Where did I leave my systematic review protocol, and what should it contain regarding Trial Sequential Analysis?

TL;DR

This study examines how Trial Sequential Analysis is included in systematic review protocols and finds inconsistencies in reporting, suggesting protocols should be required to improve transparency and reduce bias.

Contribution

The study identifies gaps in the use and consistency of Trial Sequential Analysis parameters in systematic review protocols.

Findings

Half of the systematic reviews planned to use Trial Sequential Analysis in their protocols.

TSA parameters were inconsistently reported, especially for power and heterogeneity.

Predefined protocols improved transparency and reporting of key parameters.

Abstract

Publishing a protocol before starting a systematic review is necessary to ensure data integrity and trustworthiness of results. Trial Sequential Analysis (TSA) is a program and analysis method to control random type I and type II errors often used in systematic reviews (having a protocol in some form) and meta-analyses publications (having no protocol). We scrutinised protocols for systematic reviews and meta-analyses reports to examine their quality and transparency focusing on TSA parameters.

We searched Medline and the Cochrane Database of Systematic Reviews from 1 January 2018 to 31 December, 2021 for published systematic reviews and meta-analysis reports that included a TSA. We included reports that had at least two randomised clinical trials analysed in a forest plot plus in a TSA. Two independent researchers extracted data to assess coherence between the protocol and its systematic review. Disagreements were resolved through discussion or arbitration.

We included 270 systematic reviews with pre-published protocols and 274 meta-analyses without protocols. Among the systematic reviews, 135 (50%) planned to perform TSA in their protocols (112 dichotomous and 23 continuous outcomes). TSA parameters between protocols and published reviews were often inconsistent. For dichotomous outcomes, consistency was highest for alpha (50.9%) and lowest for power (1.8%). For continuous outcomes, consistency was minimal across all parameters. Statistically significant deviations between planned and actual TSA values were observed for statistical power (mean increase 8.95%; 95% CI 7.43% to 10.47%; p < 0.001) and heterogeneity (43.8% vs. 57.1%; p = 0.048) in dichotomous outcomes. Systematic reviews with predefined TSA parameters in protocols demonstrated significantly better reporting of the proportion of participants with the outcome in the control groups, alpha, and heterogeneity compared to those without protocol information. These patterns persisted when including the meta-analyses reports. Systematic reviews with continuous outcomes showed similar trends, though differences were not statistically significant.

TSA is often neglected in systematic reviews protocols and when planned the TSA parameters are inconsistently reported and followed. Predefined protocols improve transparency and reporting, and should be required to enhance reproducibility, reduce bias, and prevent research waste in systematic reviews.