Research Letter
March 30, 2023
Comparison of Availability of Trial Results in ClinicalTrials.gov and PubMed by Data Source and Funder Type
Julianne T. Nelson, Tony Tse, Yvonne Puplampu-Dove, et al
JAMA. Published online March 30, 2023. doi:10.1001/jama.2023.2351
Timely dissemination of clinical trial results is essential for improving transparency and ensuring that findings are available to inform evidence-based medicine.1 Prior analyses of results dissemination assessed PubMed-indexed publications for National Institutes of Health (NIH)–funded trials completed by 20082; trial results published by academic and industry funders with results posted on ClinicalTrials.gov in 20123; and published or posted results of industry- and nonindustry-funded trials completed by 2015.4 In this study, the proportion and timing of results dissemination for registered trials were examined by data source (ie, ClinicalTrials.gov5and PubMed) and funder type.
Methods
On August 1, 2021, ClinicalTrials.gov was searched for registered trials with at least 1 US facility, initiated on or after January 1, 2015, and with primary completion dates by August 1, 2018, to allow a minimum of 36 months for results dissemination. We randomly sampled 400 trials, 100 from each key funder type: NIH, non-NIH US federal agency, industry, and other (eg, foundation). We reviewed ClinicalTrials.gov records for final primary results availability and date first posted on August 1, 2021.
To identify PubMed-indexed articles reporting primary outcomes and published by August 1, 2021, 4 authors (J.T.N., T.T., Y.P.D., and E.G.) searched PubMed for the sampled trials between August 6 and October 8, 2021, using methods reported previously,3 and the fifth author (D.A.Z.) resolved ambiguities. If electronic and print publications were listed, the earlier date was used. We determined proportions of trials with final primary outcome results, available in 1 or both data sources and first available in either, by funder type. We evaluated time from primary completion date to first results availability stratified by funder type using Kaplan-Meier analysis with log-rank testing (2-sided α = .05) in R (version 4.1.2; The R Foundation). Details about search, sampling, and screening strategies are available in the eAppendix in the Supplement.
Results
Of 9102 eligible trials, the majority involved drug, biologic, or genetic interventions (53.7%), an enrollment of 50 or fewer participants (55.5%), and multiple groups (73.2%). The distribution by key funder type was 714 (7.8%) by the NIH, 284 (3.1%) by a non-NIH federal agency, 4477 (49.2%) by industry, and 3627 (39.8%) by other. The median time for follow-up from primary completion date was 47.8 months (IQR, 41.9-56.8).
Results were identified for 245 of 400 trials (61.3%): 60 (24.5%) on ClinicalTrials.gov only, 128 (52.2%) on PubMed only, and 57 (23.3%) on both (Table). Of 245 trials with results, 98 (40.0%) were first available on ClinicalTrials.gov and 147 (60.0%) on PubMed. The median time from primary completion date to results availability was 18.8 months (IQR, 13.4-29.3) to posting on ClinicalTrials.gov (n = 117), 25.6 months (IQR, 18.5-36.0) to PubMed date of publication (n = 185), and 22.1 months (IQR, 14.9-32.9) for all 245 records.
Non-NIH federal agency–funded trials had the highest overall proportion with results available (71/100 trials), while industry-funded trials had the lowest (50/100) (Table). Other funder trials had the highest proportion in PubMed (61/100), while NIH- and non-NIH federal agency–funded trials had the highest proportions in ClinicalTrials.gov (both 42/100). Time to first results availability by funder type showed significant differences (log-rank testing: P = .003) (Figure).
Discussion
In this study, 39% of trials lacked results availability on ClinicalTrials.gov or PubMed after a minimum follow-up of 36 months following primary completion date. Nearly a quarter of all identified trial results were solely available on ClinicalTrials.gov, and 40% with available results were first available on ClinicalTrials.gov. Consistent with prior work, these findings suggest that searching both ClinicalTrials.gov and PubMed maximizes discovery of trial results.3,6Differences by key funder type likely reflect the influence of a patchwork of incentives for disclosing trial results.
Study limitations include the sampling period, which excluded longer-running trials and spanned the effective dates of 2 key federal reporting regulations: 42 CFR part 11 and the NIH Policy on the Dissemination of NIH-Funded Clinical Trial Information.5 Second, miscategorization of key funder type is possible. Third, data on differences in time to first results availability for trials available on both ClinicalTrials.gov and PubMed were not collected. Fourth, factors associated with nonreporting were not examined.
Future research on results disclosure behavior is needed to understand the gaps and develop ways to ensure more timely availability of trial results.
