Premium
Identifying predictive factors of patient dropout in Alzheimer’s disease clinical trials
Author(s) -
Crimin Kimberly,
Allen Patricia J,
Abba Iman,
Ahlberg Corinne,
Benz Luke,
Lau Hiuyan,
Liu Jingshu,
Melhem Fareed,
Fisseha Nahome,
Florian Hana
Publication year - 2021
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.052361
Subject(s) - dropout (neural networks) , clinical trial , medicine , disease , randomized controlled trial , machine learning , computer science
Background Improving retention of clinical trial participants plays a vital role in developing treatment and prevention strategies for Alzheimer’s disease (AD). Participant retention is critical to maintaining statistical power, preventing scientific error, and minimizing bias in research (Grill JD, et al., 2020). The purpose of this analysis was to identify factors that are predictive of participant dropout in AD trials using pooled data from randomized clinical trials run on the Medidata platform. These factors can be used to improve operational aspects of future trial design as well as to support patients with prevention strategies during future trials. Method 8,103 AD patients from 7 completed Phase 3 clinical trials were included in the analysis. The follow‐up time across studies was 1.4 ± 1.6 years (m ± sd) and the dropout rate was 21.2 ± 10.8%. A limited set of clinical and operational data were standardized across trials using proprietary machine learning algorithms combined with human review. Factors were assessed by survival analysis using Cox proportional hazards modeling. A 70/30 train‐test split was used to evaluate model performance in the prediction at baseline of dropouts within 3, 6, and 12 months of the trial. Additional models specific to dropouts due to reasons other than adverse events (AE) or death were also explored. Result The models were observed to predict dropouts at an acceptable level above random guessing with an ROC‐AUC ≥ 0.60 within 3, 6 and 12 months. However, the models typically overestimated the probability of dropout, leading to more false positives and reduced precision overall (PR‐AUC < 0.24). Several patient and site factors were associated with increased risk of dropout, including age, race, certain medical histories and AEs (e.g. anxiety), friend as caregiver, low surrounding population density & the size of the clinical staff at the site. Significant interactions between factors were also identified. Conclusion This study offers additional evidence that a combination of patient‐ and site‐specific factors influence risk of patient dropout in AD trials. These results both confirmed hypotheses of subject matter experts and suggest additional factors for consideration in planning future AD clinical trials, including supportive care for patients.