How many is enough? Sample size estimation for imaging‐based outcomes in clinical trials of neurodegenerative disease

Abstract

AbstractBackgroundEffective treatments for neurodegenerative diseases are a global health priority. Appropriately powered clinical trials are critical to showing evidence of efficacy ‐ these are expensive and take years to complete. Depending on the study aims, imaging biomarkers can provide high precision measurements that reflect underlying pathology or downstream effects and may help in reducing the number of participants and/or duration of the trial. Identifying how to design trials that make the most efficient use of imaging endpoints, while maintaining sufficient statistical power to detect the desired therapeutic effect, will be critical to reduce approval time for effective therapies.MethodSample size estimates were computed for a wide range of imaging biomarkers, trial designs, and target therapeutic effects. We examined how sample size estimates were affected by the target population (inclusion/exclusion criteria), visit schedule, and the rate of participant dropout. From the resulting trial designs, appropriate formulae for sample size estimation were derived. These formulae require estimates of a number of parameters including both the mean rate of change over time in the outcome measure and the related variances. These were obtained from natural history studies, including ADNI, DIAN, and TRACK‐HD. The precision of these estimates is calculated by bootstrap resampling.ResultAcross target populations, imaging biomarkers produced lower sample sizes to detect a clinically relevant slowing of change compared to conventional primary endpoints, particularly in shorter duration (i.e. 1‐2 years) trials. Imaging biomarkers that incorporated longitudinal information into the measurements were often the best performers. If the participant dropout rate is non‐negligible, this can result in noticeable increases in sample size, which can be partially mitigated by including interim visits into the analysis. The precision of sample sizes estimates can vary substantially between biomarkers and should be considered when making decisions on trial design.ConclusionImaging based outcomes can play an important role in providing evidence of disease modification in trials, particularly in primary and secondary prevention studies. Accurate sample size estimates for the desired treatment effect are critical for the success of these studies.