We performed a cross-sectional analysis of clinical trials launched in response to the H1N1, Ebola, and Zika PHEIC events to assess outcome reporting patterns among trials performed during these pandemics.
We defined a clinical trial according to the definition used by the WHO: any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions to evaluate the effects on health outcomes . Clinical trials were eligible for inclusion if they addressed the prevention, diagnosis, or treatment of organisms responsible for the 2009 H1N1 influenza outbreak, the 2014 Ebola outbreak, or the 2016 Zika outbreak. In order to be eligible, trials must have initiated enrollment no earlier than one year prior to the beginning of the PHEIC declaration, and no later than one year after conclusion of the PHEIC, as defined by the World Health Organization. Because clinical trial registries were used to identify eligible trials, registration in ClinicalTrials.gov or one of the Primary Registries in the WHO Registry Network was also required . Trials with a registered completion date which was after February 2019 were excluded to allow time for data analysis and dissemination of results among the included sample. Trials were also excluded if the registry entry indicated that the trial was halted prior to enrolling any participants.
We identified eligible trials by searching ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP) for all trials, as well as the Pan African Clinical Trial Registry for Ebola, and the Brazilian Clinical Trials Registry, Cuban Public Registry of Clinical Trials and Peruvian Clinical Trials Registry for Zika. An investigator experienced in the use of trial registries (CWJ) searched each registry using key words and MeSH terms relevant to each PHEIC event (Supplementary Appendix A). After these searches, we removed duplicate trial entries and reviewed the full text of the remaining registry records to assess eligibility. For each included trial, we then downloaded a dataset from the relevant registry database containing key methodologic and logistical information, including the intervention being tested, funding source, and enrollment status.
Search for trial results and publications
For each included trial we determined whether results had been posted directly to the registry, and the date on which the results were posted. We also performed a comprehensive literature search to identify peer reviewed publications describing trial results. Some registries, including ClinicalTrials.gov, encourage investigators to update registry entries with a link to PubMed-indexed manuscripts containing trial results. Additionally, ClinicalTrials.gov uses each entry’s unique trial identification number (NCT number) to automatically search for and link to relevant PubMed entries. We reviewed each linked publication to determine whether it contained results from the relevant trial. For registry entries without a linked publication containing trial results, we conducted literature searches to identify relevant publications. The publication search strategy was created in consultation with a team of health-sciences research librarians, and involved searching MEDLINE via PubMed, Google Scholar, and EMBASE by trial registration number, keywords, trial title, and investigator name for manuscripts matching each included trial. Study investigators performed three independent searches before a trial was considered unpublished, including searches by the study’s Principal Investigator (CWJ) and a health-sciences research librarian for each trial. The final assessment of results within the registry and the final literature search occurred in September and October 2020.
Matching of registered trials and publications
We determined whether registry entries and publications identified by our search strategy matched by comparing the study title, trial design, interventions, number of participants, recruitment dates, study locations, investigators, and funding sources. A trial was considered published if we identified a peer-reviewed manuscript reporting un-pooled outcome data from the trial in question. Therefore, manuscripts that only described study methods without reporting trial results and those reporting results only as part of a pooled analysis were not considered to contain published results.
We utilized the standards for disseminating clinical trial results established by the WHO as the basis for the outcomes of the present study. These standards require both the posting of results to a trial registry within 12 months of study completion and publication of trial results in a peer reviewed journal within 24 months of study completion . The primary outcome for this study was the presence or absence of publicly available results meeting either one of these established standards. To maintain consistency with the WHO standards, we did not consider non-peer reviewed publications or conference abstracts to fulfill the publication requirement. Additionally, we considered each trial’s completion date to be the trial’s primary completion date, or the final date on which data were collected for the trial’s primary outcome measure.
Our secondary outcomes include full compliance with the WHO results dissemination standards, defined by meeting both the 12-month deadline to post results in a trial registry and the 24-month publication standard. We also report publication status in a peer reviewed journal at any time (regardless of whether publication met the WHO’s 24-month standard), and availability of results in a trial registry at any time (regardless of whether this met the 12-month standard). Finally, we report the elapsed time between trial completion and public availability of results in a peer reviewed publication and in a trial registry.
We grouped eligible registered trials into prespecified subgroups according to the disease being studied (H1N1 influenza, Ebola, Zika), intervention type (drug/biologic, vaccine, other), funding source (industry, federal government, university, other), and trial phase. For drug trials containing at least one intervention arm and one control arm, we reviewed published manuscripts and registered outcome data to classify the study results as positive, negative, or neutral. Superiority trials were considered positive if they reported a statistically significant primary outcome result favoring the intervention arm, based on the significance threshold chosen by the study investigators, negative if they found a significant primary outcome result favoring the control group, and neutral if there was no difference in the primary outcome. Finally, we performed a sensitivity analysis involving only those trials with a registered enrollment status indicating that enrollment had concluded and would not re-open.
Results describing categorical data are presented using descriptive statistics. We describe continuous data using medians and interquartile ranges (IQR). We used Kaplan–Meier methods to estimate the cumulative percentage of trials having publicly available outcome data over time, censoring unpublished trials on the date of the final unsuccessful manuscript search. We also report study outcomes according to our prespecified subgroups using descriptive statistics. We report the number of cases with missing data when relevant. These analyses were performed using SPSS version 27 (IBM Corp, Armonk NY).