Dataset Information

Estimating publication bias in meta-analyses of peer-reviewed studies: A meta-meta-analysis across disciplines and journal tiers.

ABSTRACT: Selective publication and reporting in individual papers compromise the scientific record, but are meta-analyses as compromised as their constituent studies? We systematically sampled 63 meta-analyses (each comprising at least 40 studies) in PLoS One, top medical journals, top psychology journals, and Metalab, an online, open-data database of developmental psychology meta-analyses. We empirically estimated publication bias in each, including only the peer-reviewed studies. Across all meta-analyses, we estimated that "statistically significant" results in the expected direction were only 1.17 times more likely to be published than "nonsignificant" results or those in the unexpected direction (95% CI: [0.93, 1.47]), with a confidence interval substantially overlapping the null. Comparable estimates were 0.83 for meta-analyses in PLoS One, 1.02 for top medical journals, 1.54 for top psychology journals, and 4.70 for Metalab. The severity of publication bias did differ across individual meta-analyses; in a small minority (10%; 95% CI: [2%, 21%]), publication bias appeared to favor "significant" results in the expected direction by more than threefold. We estimated that for 89% of meta-analyses, the amount of publication bias that would be required to attenuate the point estimate to the null exceeded the amount of publication bias estimated to be actually present in the vast majority of meta-analyses from the relevant scientific discipline (exceeding the 95th percentile of publication bias). Study-level measures ("statistical significance" with a point estimate in the expected direction and point estimate size) did not indicate more publication bias in higher-tier versus lower-tier journals, nor in the earliest studies published on a topic versus later studies. Overall, we conclude that the mere act of performing a meta-analysis with a large number of studies (at least 40) and that includes non-headline results may largely mitigate publication bias in meta-analyses, suggesting optimism about the validity of meta-analytic results.

SUBMITTER: Mathur MB

PROVIDER: S-EPMC7954980 | biostudies-literature | 2021 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Estimating publication bias in meta-analyses of peer-reviewed studies: A meta-meta-analysis across disciplines and journal tiers.

Mathur Maya B MB VanderWeele Tyler J TJ

Research synthesis methods 20201114 2

Selective publication and reporting in individual papers compromise the scientific record, but are meta-analyses as compromised as their constituent studies? We systematically sampled 63 meta-analyses (each comprising at least 40 studies) in PLoS One, top medical journals, top psychology journals, and Metalab, an online, open-data database of developmental psychology meta-analyses. We empirically estimated publication bias in each, including only the peer-reviewed studies. Across all meta-analys ...[more]

PMID: 33108053

Similar Datasets

Project description:Publication bias is a substantial problem for the credibility of research in general and of meta-analyses in particular, as it yields overestimated effects and may suggest the existence of non-existing effects. Although there is consensus that publication bias exists, how strongly it affects different scientific literatures is currently less well-known. We examined evidence of publication bias in a large-scale data set of primary studies that were included in 83 meta-analyses published in Psychological Bulletin (representing meta-analyses from psychology) and 499 systematic reviews from the Cochrane Database of Systematic Reviews (CDSR; representing meta-analyses from medicine). Publication bias was assessed on all homogeneous subsets (3.8% of all subsets of meta-analyses published in Psychological Bulletin) of primary studies included in meta-analyses, because publication bias methods do not have good statistical properties if the true effect size is heterogeneous. Publication bias tests did not reveal evidence for bias in the homogeneous subsets. Overestimation was minimal but statistically significant, providing evidence of publication bias that appeared to be similar in both fields. However, a Monte-Carlo simulation study revealed that the creation of homogeneous subsets resulted in challenging conditions for publication bias methods since the number of effect sizes in a subset was rather small (median number of effect sizes equaled 6). Our findings are in line with, in its most extreme case, publication bias ranging from no bias until only 5% statistically nonsignificant effect sizes being published. These and other findings, in combination with the small percentages of statistically significant primary effect sizes (28.9% and 18.9% for subsets published in Psychological Bulletin and CDSR), led to the conclusion that evidence for publication bias in the studied homogeneous subsets is weak, but suggestive of mild publication bias in both psychology and medicine.

Project description:If a doctoral dissertation represents an original investigation that makes a contribution to one's field, then dissertation research could, and arguably should, be disseminated into the scientific literature. However, the extent and nature of dissertation publication remains largely unknown within psychology. The present study investigated the peer-reviewed publication outcomes of psychology dissertation research in the United States. Additionally, we examined publication lag, scientific impact, and variations across subfields. To investigate these questions, we first drew a stratified random cohort sample of 910 psychology Ph.D. dissertations from ProQuest Dissertations & Theses. Next, we conducted comprehensive literature searches for peer-reviewed journal articles derived from these dissertations published 0-7 years thereafter. Published dissertation articles were coded for their bibliographic details, citation rates, and journal impact metrics. Results showed that only one-quarter (25.6% [95% CI: 23.0, 28.4]) of dissertations were ultimately published in peer-reviewed journals, with significant variations across subfields (range: 10.1 to 59.4%). Rates of dissertation publication were lower in professional/applied subfields (e.g., clinical, counseling) compared to research/academic subfields (e.g., experimental, cognitive). When dissertations were published, however, they often appeared in influential journals (e.g., Thomson Reuters Impact Factor M = 2.84 [2.45, 3.23], 5-year Impact Factor M = 3.49 [3.07, 3.90]) and were cited numerous times (Web of Science citations per year M = 3.65 [2.88, 4.42]). Publication typically occurred within 2-3 years after the dissertation year. Overall, these results indicate that the large majority of Ph.D. dissertation research in psychology does not get disseminated into the peer-reviewed literature. The non-publication of dissertation research appears to be a systemic problem affecting both research and training in psychology. Efforts to improve the quality and "publishability" of doctoral dissertation research could benefit psychological science on multiple fronts.

Project description:BackgroundPreprints are increasingly used to disseminate research results, providing multiple sources of information for the same study. We assessed the consistency in effect estimates between preprint and subsequent journal article of COVID-19 randomized controlled trials.MethodsThe study utilized data from the COVID-NMA living systematic review of pharmacological treatments for COVID-19 (covid-nma.com) up to July 20, 2022. We identified randomized controlled trials (RCTs) evaluating pharmacological treatments vs. standard of care/placebo for patients with COVID-19 that were originally posted as preprints and subsequently published as journal articles. Trials that did not report the same analysis in both documents were excluded. Data were extracted independently by pairs of researchers with consensus to resolve disagreements. Effect estimates extracted from the first preprint were compared to effect estimates from the journal article.ResultsThe search identified 135 RCTs originally posted as a preprint and subsequently published as a journal article. We excluded 26 RCTs that did not meet the eligibility criteria, of which 13 RCTs reported an interim analysis in the preprint and a final analysis in the journal article. Overall, 109 preprint-article RCTs were included in the analysis. The median (interquartile range) delay between preprint and journal article was 121 (73-187) days, the median sample size was 150 (71-464) participants, 76% of RCTs had been prospectively registered, 60% received industry or mixed funding, 72% were multicentric trials. The overall risk of bias was rated as 'some concern' for 80% of RCTs. We found that 81 preprint-article pairs of RCTs were consistent for all outcomes reported. There were nine RCTs with at least one outcome with a discrepancy in the number of participants with outcome events or the number of participants analyzed, which yielded a minor change in the estimate of the effect. Furthermore, six RCTs had at least one outcome missing in the journal article and 14 RCTs had at least one outcome added in the journal article compared to the preprint. There was a change in the direction of effect in one RCT. No changes in statistical significance or conclusions were found.ConclusionsEffect estimates were generally consistent between COVID-19 preprints and subsequent journal articles. The main results and interpretation did not change in any trial. Nevertheless, some outcomes were added and deleted in some journal articles.

Dataset Information

Estimating publication bias in meta-analyses of peer-reviewed studies: A meta-meta-analysis across disciplines and journal tiers.

Publications

Estimating publication bias in meta-analyses of peer-reviewed studies: A meta-meta-analysis across disciplines and journal tiers.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets