Poor reporting quality and high proportion of missing data in economic evaluations alongside pragmatic trials: a cross-sectional survey.

Abstract

Background: Lack of data integrity is a common problem in randomized clinical trials and is more serious in economic evaluations conducted alongside explanatory clinical trials. Despite pragmatic randomized controlled trials (pRCTs) becoming recognized as the best design for economic evaluations, information on the proportion, handling approaches, and reporting quality of missing data in pRCTs-based economic evaluations remains limited. This study aimed to investigate the quantity and reporting quality of missing data in economic evaluations conducted alongside pragmatic clinical trials.

Methods: In this cross-sectional survey, data were extracted from PubMed and OVID (Embase, CENTRAL, HTA database, and NHS EED) from January 1, 2010, to April 24, 2022. Economic evaluations conducted alongside pRCTs were included. Two independent reviewer groups identified relevant articles, and data were extracted by three groups comprising two reviewers each. Descriptive analyses were performed to assess the characteristics of the included studies, missingness in the included studies, and handling of missing data.

Results: Overall, 715 studies were identified, of which 152 met the inclusion criteria. In total, 113, 119, and 132 articles reported missing data, costs, and effects, respectively. More than 50% (58/113) of the articles reported the proportion or quantity of overall missingness, and 64.71% and 54.55% reported missing costs and effects, respectively. The proportion of missingness of < 5% in the overall group was 3.45%, whereas the proportions of missing costs and effects were both < 10% (5.26% vs. 8.45%, respectively). In terms of the proportion of missing data, the overall missingness rate was 30.22% in 58 studies, whereas the median proportion of missing data was slightly higher than that of missing effects (30.92% vs. 27.78%). Of the included studies, 56 (36.84%) conducted a sensitivity analysis on handling missing data. Of these, 12.50% reported missing mechanisms, and 83.93% examined handling methods.

Conclusions: Insufficient description and reporting of missing data, along with a high proportion of missing data in pRCT-based economic evaluations, could decrease the reliability and extrapolation of conclusions, leading to misleading decision-making.