A scoping review of the measurement and analysis of frailty in randomised controlled trials

Background Frailty is of increasing interest in trials, either as a target of intervention, as an outcome or as a potential treatment modifier. However, frailty measurement is often highly variable. This scoping review assessed how frailty is quantified in randomised controlled trials (RCTs), in what context and for what purposes. Methods We searched five electronic databases for RCTs in which frailty was measured among trial participants. We extracted data on intervention type, the frailty measure used and the purpose for which frailty was assessed. We then compared these data according to reasons for frailty assessment. Results We identified 415 RCTs assessing frailty across a range of interventions. Frailty was used to define the target population (166 trials), as an outcome (156 trials), as an effect modifier examining interaction of frailty on treatment effect (61 trials), as a purely descriptive characteristic (42 trials) or as a prognostic marker examining the impact of frailty on future health outcome (78 trials). The trials used 28 different measures of frailty (plus 29 additional trial-specific measures). The frailty phenotype model was the most common overall (164 trials), for defining the target population (90/166 trials) and as an outcome (81/156 trials). The cumulative deficit model frailty index was also frequently used (102 trials) and was most common among trials assessing treatment effect modification (21/61 trials). Conclusion Frailty measurement in RCTs is highly variable. Understanding the properties of respective frailty measures and how these relate to frailty as encountered in clinical practice is a priority to ensure that trial findings can inform healthcare delivery for people living with frailty.