Metaanalyses, Network Metaanalyses, and Systematic Reviews: The Perpetual Motion Machine All Over Again
Bibliographic record
Abstract
The term metaanalysis was first used in the mid-1970s for describing methods designed to characterize and combine the findings of prior studies to increase statistical power, along with providing quantitative summary estimates, and to identify data gaps and biases1. (In this editorial I will use the term metaanalysis to encompass not only metaanalyses but also systematic reviews and network metaanalyses, because the issues I raise apply to all of them and their variations.) When applied to studies conducted with similar populations and methods, metaanalyses can be useful. However, this is not the case with many metaanalyses where the findings of studies that differ in important ways have been combined, prompting the comment that “they have mixed apples and oranges” — and sometimes “apples, lice, and killer whales — yielding meaningless conclusions”1,2. Combining the results of individual studies potentially increases the total number of participants, and this should mean increased statistical power, yet differences in participant demographics and study methods may actually lead to decreased power owing to variability in the patient characteristics1. This then leads to more difficulty in ascertaining the real effects. Add to this the issue of unpublished research to potentially skew the conclusions, because positive findings get published more often than negative results, starting with the decision to submit them in the first place3. It has been reported that falsified data also make it into metaanalyses4. In one example authors showed that 46% of all metaanalysis publications had their conclusions changed by publications with falsified data and 32% of all the analyses had a considerable change in the outcome5. There has also been … Address correspondence to Dr. Y. Yazici, 333 East 38th St., New York, New York 10016, USA. E-mail: yusuf.yazici{at}nyumc.org
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.189 | 0.435 |
| Meta-epidemiology (narrow) | 0.004 | 0.003 |
| Meta-epidemiology (broad) | 0.009 | 0.006 |
| Bibliometrics | 0.019 | 0.015 |
| Science and technology studies | 0.005 | 0.037 |
| Scholarly communication | 0.031 | 0.046 |
| Open science | 0.007 | 0.015 |
| Research integrity | 0.021 | 0.047 |
| Insufficient payload (model declined to judge) | 0.013 | 0.008 |
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 itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".