Are the MORECare guidelines on reporting of attrition in palliative care research populations appropriate? A systematic review and meta-analysis of randomised controlled trials.

Background: Palliative care trials have higher rates of attrition. The MORECare guidance recommends applying classifications of attrition to report attrition to help interpret trial results. The guidance separates attrition into three categories: attrition due to death, illness or at random. The aim of our study is to apply the MORECare classifications on reported attrition rates in trials.

Methods: A systematic review was conducted and attrition classifications retrospectively applied. Four databases, EMBASE; Medline, CINHAL and PsychINFO, were searched for randomised controlled trials of palliative care populations from 01.01.2010 to 08.10.2016. This systematic review is part of a larger review looking at recruitment to randomised controlled trials in palliative care, from January 1990 to early October 2016. We ran random-effect models with and without moderators and descriptive statistics to calculate rates of missing data.

Results: One hundred nineteen trials showed a total attrition of 29% (95% CI 28 to 30%). We applied the MORECare classifications of attrition to the 91 papers that contained sufficient information. The main reason for attrition was attrition due to death with a weighted mean of 31.6% (SD 27.4) of attrition cases. Attrition due to illness was cited as the reason for 17.6% (SD 24.5) of participants. In 50.8% (SD 26.5) of cases, the attrition was at random. We did not observe significant differences in missing data between total attrition in non-cancer patients (26%; 95% CI 18-34%) and cancer patients (24%; 95% CI 20-29%). There was significantly more missing data in outpatients (29%; 95% CI 22-36%) than inpatients (16%; 95% CI 10-23%). We noted increased attrition in trials with longer durations.

Conclusion: Reporting the cause of attrition is useful in helping to understand trial results. Prospective reporting using the MORECare classifications should improve our understanding of future trials.