Multiple imputation approaches for epoch-level accelerometer data in trials.

Clinical trials that investigate physical activity interventions often use accelerometers to measure step count at a very granular level, for example in 5-second epochs. Participants typically wear the accelerometer for a week-long period at baseline, and for one or more week-long follow-up periods after the intervention. The data is aggregated to provide daily or weekly step counts for the primary analysis. Missing data are common as participants may not wear the device as per protocol. Approaches to handling missing data in the literature have defined missingness on the day level using a threshold on daily weartime, which leads to loss of information on the time of day when data are missing. We propose an approach to identifying and classifying missingness at the finer epoch-level and present two approaches to handling missingness using multiple imputation. Firstly, we present a parametric approach which accounts for the number of missing epochs per day. Secondly, we describe a non-parametric approach where missing periods during the day are replaced by donor data from the same person where possible, or data from a different person who is matched on demographic and physical activity-related variables. Our simulation studies show that the non-parametric approach leads to estimates of the effect of treatment that are least biased while maintaining small standard errors. We illustrate the application of these different multiple imputation strategies to the analysis of the 2017 PACE-UP trial. The proposed framework is likely to be applicable to other digital health outcomes and to other wearable devices.

A comparison of covariate adjustment approaches under model misspecification in individually randomized trials.

Adjustment for baseline covariates in randomized trials has been shown to lead to gains in power and can protect against chance imbalances in covariates. For continuous covariates, there is a risk that the the form of the relationship between the covariate and outcome is misspecified when taking an adjusted approach. Using a simulation study focusing on individually randomized trials with small sample sizes, we explore whether a range of adjustment methods are robust to misspecification, either in the covariate-outcome relationship or through an omitted covariate-treatment interaction. Specifically, we aim to identify potential settings where G-computation, inverse probability of treatment weighting (IPTW), augmented inverse probability of treatment weighting (AIPTW) and targeted maximum likelihood estimation (TMLE) offer improvement over the commonly used analysis of covariance (ANCOVA). Our simulations show that all adjustment methods are generally robust to model misspecification if adjusting for a few covariates, sample size is 100 or larger, and there are no covariate-treatment interactions. When there is a non-linear interaction of treatment with a skewed covariate and sample size is small, all adjustment methods can suffer from bias; however, methods that allow for interactions (such as G-computation with interaction and IPTW) show improved results compared to ANCOVA. When there are a high number of covariates to adjust for, ANCOVA retains good properties while other methods suffer from under- or over-coverage. An outstanding issue for G-computation, IPTW and AIPTW in small samples is that standard errors are underestimated; they should be used with caution without the availability of small-sample corrections, development of which is needed. These findings are relevant for covariate adjustment in interim analyses of larger trials.

A framework for handling missing accelerometer outcome data in trials.

Accelerometers and other wearable devices are increasingly being used in clinical trials to provide an objective measure of the impact of an intervention on physical activity. Missing data are ubiquitous in this setting, typically for one of two reasons: patients may not wear the device as per protocol, and/or the device may fail to collect data (e.g. flat battery, water damage). However, it is not always possible to distinguish whether the participant stopped wearing the device, or if the participant is wearing the device but staying still. Further, a lack of consensus in the literature on how to aggregate the data before analysis (hourly, daily, weekly) leads to a lack of consensus in how to define a "missing" outcome. Different trials have adopted different definitions (ranging from having insufficient step counts in a day, through to missing a certain number of days in a week). We propose an analysis framework that uses wear time to define missingness on the epoch and day level, and propose a multiple imputation approach, at the day level, which treats partially observed daily step counts as right censored. This flexible approach allows the inclusion of auxiliary variables, and is consistent with almost all the primary analysis models described in the literature, and readily allows sensitivity analysis (to the missing at random assumption) to be performed. Having presented our framework, we illustrate its application to the analysis of the 2019 MOVE-IT trial of motivational interviewing to increase exercise.