In this analysis of MCI and mild-to-moderate AD trials, we found that participant retention was negatively associated with trial duration. The proportions of participants in the placebo group completing 6-, 12-, and 18-month trials were estimated to be 85.2%, 80.0%, and 73.3% for mild-to-moderate AD trials and 91.9%, 84.2%, and 71.3% for MCI trials, respectively. Trials with a duration greater than 6 months that incorporate common attrition estimates of 10% per year or 20% overall, therefore, may be at risk of being underpowered. This is concerning, since AD trials have increased in length over time [8].

While trials in both diagnostic groups presented similar patterns of attrition with increasing duration, mild-to-moderate AD trials were estimated to have lower completion estimates at 6- and 12-months while MCI trials were estimated to have lower completion rates at 18 months. Age may be a risk factor for participant dropout [9, 10], and participants in the mild-to-moderate AD trials are likely to be older than those in MCI trials [11]. AD is also a progressive disease. Previous examinations of data from NIH-funded Alzheimer’s Disease Research Centers (ADRC) identified worsening cognitive impairment as a risk factor for dropout [12]. This may be most relevant in longer MCI trials, where cognitive decline may be accompanied by the onset of functional impairment.

AD trial participants must co-enroll with a study partner [7]. Longer trials bring an added burden for dyads and may require more careful planning and resources to reduce modifiable barriers to trial completion [13]. For example, a previous study observed an association between the number of retention tactics used at NIH-funded ADRCs and rates of retention at 1 and 2 years [14]. Trials that require longer participation may also benefit from adjusting other study design features like reducing/replacing high-burden assessments (e.g., lumbar puncture) and increasing motivating factors such as returning test results [15] and even financial bonuses for trial completion [16].

While our study underscores the association between trial duration and retention, trial duration decisions are complex. For example, financial considerations may influence duration decisions [17]. Trials with objectives related to pharmacodynamics and pharmacokinetics may require shorter durations. Trials that enroll participants at earlier stages of the disease (e.g., MCI or early AD) may observe less change over time, therefore requiring longer durations [18, 19]. Our results suggest relatively similar associations between duration and retention between MCI and dementia trials. Symptomatic agents may produce measurable effects more rapidly, requiring shorter durations, while disease-modifying treatments may require a longer duration to detect measurable attenuation of AD progression [18]. Though the association with duration was similar between symptomatic and disease-modifying trials here, we note that symptomatic trials were heavily skewed toward shorter durations. We also found similar associations between phase II and phase III trials and between single- and multi-site trials. Overall, trial duration should not be determined based on anticipated retention rates; however, our findings suggest that duration does impact retention and can provide valuable insights into projected retention rates for investigators during the design stage.

The retention estimates in this study may not be representative of all MCI and AD trials due to publication bias. We did not adjust the models in our study for other potential confounding factors that may be associated with retention, such as personal characteristics of the participants (e.g., study partner types, race and ethnicity, education) or trial design features (e.g., number of visits, alternate allocation, mode of treatment administration). Trials that were terminated before completion were excluded from our analysis, which could have biased our estimates, particularly if high dropout contributed to early termination. Some of the trials included in our analyses were ongoing during the COVID-19 pandemic, potentially resulting in a change in their protocols, including adopting a decentralized approach or extending trial duration. We were unable to assess how such changes affected retention estimates.