Use of wearable biometric monitoring devices to measure outcomes in randomized clinical trials: a methodological systematic review

Abstract

Abstract Background Wearable biometric monitoring devices (BMDs) have the potential to transform the conduct of randomized controlled trials (RCTs) by shifting the collection of outcome data from single measurements at predefined time points to dense continuous measurements. Methods Methodological systematic review to understand how recent RCTs used BMDs to measure outcomes and to describe the reporting of these RCTs. Electronic search was performed in the Cochrane Central Register of Controlled Trials, PubMed, and EMBASE and completed a page-by-page hand search in five leading medical journals between January 1, 2018, and December 31, 2018. Three reviewers independently extracted all primary and secondary outcomes collected using BMDs, and assessed (1) the definitions used to summarize BMD outcome data; (2) whether the validity, reliability, and responsiveness of sensors was reported; (3) the discrepancy with outcomes prespecified in public clinical trial registries; and (4) the methods used to manage missing and incomplete BMD outcome data. Results Of the 4562 records screened, 75 RCTs were eligible. Among them, 24% tested a pharmacological intervention and 57% used an inertial measurement sensor to measure physical activity. Included trials involved 464 outcomes (average of 6 [SD = 8] outcomes per trial). In total, 35 trials used a BMD to measure a primary outcome. Several issues affected the value and transparency of trials using BMDs to measure outcomes. First, the definition of outcomes used in the trials was highly heterogeneous (e.g., 21 diabetes trials had 266 outcomes and 153 had different unique definitions to measure diabetes control), which limited the combination and comparison of results. Second, information on the validity, reliability, and responsiveness of sensors used was lacking in 74% of trials. Third, half (53%) of the outcomes measured with BMDs had not been prespecified, with a high risk of outcome reporting bias. Finally, reporting on the management of incomplete outcome data (e.g., due to suboptimal compliance with the BMD) was absent in 68% of RCTs. Conclusions Use of BMDs to measure outcomes is becoming the norm rather than the exception in many fields. Yet, trialists need to account for several methodological issues when specifying and conducting RCTs using these novel tools.