Content area
Full text
Background
Trial attrition, or the non-completion of a trial for any reason, presents several risks for the validity of randomised controlled trials (RCTs). Attrition to varying degrees is an inevitable occurrence in RCTs. Publicly funded RCTs typically lose up to 12% of participants to attrition [1, 2], but rates as high as 70% have been reported [2].Considerably higher rates have been reported among RCTs for cancer [3, 4–5], obesity [6] and psychological conditions [7, 8–9]. The impact of attrition on trial inferences is complex and depends on the type of outcome (e.g. events versus measures), effect measures and analysis strategy (e.g. intention-to-treat (ITT) versus per protocol) [10]. Nonetheless, attrition can introduce bias if the characteristics of retained participants differ from those lost to attrition [11]. Moreover, if related to an intervention, attrition can undermine randomisation and the exchangeability of treatment groups, threatening the core strength of RCT designs [12]. Furthermore, it can impact statistical power [13], treatment effect estimates [14] and the generalisability of findings [15]. Given its commonness and potential impacts on RCTs, addressing attrition is a top priority in the trial methodology research agenda [16].
To better understand why attrition occurs in RCTs, previous studies have explored influential factors. Quantitative studies have identified participant and trial characteristics associated with attrition, including participant sex [6, 17, 18–19] and comorbidities [17, 20], as well as trial duration [4, 5] and recruitment strategies [19]. Qualitative studies have identified participant and trialist-perceived barriers to retention in RCTs. From the participants’ perspective, these relate to personal beliefs, capabilities and life circumstances [21], while trialists perceive the presence of severe comorbidities, adverse events and study procedures to facilitate attrition [22, 23]. Evidence for influential factors has rarely been synthesised across RCTs for multiple conditions, particularly for participant and trial characteristics. Given heterogeneous trial designs, interventions and populations, the types of characteristics associated with attrition might differ depending on the condition studied. Conversely, there could be frequently associated characteristics across multiple conditions. To explore this, it would be useful to know which characteristics are reported in analyses of attrition among trials of individual conditions as well as across multiple conditions. This would inform future studies intending to evaluate characteristics that are predictive of trial attrition and, in turn, trialist decision...