Background
Clinical trial transparency is important to participants, sponsors, trialists, publishers, and regulators [
1‐
10]. Responsible sharing of participant-level data (also known as individual participant data, IPD) is one aspect of clinical trial transparency which enables novel secondary analyses, verification of results, and optimisation of future study designs [
1,
11,
12]. Effective 1 January 2014, the biopharmaceutical industry — via the Pharmaceutical Research and Manufacturers of America (PhRMA) and the European Federation of Pharmaceutical Industries and Associations (EFPIA) — endorsed a commitment to share de-identified IPD for approved medicines and indications upon request with qualified researchers [
3].
This endorsement was an important milestone in the pharmaceutical industry transitioning towards data sharing. It has sparked significant discussion and debate on several issues related to data sharing, including balancing the ease of access to shared data and the need to safeguard participant anonymity [
13‐
16], ensuring participants continue to enrol in clinical trials [
10], developing the required resources and support systems to facilitate valid research output [
15‐
18], the opportunity cost of data sharing [
19,
20], and mechanisms to provide adequate attribution to original investigators [
16,
20,
21]. However, little is known on the eligibility of independent researchers to request access to participant-level data. Evidence to date suggests that the scientific review process of submitted proposals to access clinical trial data is not a barrier [
19,
22], and thus a key step is whether a study is in scope. Whether a study is in scope for data sharing depends on whether the trial sponsor has a data sharing policy and, if so, the specific conditions of the policy. A recent audit of 42 selected industry sponsors found that 74% had a policy to share IPD, but it did not evaluate the conditions or the degree to which the policies affected published trials from being in scope to share [
23]. Herein the aim was to evaluate the proportion of prominent contemporary clinical trials sponsored by the pharmaceutical industry available for sharing of participant-level data with independent researchers 2 years after first publication of the primary results.
Discussion
To our knowledge, this is the largest structured assessment of the eligibility of independent researchers to request IPD from a broad selection of contemporary clinical trials that have been recently published. This study was also the first to identify data sharing policy decisions that have a negative impact on trial data sharing and to evaluate and identify the influence of PhRMA/EFPIA membership of industry sponsors on data sharing public policy and trial eligibility for data sharing.
With respect to study limitations, the eligibility of data sharing was investigated for clinical trials published 2–2.5 years ago in the top 10 general and internal medicine journals by impact factor. This time period was chosen as a reasonable balance between opportunity for trialists to publish key findings and independent data analysts to access data to contemporary trials which are affecting current health care — noting that 48 (79%) of the 61 clinical trials investigated a medicine and indication that was approved by either/both the FDA or/and EMA as of 1 October 2017. Eligibility may be different at other time points following publication, which is relevant as the Institute of Medicine committee recommend IPD sharing no later than 6 months after publication [
26], and the biopharmaceutical industry statement on principles for data sharing specifies no timeframe [
3]. Eligibility may also differ for clinical trials not published in the sampled journals or for the portion of clinical trials that go unpublished. Additionally, all 9 clinical trials that were identified as eligible for data sharing required a review board to accept a submitted research proposal prior to granting access to IPD, and this process was not assessed. Future studies should investigate the time from proposal submission to data sharing, data completeness upon sharing, researcher support initiatives, and data sharing eligibility of non-industry-sponsored trials. The importance of such investigations was recently reinforced by Naudet et al. [
27], who conducted a primary outcome reanalysis study in which only 46% (17 of 37) of trial authors provided complete IPD to randomised controls trials published in
The BMJ and
PLOS Medicine, journals with a strong data sharing policy (note that 26 of 37 sampled trials had no industry funding) [
28].
A recent audit of 42 selected industry sponsors reported that 74% had a policy for clinical trial data sharing [
23]. However, there was no assessment of how often clinical trials would be eligible for data sharing under reasonable circumstances (e.g. 2 years after publication). In comparison, 52% of our sampled industry sponsors had a public data sharing policy. The difference was largely driven by 46% of our sampled sponsors being a PhRMA or EFPIA member, compared with 81% in the prior study [
23]. Thus, while the PhRMA and EFPIA jointly committed to sharing de-identified IPD for approved medicines on 1 January 2014 [
3], this does not currently represent an industry-wide commitment, as the actioning of these principles for data sharing has been minimal by PhRMA and EFPIA non-members. Although PhRMA and EFPIA non-members are often smaller pharmaceutical companies, this study highlights the fact that they conduct and publish a significant portion (51% of trials in this sample) of contemporary clinical trials of medicines.
In 2015 Murugiah et al. assessed the data sharing eligibility of 60 clinical trials investigating cardiovascular medicines (> 5000 participants) sponsored by the top 20 revenue pharmaceutical companies [
29]. The completion date of the trials sampled in the Murugiah et al. study ranged from 1 to 14 years prior to their enquiry submission [
29]. Despite the current study sampling a wider range of medicines and sponsors and the focus on clinical trials published 2–2.5 years previous (avoiding historical data), our finding of 15% (9 of 61) of trials being eligible for data sharing was similar to the 15% (9 of 60) eligibility reported by Murugiah et al. This suggests that there has been minimal improvement in accessing industry-sponsored trial data in the 2.5 years since the Murugiah et al. study was conducted [
29].
The two most common reasons for confirmed data sharing ineligibility according to a sponsor’s policy were that the study was ongoing and regulatory reasons. Many modern trials continue to follow up participants well after the primary outcome of the trial has been completed and the results published. For instance, medicines for advanced cancers commonly evaluate progression-free survival as the primary outcome but continue to follow up for overall survival. Registration of a medicine is usually based on the primary outcome, and hence widespread use of a drug may occur well before study follow-up is completed. This is problematic for many data sharing policies that indicate the trial is eligible for data sharing after (often 1 or 2 years after) a study is completed. Of the 61 clinical trials evaluated, a substantial proportion, particularly for oncology trials, had not passed the final completion date at publication and 2 years after publication, with completion dates ranging up to December 2020. If a study has results worthy of being published or has contributed to the registration of a medicine, it is unclear why incomplete follow-up for a secondary outcome should prevent sharing of the data supporting the published primary outcomes.
Additionally, five trials were confirmed as ineligible for data sharing because either the medicine was not approved by both the FDA and EMA or there was an ongoing regulatory submission. Registration by both the FDA and the EMA is a condition currently stipulated in data sharing policies of many industry sponsors [
23,
26]. Arguably, once a medicine has been registered and is in widespread use in one jurisdiction, data sharing of the pivotal trials supporting the registration decision is in the public interest irrespective of the registration status in another jurisdiction.
Submission of a research proposal was required to assess eligibility of four clinical trials, and this was outside the scope of this study. Writing a research proposal is a major endeavour, and it is useful to be able to clarify whether a study is in scope for data sharing prior to developing a research proposal. As such, mandating a research proposal to assess data sharing eligibility for a specific clinical trial may discourage applications for data sharing.
Notably, of the 61 clinical trials to which eligibility for data sharing was enquired, no response/confirmation was received within 3 months for 22 trials, including 10 for which an explicit data sharing process was identified. This subgroup may partly represent trials in scope for data sharing; however, at 9 months after initial enquiry only 3 additional responses (2 trials eligible and 1 trial ineligible) were received, and thus the subgroup for the majority represents trials either ineligible for data sharing or trials that, due to severe process issues, were equated as ineligible for data sharing with independent researchers. As data sharing becomes more prevalent, this highlights the importance of considering and addressing the resource challenges faced by data custodians inherent in managing enquiries and proposals, preparing data for sharing, and appropriately managing data access [
6,
9]. Data sharing systems and processes are not inexpensive to establish, and to date the cost has largely been absorbed by the pharmaceutical industry [
19,
20].