Guest blog: Data sharing without accessible reporting is not enough

The latest Good Pharma Scorecard provides a welcome update on the disclosure of clinical trials. It rates the 20 largest pharmaceutical companies based on their disclosure of results for drugs that were approved by the US Food and Drug Administration (FDA) in a given year.

RESULTS REPORTING

For drugs approved in 2015, results were publicly available for a median of 100% of trials in patients by February 2017, showing that, for most drugs approved by the FDA, all trial results in patients are available at least a year after FDA approval.

Comparing with results from the previous year (2014), this represents a slight increase from an already high level (96%). When looking at means rather than medians, by my calculations, the overall disclosure rate came in at 87% of trials, compared with 90% the previous year.

In this field, terminology matters, and so it is worth unpacking what is meant by ‘publicly available’. To assess public availability, the creators of the Good Pharma Scorecard – Jennifer Miller at Bioethics International together with colleagues at Stanford and Yale – looked at whether trial results or clinical study report synopses were published in a journal or posted in a public registry (including corporate, national and international registries such as ClinicalTrials.gov).

Looking at these measures individually, they saw that the rates of publishing lag behind those of posting, perhaps because of delays in the peer review and editorial processes that are beyond the control of the company sponsors.

TRIAL REGISTRATION

The Good Pharma Scorecard also looks at whether trials have been registered. Registration aims to combat publication bias by enabling researchers to detect when trial results have been reported selectively or not at all.

Trial registration was consistently high in both years, with a median of 100% of trials registered. When looking at means rather than medians, the overall registration rate was 88% of 2015 trials, compared with 91% of 2014 trials (again, by my calculations). I hope that future analyses will be undertaken to confirm that registration took place before the trial began.

OPEN ACCESS PUBLISHING

Taken in context with other research from academia and the pharmaceutical industry, these findings show how well pharmaceutical companies are now doing in terms of disclosing the results of their trials. However, true transparency means making the results and interpretation of clinical trials available to everyone who needs them.

Researchers from Open Pharma analysed the publications included in the 2019 Good Pharma Scorecard to see whether or not they were available open access. The resulting research, which was presented at the 2020 European Meeting of the International Society for Medical Publications Professionals, showed that 61% of industry-sponsored publications in the 2015 data set were open access, up from 58% in the previous year.

To continue this progress, Open Pharma has issued a position statement calling on publishers to offer pharmaceutical companies the same right other funders have to publish their research open access. With over 150 endorsements from biopharmaceutical companies such as Galápagos and Ipsen, publishers such as PLOS and Wiley, and numerous patient advocates, academic researchers and policymakers, the position statement highlights the crucial role open access plays in ensuring transparent medical research. You can add your endorsement here.

DATA SHARING

Pharmaceutical companies have faced calls not just to report the results of all of their clinical trials but also to make the raw data available for checking by other researchers. As a result, they have invested tremendous resources in platforms such as ClinicalStudyDataRequest.com, Vivli and Yale Open Data Access.

The value of this investment is, in large part, pharmaceutical companies being able to demonstrate that they have nothing to hide. Yet, understanding of companies’ commitment to data sharing is currently limited, and the diversity of platforms and rules is unlikely to help.

A recent meeting of Open Pharma identified the need to specify and communicate the underlying data-sharing principles to which different pharmaceutical companies are signing up. Now, the latest Good Pharma Scorecard has done just that.

Starting with the landmark Institute of Medicine guidelines, Jennifer Miller and colleagues worked with diverse stakeholders to develop what are essentially the criteria for good data-sharing practice.

The outcome is that, for the first time this year, the Good Pharma Scorecard included an assessment of data sharing. To score 100%, pharmaceutical companies had to have a publicly available data-sharing policy that

(a) allows access to the analysis-ready dataset and associated clinical study report,

(b) explains how the data can be requested,

(c) reports the number and outcomes of data requests,

(d) states a deadline for sharing data of no longer than 18 months after trial completion or 6 months after regulatory approval (whichever is later).

To be compliant, companies also needed to have registered 100% of their trials on a registry.

Overall, the median score for data sharing was 80%, and four of the top 20 pharmaceutical companies scored 100%. Looking at the criteria individually, 18 of the top 20 companies have a data-sharing policy meeting criteria (a) and (b). However, only 10 reported the number and outcome of data requests (c) and only five set a deadline for data sharing that met the Good Pharma Scorecard view of best practice (d).

USE OF SHARED DATA

We can see from the Good Pharma Scorecard and other sources that pharmaceutical companies have data-sharing policies in place. Other studies have investigated how widely shared data are used.

Although most requests for data sharing are approved, the use of shared data has been limited to a small number of individual patient data meta-analyses and pooled analyses, with very few replication analyses undertaken to confirm the original findings.

Furthermore, of 177 approved analyses of shared data, only one led to a full publication. In light of the tremendous infrastructure required, in the words of one editorial, “Is the juice worth the squeeze?”

[TranspariMED comment: For a recent high profile example of data sharing, see the recently published meta-analysis of systemic corticosteroids for critically ill patients with Covid-19.]

THE ROAD AHEAD

It remains to be seen what impact industry’s investment in data sharing and open access will have on confidence in industry-funded research and, indeed, on human health.

Concerns about reidentification mean that managed access to individual patient data remains essential. In July 2019, researchers at Imperial College London demonstrated the ease with which almost any pseudonymized patient can been reidentified from just 15 characteristics.

However, perhaps my biggest concern is that the hot topic of data sharing may distract attention from trial reporting. Although the pharmaceutical industry is now performing well, academia is lagging well behind.

Unless trials are reported in a peer-reviewed publication or registry entry, no one will know that they have data available for sharing. It would be a tragedy if efforts to push forward the frontier of transparency diverted attention from the timely reporting of clinical trials.

Note: This guest blog was written by Chris Winchester, CEO of Oxford PharmaGenesis, a company that manages communications for pharmaceutical industry clients. The opinions expressed here are his own, and TranspariMED has not independently verified the data cited in his blog. TranspariMED does not receive funding from Oxford PharmaGenesis or any other commercial source.