Research in the Age of Transparency

Don't look now, but data transparency is all the rage these days. This current movement toward transparency goes beyond recent scientific initiatives by organizations like the Critical Path Institute to make certain collections of data for a specific therapeutic area available for specific research purposes—it has extended into a teeming global movement to require any and all clinical research data relevant to marketed medical therapies be fully available to the full research and healthcare community. The idea is that doctors and scientists should have access to the full body of knowledge relevant to the use of a medicinal product—not just the information published in the product labeling.

In his controversial and provocative book Bad Pharma, Ben Goldacre calls for full disclosure of all clinical trial results data, so that physicians and patients can use that information to make the most informed treatment decisions. This is the same motivation that prompted Joe Kanter, when frustrated by the lack of information about treatment options for his own condition, to create the Joseph H. Kanter Family Foundation, which is working to establish a Learning Health System in the United States. Goldacre's book also includes other indictments on the industry, though one wishes he would balance these with the many benefits to patients that have come even under the flawed system he so bluntly criticizes.

And there are counterpoints to be made. For example, given the extraordinarily high costs of research and development, biopharmaceutical companies need to remain viable as successful business entities in a highly competitive environment, and thus may understandably be reluctant to expose the details of their developmental product portfolio to their direct competitors.

On the other hand, once a product is on the open market, the arguments against disclosure are harder to justify. And a few high profile cases of drugs withdrawn from the market due to serious safety events raise the question of whether lives are jeopardized when research data is withheld from independent researchers. Thus, some companies are now pledging greater data openness in the future, despite objections from elsewhere in the industry. GlaxoSmithKline in particular has stepped forward with a program committing to making patient-level from its clinical trials available to independent researchers, with others beginning to follow suit.

When it comes to improving patient health and potentially saving lives, any prior research, whether positive or negative, should add to the cumulative knowledge pool. This principle should also apply to data about products that are abandoned during clinical development without achieving approval—we have seen past examples where failed drugs later rose from the dead to provide useful treatments for other indications. There may be gold buried in those junk heaps of past failures.

And yet, there are many motivations other than fear of commercial competition that may make companies (or academics) reluctant or unable to share clinical research data.

For example, Goldacre mentions the historical impact of publication bias against studies that lack any groundbreaking revelations.

And some of the reluctance can be attributed to insecurity—or "Emperor's New Clothes" syndrome, when full transparency may be more revealing than modesty dictates. Some protocols fail due to shortcomings of design or errors in execution which nobody wants to broadcast—especially when the perceived contribution to research may be moot, or, worse, detrimental to otherwise positive effects demonstrated by more vigorously run trials. As Mom always warned, you want to make sure you're wearing clean, white underwear just in case you suddenly wake up in the ER at some point.

And then there's the fear of enabling misleading conclusions by the hopelessly uninformed. This is not unlike the traditional reluctance of some statisticians afraid to share data with clinicians for fear that they would jump to conclusions or, worse, respond with endless, naïve questions—better to have them wait for the printed table of results. Making data widely available to people with incomplete knowledge of the experimental design and uncertain skill—particularly those who may have axes to grind in today's Internet-leveled world—seems akin to giving the car keys and a credit card to a self-centered pre-teen.

Yet suppressing such research doesn't instill confidence, since this can also be an excuse to exaggerate a limited set of positive results.

One of the most important recent efforts, the European Medicines Agency's Clinical Trial Data Transparency Initiative, is setting an aggressive timeline to make clinical study reports (CSRs) more available. While CSRs can provide data for conducting meta-analyses, the agency also foresees requiring raw, patient-level data as part of their roadmap. The EMA has clearly been influenced by organizations such as the Cochrane Collaboration, who with Goldacre and others have established http://alltrials.net/—which proudly espouses the succinct goal of "all trials registered; all results reported."

But another rationale for withholding data is the fear of unintentionally violating patient privacy by disclosing protected health information—which could have serious legal and business consequences. Data can't be shared without the proper informed consent of the patient, and historical consent forms may not have anticipated the full transparency expected today. While there are now well-established techniques for anonymizing or pseudonymizing research data, these are not foolproof, and the risk/benefit tradeoff does not always point the same way for all data providers.

Despite legitimate fears of violating privacy, it's not likely that patients in need of a cure are obstructing this. Many are already sharing their data among Internet patient community groups, and others who are experiencing a serious disease would be willing to share data that might help others in the future—if only they knew how. Initiatives such as http://meforyou.org/ personalize the importance of safely contributing data to help the patients you love. John Wilbanks has established weconsent.us to allow anyone to donate data under a "consent to research" legal agreement that would allow your data to be shared under open access as long as you are not identified and not harmed. So patient privacy shouldn't be the problem.

Assuming we can actually have patient-level data, how would we use such data for research? Well, the answer depends on what we consider to be "research."

According to Martyn Shuttleworth, "the (broad) definition of research includes any gathering of data, information and facts for the advancement of knowledge." Students do research to prepare a history paper, and writers do research before composing a story.

Lots of useful research can be performed without using clinical trials data. The global movement toward expanded use of Electronic Healthcare Records (EHR) systems provides a rich source of vast amounts of patient medical data that can be harvested for epidemiology research and for drug safety initiatives like the Observational Medical Outcomes Partnership, the FDA's Mini-Sentinel, and the EU-ADR project in Europe. Other initiatives use EHR data to explore medical errors, quality and cost of patient care, comparative effectiveness of alternative treatments, and more. But epidemiologists are well aware of the limitations of EHR data, which is often sparse, incomplete, miscoded and subject to many confounding factors that are never directly recorded in the database.

The shortcomings of such observational healthcare data is often contrasted with what is considered the gold standard of research data gathered from protocol-driven randomized clinical trials, which use consenting human subjects to test the effectiveness and safety of a treatment, diagnostic, or prophylactic intervention such as a vaccine. In this world, research must systematically follow a rigid protocol in accordance with good clinical practice and government regulations to directly compare a treatment against a control. And the value of CRF data from clinical trials only increases when data standards are employed, so data can be effectively aggregated and compared to support discovery of new insights.

While there are cases—such as lab and past or concomitant drug use—when both healthcare and research can and ought to use the same data for multiple purposes, let's not confuse reusing the relatively small set of patient data collected during ordinary encounters with the extensive, high-quality data collected so rigorously under the precisely dictated conditions of randomized controlled clinical trial protocols. Most outcomes and efficacy data (often gathered on extensive research-specific questionnaires) typically do not even exist in EHRs at the source. But reusing the common data fit for both healthcare and research can only improve the quality and completeness of EHR data, which in turn provides better evidence for other research purposes overall.

This is not to minimize the potential value for EHRs to inform different types of research relevant to improving patient care. Goldacre provides an example of a UK observational study that is seeking to comparatively evaluate different statins that are prescribed to patients. Right now, the choice of product may be based on individual physician choice—but wouldn't it be better to continually collect more and more evidence on which products work best on which types of populations and start with those? This is precisely the type of thing that Kanter's Learning Health System ultimately intends to do: to have nearly every patient engage in a live trial that observes how they respond to each therapy—and pass that knowledge back into a knowledgebase so it can provide more evidence of what should work best for the next patient who comes along. Making all clinical trials data available—in standardized format—is also a key step toward that goal. So viva la transparence.

Wayne R. Kubick is Chief Technology Officer for the Clinical Data Interchange Standards Consortium (CDISC). He resides near Chicago, IL, and can be reached at [email protected].