Predictors of clinical trial data sharing: exploratory analysis of a cross-sectional survey

A number of major clinical research funders[1‐3] and biomedical journals[4‐7] have adopted policies supporting or mandating clinical trial data sharing in an effort to maximize the reproducibility and reliability of the results of clinical trials and maximize return on investment in research. Sharing data can increase its value in a number of ways, including by allowing external investigators to investigate secondary hypotheses, aiding future trial design, and providing training resources for students and fellows[8]. Clinical researchers most commonly share data either by depositing it in repositories, with terms of access defined by maintaining organizations, or by granting personal requests for the data on their own terms. Secondary users of trial data, such as the Cochrane Collaboration, are now advocating for stronger data sharing policies as a means to safeguard against selective reporting of outcomes and improve the medical evidence base for clinical decision-making[9‐17]. These growing calls for clinical trial data sharing have prompted the European Medicines Agency[18] and several companies in the pharmaceutical industry[19‐21] to commit to far-reaching data access policies for the first time. It now appears that many policymakers are no longer contemplating the need to share data, but rather the best way to do so[22]. Indeed, the United States Food and Drug Administration has recently proposed to share de-identified clinical trial data for the first time, albeit on a limited basis and such that product identity would not be disclosed, only therapeutic class[23].

Recently, we conducted a survey of clinical trialists to inform these ongoing data sharing efforts. Despite previous research finding clinical researchers to be amongst the least likely to share data[24], we found strong support for clinical trial data sharing, both in principle and in practice[25]. Most, but not all, respondents to our survey agreed that the clinical research community should promote and facilitate data sharing, and the majority of respondents who were required or requested to share data had done so. However, respondents identified a number of practical concerns with sharing trial data. Because trialists represent a diverse community of investigators with respect to funding sources, research aims and scope, and academic aspirations, they may vary in their concerns or support for data sharing. For instance, trialists receiving industry funding and investigators who do not publish frequently may both be less inclined to share data, albeit for different reasons. Trialists receiving industry funding may withhold data because of restrictions on their control over the data[26‐29] or potential conflicts with their research funders, whereas investigators who do not publish frequently may withhold data in an effort to optimize their own academic productivity. Similarly, trialists conducting either very large or small clinical studies may be less inclined to share data for different reasons, as trialists conducting bigger studies may be concerned with receiving adequate return on investment from large-scale undertakings in data collection, and trialists conducting smaller studies may be concerned with protecting patient confidentiality. On the other hand, trialists who published their clinical studies in journals advocating for data sharing and those located in Western Europe may be more inclined to share data in accordance with journal data sharing policies[5, 30], and a cultural shift towards data sharing in their scientific community[18], respectively. To date, no research has addressed this issue.

Accordingly, using the data collected during the previously described survey, we conducted a secondary analysis to explore whether concerns and support for data sharing varied among trialists according to professional background and aspects of the trials they conducted. In particular, we examined the following key characteristics, which we anticipated may influence trialists’ perceptions of data sharing: trialists’ academic productivity, geographic location, and trial characteristics including funding source, size, and the journal in which it was published. We expect these findings will help inform policy-makers by identifying groups of trialists who may be more resistant to new requirements and regulations on data sharing, or who may have specific concerns about data sharing that could be addressed, in order to facilitate data sharing.

There were 317 corresponding authors of clinical trials who responded to our survey. The majority of respondents were between 50 and 64-years-old (50%) and male (77%, Table 1). A minority of respondents published 10 articles or fewer over the past three years (22%), with roughly equal numbers publishing between 11 and 25 articles (37%), and more than 25 articles (41%) during the same period. More than half of our respondents were located in United States or Canada (53%), with over one third located in Western Europe (36%), and the remaining located elsewhere (11%). Nearly half of all respondents received industry funding for their study (48%), whereas a much smaller group was supported solely by non-profit funding sources (14%). With respect to trial size, a quarter of respondents were corresponding authors of trials with approximately 240 subjects or fewer, and another quarter with approximately 2,000 subjects or more. The majority of respondents published their trial in New England Journal of Medicine (36%), Lancet (22%), or Journal of the American Medical Association (14%). Approximately one fourth of respondents (28%) published their trial in one of the three journals advocating for data sharing, which included Annals of Internal Medicine (16%), The BMJ (8%), and PLoS Medicine (4%). Survey respondents did not differ from non-respondents with respect to corresponding author location or affiliation, trial enrollment, or journal in which the article was published. However, trial funders did differ between respondents and non-respondents; corresponding authors of trials funded solely by government sources responded more often than corresponding authors of trials funded solely by industry or by mixed funding sources (data not shown, provided in prior publication[25]).

In our survey assessing clinical trialists’ views on data sharing (the first study of this scope) we conducted exploratory analyses to determine whether certain characteristics of investigators and their trials were associated with more favorable perceptions of sharing. No prior study has examined how perceptions of data sharing may differ among individuals in this diverse community of clinical trial investigators. We found few notable differences in how respondents viewed the benefits and risks of data sharing. When respondents were categorized by trialists’ academic productivity and geographic location, trial funding source and size, and the journal in which it was published, respondents consistently indicated strong support for data sharing. Our findings suggest that, among authors of trials published in general medical journals with high impact factors, no particular group of trialists is more likely than others to be resistant to new requirements and regulations on data sharing.

However, while trialists were generally in agreement about the benefits of data sharing, we found several specific differences in their perceptions of data sharing that merit further discussion. With respect to academic productivity, respondents who published most often (>25 articles over the past three years) cited the protection of research subjects as a reason for withholding data less often. Investigators achieving such a high level of academic productivity are likely to have substantial experience handling data, and this finding may reflect their comfort and familiarity with implementing measures to ensure data de-identification. Future data sharing initiatives may be optimized for less experienced trialists by offering educational materials on best practice for data de-identification[31], and making ethical consultation available when there is no explicit consent for data sharing[43].

Respondents who conducted studies funded by industry cited the protection of research subjects as a reason for withholding data less often. These respondents were more often concerned with protecting investigator or funder interests when sharing data through a repository, a finding which lay just inside the region of statistical significance. These findings may in part reflect respondents’ lack of perceived ownership of data or access to data from industry-funded trials[26‐29]. Furthermore, these findings suggest that avoiding potential conflicts with corporate research sponsors is of primary importance to industry-funded trialists. Future data sharing policies may address this concern and secure engagement from trialists involved in industry-funded studies by ensuring robust data access, ownership and stewardship agreements and, where applicable, copyright or other licensing agreements[44].

There were also interesting patterns in the reasons respondents noted for sharing data. More than half of respondents located in Western Europe cited the potential for academic benefits or recognition as a reason for sharing data, whereas respondents located in the United States or Canada cited this reason less often. This finding suggests that the professional desirability of data sharing depends on cultural norms within scientific communities, and the community of trialists based in Western Europe has already begun to foster a culture in which investigators recognize that their academic aspirations may be furthered by sharing data. In the near future, data sharing policies which ensure co-authorship rights to data originators[8] may help incentivize more trialists to share data and align their professional interests with those of the public good. However, academic institutions and promotions committees will need to begin rewarding trialists for creating data and sharing it with other investigators if the global culture of academic medicine is to shift from ‘publish or perish’ to more open science[45, 46]. Some funding agencies, including the National Science Foundation, have already begun to consider data as a research product when evaluating investigators[47].

In our survey of clinical trialists’ perceptions and experiences with data sharing, we found few notable differences among subgroups of trialists in how they viewed the benefits and risks of data sharing. When respondents were categorized by trialists’ academic productivity and geographic location, trial funding source and size, and the journal in which it was published, strong support for data sharing remained. These findings suggest that no particular group of trialists is more likely to be resistant to new requirements and regulations on data sharing, as the clinical trials research community appears to have adopted a more favorable outlook on sharing. However, support for data sharing among trialists in principle may not translate to practice, as research funders and other stakeholders often figure prominently in the decision of whether or not to share. A better understanding of the context in which trialists conduct research may thus be helpful in identifying and addressing specific, practical barriers to sharing in order to align public and private interests and fully realize the promise of more open science.