knapp M. Cognitive Stimulation Therapy (CST): Summary of evidence on cost-effectiveness [Internet]. 2016 [cited 2026 Feb 6]. Available from: https://www.england.nhs.uk/wp-content/uploads/2018/01/dg-cognitive-stimulation-therapy.pdf

Facts & figures - Alzheimer's Disease [Internet]. 2025 [cited 2026 Feb 15]. Available from: https://www.brightfocus.org/alzheimers/facts-figures/#:~:text=The%20percentage%20of%20people%20with,3

Gong C-X, Iqbal K. Hyperphosphorylation of microtubule-associated protein tau: A promising therapeutic target for Alzheimer’s disease [Internet]. 2009 [cited 2026 Feb 15]. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC2656563/

Hampel H, Hardy J, Blennow K, Chen C, Perry G, Kim SH, et al. The amyloid-β pathway in Alzheimer's disease [Internet]. Nature Publishing Group; 2021 [cited 2026 Feb 5]. Available from: https://www.nature.com/articles/s41380-021-01249-0

Herrup K. The case for rejecting the amyloid cascade hypothesis [Internet]. 2015 [cited 2026 Feb 16]. Available from: http://behavioralhealth2000.com/wp-content/uploads/sites/3661/2017/02/The-case-for-rejecting-the-amyloid-cascade-hypothesis2.pdf

How is Alzheimer's disease treated? | National Institute on Aging [Internet]. 2023 [cited 2026 Feb 6]. Available from: https://www.nia.nih.gov/health/alzheimers-treatment/how-alzheimers-disease-treated

Mann SK, Malhi NK. Repetitive transcranial magnetic stimulation [Internet]. U.S. National Library of

Maccioni RB, Farías G, Morales I, Navarrete L. The revitalized tau hypothesis on Alzheimer's disease. Archives of medical research. 2010 Apr 1;41(3):226-31.Francis PT, Palmer AM, Snape M, Wilcock GK. The cholinergic hypothesis of Alzheimer’s disease: a review of progress. Journal of Neurology, Neurosurgery & Psychiatry. 1999 Feb 1;66(2):137-47.

Medicine; 2023 [cited 2026 Feb 5]. Available from: https://www.ncbi.nlm.nih.gov/books/NBK568715/

Mini-mental status exam (MMSE) [Internet]. 2021 [cited 2026 Feb 6]. Available from: https://www.psychdb.com/cognitive-testing/MMSE

Mitchell AJ. A meta-analysis of the accuracy of the mini-mental state examination in the detection of dementia and mild cognitive impairment [Internet]. 2009 [cited 2026 Feb 6]. Available from: https://www.researchgate.net/publication/351163114_httpswwwsciencedirectcomsciencearticleabspiiS1051200421000968

Morris GP, Clark IA, Vissel B. Inconsistencies and Controversies Surrounding the Amyloid Hypothesis of Alzheimer’s Disease [Internet]. 2014 [cited 2026 Feb 16]. Available from: https://link.springer.com/content/pdf/10.1186/s40478-022-01441-5.pdf

Mukhopadhyay S, Banerjee D. A primer on the evolution of aducanumab: The first antibody approved for treatment of Alzheimer's disease [Internet]. U.S. National Library of Medicine; 2021 [cited 2026 Feb 5]. Available from: https://pubmed.ncbi.nlm.nih.gov/34366359/

OpenAI. ChatGPT (Feb 14 version) [Large language model]. 2023. Available from: https://chat.openai.com/chat

Rockwood K, Fay S, Gorman M, Carver D, Graham JE. The clinical meaningfulness of Adas-cog changes in Alzheimer's disease patients treated with Donepezil in an open-label trial [Internet]. U.S. National

Shasteen ME, Wurzelmann MK, McGregor AJ, Raukar NP. Heart Breaking Differences: A Narrative Review of Sex and Gender Disparities in Sports-Related Sudden Cardiac Death. Clin Ther. 2024;46(12):982-987. doi: 10.1016/j.clinthera.2024.11.002.

Schindler, S. E., Musiek, E. S., & Morris, J. C. (2025). Anti-amyloid treatments: Why we think they are worth it. Alzheimer's & dementia (New York, N. Y.), 11(1), e70055. https://doi.org/10.1002/trc2.70055

Library of Medicine; 2007 [cited 2026 Feb 5]. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC2034585/

Van Dyck CW, Swanson CJ, Aisen P, Bateman RJ, Chen C, Gee M, et al. The New England Journal of Medicine [Internet]. 2022 [cited 2026 Feb 6]. Available from: https://www.nejm.org/doi/pdf/10.1056/nejmoa2212948

References

Aguirre, E., Woods, R. T., Spector, A., & Orrell, M. (2013). Cognitive stimulation for dementia: A systematic review of the evidence of effectiveness from randomized controlled trials. Ageing Research Reviews, 12(1), 253–262. https://doi.org/10.1016/j.arr.2012.07.001

Chen, T., O'Gorman, J., Castrillo-Viguera, C., Rajagovindan, R., Tian, Y., Patel, D., ... Sandrock, A. (2024). Results from the long-term extension of PRIME: A randomized Phase 1b trial of aducanumab. Alzheimer's & Dementia, 20(5), 3406–3415. https://doi.org/10.1002/alz.13755

Cooper, C., Mukadam, N., Katona, C., Lyketsos, C. G., Ames, D., Rabins, P., ... Livingston, G. (2012). Systematic review of the effectiveness of non-pharmacological interventions to improve quality of life of people with dementia. International Psychogeriatrics, 24(6), 856–870. https://doi.org/10.1017/S1041610211002614

Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini-mental state": A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12(3), 189–198. https://doi.org/10.1016/0022-3956(75)90026-6

Gates, N. J., Sachdev, P. S., Fiatarone Singh, M. A., & Valenzuela, M. (2011). Cognitive and memory training in adults at risk of dementia: A systematic review. BMC Geriatrics, 11, 55. https://doi.org/10.1186/1471-2318-11-55

Gormley, N., Lyons, D., & Howard, R. (2001). Behavioural management of aggression in dementia: A randomized controlled trial. Age and Ageing, 30(2), 141–145. https://doi.org/10.1093/ageing/30.2.141

Joshi, A. D., Pontecorvo, M. J., Lu, M., Skovronsky, D. M., Mintun, M. A., & Devous, M. D. (2015). A semiautomated method for quantification of F18 Florbetapir PET images. Journal of Nuclear Medicine, 56(11), 1736–1741. https://doi.org/10.2967/jnumed.115.158402

Klunk, W. E., Koeppe, R. A., Price, J. C., Benzinger, T. L., Devous, M. D., Jagust, W. J., ... Mathis, C. A. (2015). The Centiloid Project: Standardizing quantitative amyloid plaque estimation by PET. Alzheimer's & Dementia, 11(1), 1–15. https://doi.org/10.1016/j.jalz.2014.07.003

Koch, G., Casula, E. P., Bonnì, S., Borghi, I., Assogna, M., Di Lorenzo, F., ... Martorana, A. (2025). Effects of 52 weeks of precuneus rTMS in Alzheimer's disease patients: A randomized trial. Alzheimer's Research & Therapy, 17(1), 69. https://doi.org/10.1186/s13195-025-01709-7

Koch, G., Casula, E. P., Bonnì, S., Borghi, I., Assogna, M., Minei, M., ... Martorana, A. (2022). Precuneus magnetic stimulation for Alzheimer's disease: A randomized, sham-controlled trial. Brain, 145(11), 3776–3786. https://doi.org/10.1093/brain/awac285

Mapelli, D., Di Rosa, E., Nocita, R., & Sava, D. (2013). Cognitive stimulation in patients with dementia: Randomized controlled trial. Dementia and Geriatric Cognitive Disorders Extra, 3(1), 263–271. https://doi.org/10.1159/000353457

Mittelman, M. S., Haley, W. E., Clay, O. J., & Roth, D. L. (2006). Improving caregiver well-being delays nursing home placement of patients with Alzheimer disease. Neurology, 67(9), 1592–1599. https://doi.org/10.1212/01.wnl.0000242727.81172.91

Morris, J. C. (1993). The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology, 43(11), 2412–2414. https://doi.org/10.1212/wnl.43.11.2412-a

Olazarán, J., Reisberg, B., Clare, L., Cruz, I., Peña-Casanova, J., del Ser, T., ... Muñiz, R. (2010). Nonpharmacological therapies in Alzheimer's disease: A systematic review of efficacy. Dementia and Geriatric Cognitive Disorders, 30(2), 161–178. https://doi.org/10.1159/000316119

Raskind, M. A., Peskind, E. R., Truyen, L., Kershaw, P., & Damaraju, C. V. (2004). The cognitive benefits of galantamine are sustained for at least 36 months: A long-term extension trial. Archives of Neurology, 61(2), 252–256. https://doi.org/10.1001/archneur.61.2.252

Reisberg, B., Auer, S. R., & Monteiro, I. M. (1997). Behavioral pathology in Alzheimer's disease (BEHAVE-AD) rating scale. International Psychogeriatrics, 9(S1), 301–308. https://doi.org/10.1097/00019442-199911001-00147

Rogers, S. L., Doody, R. S., Mohs, R. C., Friedhoff, L. T., & the Donepezil Study Group. (1998). Donepezil improves cognition and global function in Alzheimer disease: A 15-week, double-blind, placebo-controlled study. Archives of Internal Medicine, 158(9), 1021–1031. https://doi.org/10.1001/archinte.158.9.1021

Sevigny, J., Chiao, P., Bussière, T., Weinreb, P. H., Williams, L., Maier, M., ... Sandrock, A. (2016). The antibody aducanumab reduces Aβ plaques in Alzheimer's disease. Nature, 537(7618), 50–56. https://doi.org/10.1038/nature19323

Spector, A., Thorgrimsen, L., Woods, B., Royan, L., Davies, S., Butterworth, M., & Orrell, M. (2003). Efficacy of an evidence-based cognitive stimulation therapy programme for people with dementia. British Journal of Psychiatry, 183(3), 248–254. https://doi.org/10.1192/bjp.183.3.248

Swanson, C. J., Zhang, Y., Dhadda, S., Wang, J., Kaplow, J., Lai, R. Y. K., ... Cummings, J. L. (2021). A randomized, double-blind, phase 2b proof-of-concept clinical trial in early Alzheimer's disease with lecanemab, an anti-Aβ protofibril antibody. Alzheimer's Research & Therapy, 13(1), 80. https://doi.org/10.1186/s13195-021-00813-8

Teri, L., Gibbons, L. E., McCurry, S. M., Logsdon, R. G., Buchner, D. M., Barlow, W. E., ... LaCroix, A. Z. (2003). Exercise plus behavioral management in patients with Alzheimer disease: A randomized controlled trial. JAMA, 290(15), 2015–2022. https://doi.org/10.1001/jama.290.15.2015

Teri, L., Logsdon, R. G., Peskind, E., Raskind, M., Weiner, M. F., Tractenberg, R. E., ... Thal, L. J. (2000). Treatment of agitation in AD: A randomized, placebo-controlled clinical trial. Neurology, 55(9), 1271–1278. https://doi.org/10.1212/wnl.55.9.1271

van Dyck, C. H., Swanson, C. J., Aisen, P., Bateman, R. J., Chen, J., Gee, M., ... Sandrock, A. (2025). Long-term safety and efficacy of lecanemab in early symptomatic Alzheimer disease: A randomized clinical trial. Alzheimer's & Dementia, 41(3), 355–380. https://doi.org/10.1002/alz.13959

Woods, B., Aguirre, E., Spector, A. E., & Orrell, M. (2012). Cognitive stimulation to improve cognitive functioning in people with dementia. Cochrane Database of Systematic Reviews, 2, CD005562. https://doi.org/10.1002/14651858.CD005562.pub2

Yamanaka, K., Kawano, Y., Noguchi, D., Nakaaki, S., Watanabe, N., Amano, T., & Spector, A. (2013). Effects of cognitive stimulation therapy Japanese version (CST-J) for people with dementia: A single-blind, controlled clinical trial. Aging & Mental Health, 17(5), 579–586. https://doi.org/10.1080/13607863.2013.777395

Yesavage, J. A., Brink, T. L., Rose, T. L., Lum, O., Huang, V., Adey, M. B., & Leirer, V. O. (1983). Development and validation of a geriatric depression screening scale: A preliminary report. Journal of Psychiatric Research, 17(1), 37–49. https://doi.org/10.1016/0022-3956(82)90033-4

Zhang, Z.-X., Hong, Z., Wang, Y.-P., He, L., Wang, N., Zhao, Z.-X., ... Strohmaier, C. (2012). Rivastigmine patch in Chinese patients with probable Alzheimer's disease: A 24-week, randomized, double-blind parallel-group study. Dementia and Geriatric Cognitive Disorders, d33(2-3), 122–131. https://doi.org/10.1159/000338453

Machine Learning References

Alzheimer’s Disease Neuroimaging Initiative. (n.d.). ADNI Database. Retrieved January 6, 2026, from

adni.loni.usc.edu

Beheshti, N. (2022, March 2). Random Forest Regression. Towards Data Science. Retrieved April 10, 2026, from

https://towardsdatascience.com/random-forest-regression-5f605132d19d/

Bukkawar, L. (2025, January 11). Understanding K-Means and K-Means++: A Comprehensive Guide. Medium. Retrieved April 20, 2026, from https://medium.com/@laakhanbukkawar/understanding-k-means-and-k-means-a-comprehensive-guide-4b288a6bf218

Emrani, S., & Sundermann, E. E. (2025). Sex/gender differences in the clinical trajectory of Alzheimer’s disease: Insights into diagnosis and cognitive reserve. Elsevier. Retrieved April 20, 2026, from

https://www.sciencedirect.com/science/article/pii/S009130222500010X

Gupta, A., Ganti, L., & Stead, T. S. (2024, October 26). Determining a Meaningful R-squared Value in Clinical Medicine. Academic Medicine & Surgery. Retrieved April 27, 2026, from https://academic-med-surg.scholasticahq.com/article/125154-determining-a-meaningful-r-squared-value-in-clinical-medicine

Scikit Learn. (n.d.). KMeans. Retrieved April 10, 2026, from

https://scikit-learn.org/stable/modules/generated/sklearn.cluster.KMeans.html

Scikit Learn. (n.d.). RandomForestClassifier. Retrieved April 10, 2026, from https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html

Scikit Learn. (n.d.). Ridge. Retrieved April 10, 2026, from https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Ridge.html

Statistics How To. (n.d.). Benjamini-Hochberg Procedure. Retrieved April 10, 2026, from https://www.statisticshowto.com/benjamini-hochberg-procedure/

Weiner, M. W., et al. (2024). Overview of Alzheimer's Disease Neuroimaging Initiative and future clinical trials. Alzheimer's & Dementia, 21. Retrieved December 12, 2025, from

https://doi.org/10.1002/alz.14321