Dataset: effect estimates obtained from randomised and non-randomised studies

This dataset provides study-level information on matched sets of published randomised and non-randomised studies answering the same clinical questions. The dataset includes 346 meta-analyses, and data on over 2,700 unique studies.

The dataset includes the following information (definitions and coding are provided in the spreadsheet):

IDs for clinical topics and source meta-analyses

IDs for individual studies

Brief descriptions of clinical topics

Therapeutic area (ATC Code first level)

Topic narrowness categorical

Indicators for high-quality source meta-analyses (Cochrane review; top journal)

Time period

Comparator type

Outcome type

Meta-analysis-level median risk of bias for NRS and RCT

Publication year of individual studies

Risk of Bias in individual studies

Standardised effect estimates (logOR) and variance for each study

Summary logOR and variance for NRS and RCT, respectively, within each meta-analysis for the full sample

The protocol for this study was registered on PROSPERO: 

Maximilian Salcher-Konrad. Comparison of treatment effects in randomised vs. non-randomised studies and the role of analytical methods to control for confounding: a meta-epidemiological study. PROSPERO 2018 CRD42018062204 Available from: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42018062204