Designing and analysing pragmatic clinical trials of complex interventions in mental health: challenges and solutions

Abstract

Designing and analysing clinical trials involving behavioural interventions is becoming increasingly complex. Trial teams are often presented with a variety of challenges requiring careful consideration, if the study is to be implemented successfully. SHIFT (Self-Harm Intervention, Family Therapy) is a pragmatic, randomised, controlled trial, comparing family therapy with treatment as usual for adolescents who have self-harmed at least twice, with a primary outcome of repetition of self-harm leading to hospital attendance. We will present challenges encountered during the design of SHIFT and steps taken to overcome them. Different therapists are involved in delivering interventions, hence we will present the rationale for choosing individual over cluster randomisation and how we minimise therapist variation using standardised training/supervision. As dropout from treatment and follow-up is common in trials of behavioural interventions, we will discuss the multiple methods employed to collect the primary outcome and how these maximise reliable data collection. Due to the nature of SHIFT's endpoints, the analysis will use techniques not often encountered in mental health trials. We will discuss the planned interim analysis; the use of Cox's Proportional Hazards Model to test for differences in repetition rates and the use of multilevel survival frailty models(1) to determine the extent of clustering on outcome due to therapists and the impact on the precision of the treatment effect. Multiple events analysis using Anderson-Gill(2) methodology making use of the timing and number of first and subsequent events will be discussed.