Reproducibility of meta-analytic results in systematic reviews of interventions: meta-research study
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Objective: To determine how often meta-analyses of effects of interventions are reproducible. Design: Meta-research study. Setting: Systematic reviews with meta-analyses of the effects of health, social, behavioural, or educational interventions indexed in five databases (PubMed, Science Citation Index, Social Sciences Citation Index, Scopus, and Education Collection), 2 November to 2 December 2020. Population: 296 reviews meeting the inclusion criteria formed the overall sample of the study. 175/296 (59%) reviews included a forest plot from Review Manager and were considered inherently reproducible. The remaining 121/296 (41%) reviews constituted the reproduction sample. Main outcome measures: Original review authors were contacted to obtain meta-analysis data files, and analytic code used to generate the first reported (index) meta-analysis; if not provided, the necessary data and statistical details of the meta-analysis methods were extracted from the review. Two investigators independently reproduced each review's first reported meta-analysis using the original computational steps and analytic code. Meta-analyses were classified as fully reproducible if the difference between the original and reproduced summary estimates and 95% confidence interval (CI) widths was less than 10%. Differences in meta-analysis results were classified as meaningful if there was a change in direction of the summary effect estimate or if the 95% CI included the null, which may alter the interpretation of the results. Results: 22 authors provided data files or analytic code, or both. 104 meta-analyses (86%) were fully reproducible, seven (6%) were not fully reproducible, and 10 (8%) had insufficient data available to attempt reproduction. No meaningful differences were found in the reproduced meta-analytic results that might alter their interpretation (eg, changes in the direction of summary effect estimate or if the 95% CI included the null). Conclusions: The findings of the study suggested that the results of meta-analyses could be reliably replicated if the original data or analytic code, or both, could be obtained, or if the necessary data were accessible in the review. Few systematic reviewers responded to requests to share data or code. Making data files and analytic code publicly available will facilitate future investigations of reproducibility.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.887 | 0.850 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.027 | 0.008 |
| Bibliometrics | 0.003 | 0.010 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.003 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.008 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it