Do The Right Thing: GSK's Case For Data Transparency
By Wayne Koberstein, Executive Editor
GlaxoSmithKline’s head of science and innovation argues that the future progress of R&D industrywide depends on open but qualified sharing of all clinical-trials data.
You must imagine this scenario because it has not yet come to pass: all patient data from all clinical trials, minus patient identities, made available through an independent “custodian” — an organization charged with administering access to the data based on criteria that ensure all granted requests have a “legitimate scientific purpose.” Now, turn back to status quo reality: highly selective or no release of clinical trials data at the total discretion of companies sponsoring the trials, except for disclosures forced by litigation or regulation.
Compared to such a reality, the imagined scenario amounts to data transparency. It may not be the naked transparency sought by Internet hackers and activists, where any person could obtain clinical data like readers of Gutenberg’s Bible received scripture. Quite frankly, the custodial model of transparency serves the industry’s interests in avoiding the chance that someone could wildly misinterpret such data to suit a nonscientific agenda, a distinct and even inevitable possibility. Yet the model, as expounded by Perry Nisen, senior vice president of science and innovation at GlaxoSmithKline, actually seems to give sponsors no real place to hide when science calls for a reexamination of the data generated in their human trials — an amazing change of climate for the industry should it become the new reality.
A basic rationale for the data transparency movement is that the data summaries published by sponsors do not always accurately reflect the underlying data. But if sponsors can distort the data in summary, why cannot their critics do the same? Therein lies a key argument for the independent custodian.
Nisen’s argument for a custodian does not rest solely on the question of summary error or bias, but also on the idea of refereed access to a massive data pool from which researchers can draw great power in their quest for safer and more-effective medicines. The custodian’s role thus becomes more a facilitator of research than a simple gatekeeper — ensuring shared data goes to qualified scientists on bona fide scientific quests, rather than amateur sleuths with axes to grind.
An article published in the Aug. 1, 2013 edition of NEJM gives a detailed account of GSK’s current and planned data-sharing program. Here, we are more concerned with the “why” than the “what” of the company’s data-transparency initiative. Why would a single company, acting on its own, go against the industry grain to push for transparency?
TRANSPARENCY, RIGHT VS. WRONG
Ben Franklin believed in “doing well by doing good” and suggested it is not enough to do good; you should be seen doing good. Similar practicality and honesty combine to explain GSK’s reasons for advocating and implementing the industry’s first companyinitiated system for voluntarily disclosing patient-level data from its clinical trials.
“I’ve always been struck by the duplication and inefficiency in clinical development — the inability to analyze in a meaningful way such a rich and deep source of data,” Nisen says. “We were getting ourselves organized to meta-analyze information, to validate methodology, to interrogate, and explore everything from placebo data to signals we would see in unexpected ways. But it was frustrating not to have access to the data in a straightforward way because of varying data standards, multiple databases, and so on.”
Although GSK has worked for years to establish standards that enable internal company access to patient-level data, Nisen says it still lacked an integrated sense of all the relevant data generated outside the company. Access to the universe of scientific information from outside trials was becoming essential, he says, particularly at the broader management level such as the company’s safety board, which he co-chairs.
On the other side of the equation, it became apparent that the often negative and poor-quality meta-analyses of GSK’s trials by people outside the company suffered from lack of data it controlled and traditionally held confidential. Custodial transparency appeared to be the only logical solution.
“We have an obligation to share our data,” asserts Nisen. “Even back in med school, I saw how powerful and useful a clinical trial with large data sets could be, as well as impossible to duplicate. To generate such magnificent amounts of data and not have a means for investigators and scientists to explore it in all kinds of ways seems so misguided. So here at GSK, one of the issues that especially matters to me is data transparency — making anonymized patient-level data available, ultimately in the interest of society — because it’s the right thing to do.”
A FEEDBACK OF BENEFITS
Three main reasons may justify the adoption of data transparency from the industry’s perspective: the potential for validation by multiple analyses of original observations or interpretations, the unleashing of data resources for researchers, and the advancement of evolutionary improvements in future clinical trials and data. Without such access to clinical trial data, each trial remains a closed book, locking up a wealth of irreplaceable information. “How many times can you reproduce a large clinical Phase 3 trial that involved 38,000 people? It’s not a doable thing,” says Nisen. “Companies are unlikely to repeat trials, especially the larger, late-stage clinical trials, just because of the enormity of the investment to generate that data in the first place.”
But if data transparency is the answer to how companies can have their trials and use them, too, it raises many other questions. Nisen says GSK’s goal is to publish all of its past trials, going back even before its merger — but how feasible is the goal? “The further we go back, the harder it is because of paper patient report forms and the ability to just track down all the bits of data and get them in shape to anonymize them,” he replies.
Similar limits apply to how far back in development data transparency may reach. “In the discovery space, we and all other sponsors have a way to go in making data available. Arguments could be made for sharing a lot of our validation work,” says Nisen. “With how much we all invest separately to validate and revalidate preclinical data, we could pool our results — and then let the winner get ahead with the best molecule, the best studies, and the best indications.”
On the clinical side, Nisen says companies could share a treasure trove of data about placebo effects, ranges of normal variation, validating methodology, and so on. “It is a shame we don’t make such data more available to investigators.”
The payoff for doing so, he says, could be nothing less than a leap in drug safety and efficacy. “Increasing benefit to risk, understanding disease, understanding the range of normal response — all that and more is possible once we unlock the data. We could apply signals, measures, and even methodology that could assess human response using biomarkers, in vivo or in vitro models, or other predictors of toxicology and efficacy.”
Companies could do a lot together to reach the goal, and some efforts are under way to set up “cloud sharing” and other cooperative programs for exchanging commonly useful data. But Nisen makes the point that the issue of transparency is not restricted to pharmacompany sponsors. “There are lots of studies undertaken in academia where we should be able to see, cross-analyze, and interrogate the data.” An independent custodian would maintain a platform for the broadest possible inclusion of all sponsors’ data, he believes.
“We hope to move to a situation where an independent custodian would have oversight of which academia and sponsors could make their data available, preserve anonymity of patients, ensure a reasonably legitimate scientific question is being asked, and verify the research teams are equipped with the necessary IT and support to handle and analyze the data. Without those protections, one of the risks is nonqualified people using our data to make nonscientifically valid assertions about benefit and risk,” he says.
GSK BLAZES THE TRANSPARENCY TRAIL
According to Nisen, the company began it efforts toward data transparency prior to his arrival in 2001. Frank Rockhold, now senior vice president, drug development sciences, and others inside the company started to work internally and externally on establishing a common database and standards and dealing with issues such as patient privacy and informed consent. Actual sharing of data began in the same time frame with GSK trials of medicines for developing-world diseases such as TB and malaria.
Finally, last May, the company made anonymized patient-level data available from more than 200 studies “within certain boundaries” on the GlaxoSmithKline Clinical Study Requests website (https://clinicalstudydata.gsk.com), and Nisen says it will continue to expand its program, doubling the number of studies available by the end of the year. GSK has also committed to publishing clinical study reports (CSRs) of its marketed and terminated/ failed medicines and the detailed summaries and interpretation of results from its clinical trials. Beginning in December, more than 1,500 reports will be made available over a two-year period. “It started in Europe, and by December we’ll be putting out thousands of the clinical summary reports.”
Meanwhile, the company has been working with SAS to build the analytical system where this patient-level data can be made available on a central hub database, along with a deployed model where outside investigators can have access to data, once their request has been approved by an independent panel.
GSK’s first step toward implementing the broader custodial idea was to establish its own expert review committee with oversight and authority over data releases. Anyone receiving the data must sign a data-sharing agreement that describes the use they will make of it.
But to establish an independent custodian would require an industrywide effort. GSK’s vision of the custodian is an organization modeled on the Structural Genomics Consortium, formed initially by reaching out to all the key constituents and structured to represent all of their interests. “Once we work out some of the issues for ourselves, hopefully a few other sponsors will join in, and we will make some adjustments to accommodate their expectations. And presumably, that will help GSK create a movement to start making the consortium happen,” Nisen says.
GSK may have some competition in influencing the form data transparency eventually takes. The EMA (European Medicines Agency) is preparing to advance an entirely different approach, in which the agency would be the central arbiter of data release and publication for trials in its jurisdiction. And activist proponents of total transparency would be loathe to surrender the initiative to any company they consider responsible for secreting the information in the first place.
Academic institutions may find it particularly difficult merely to participate in a data transparency system, partly because they typically lack the institutional history of open research, but perhaps more importantly because of the costs involved. Nisen says the resources needed to put clinical data into an analytical form according to a common data standard are expensive and perhaps impractical in such institutions, considering the high turnover among their investigators.
In the face of those and other issues, Nisen is philosophical, but determined to push ahead. “One of the only ways to get around antagonisms and distrust is to just start making our data available and let others then affirm or refute the conclusions that we’ve made. At the end of the day, there will be a benefit, predicated upon data generated to the best of our ability, along with the best analysis possible. And the more we can do that, the more acceptance we will win. We will also have less waste in generating and regenerating data.”
Nisen says the company is now in related discussions with a few other pharma sponsors about joining its transparency initiative. “I am cautiously optimistic that we’ll move forward and they, too, will start contributing.” He says GSK has also spoken with “independent parties who could potentially function as independent custodians.”
OPEN DATA, INNOVATION, and TRUST
Well done is better than well said — according to another Franklin maxim. Inside GSK, Nisen and his team turn words into action, coordinating a flexible group of personnel largely drawn as needed from the company matrix. “We receive team support, policy support, and operational support from the company,” he says. “We can put enough resources behind any particular project to make it work. And we have a commitment to data transparency as a high priority from the top of the organization down.”
One of the ongoing responsibilities of Nisen’s group is to use data sharing as a tool for clinical trials improvement. To start with, it teaches investigators in new trials how to prepare the clinical data they help generate for future sharing, including publications. But data transparency also molds the operations, regulatory routines, policy development, and overall planning of trials — essentially forcing an innovative approach to the clinical development.
Outside GSK, however, the people running small development companies may well wonder, Why should I add this to my plate? Why should I make my data transparent? “For all the same reasons that we do — to leverage the opportunity to cross-analyze and harmonize the information,” answers Nisen. “If you move to common standards, you have to give a little to get more back, on some level. But it will not be so easy for small companies to do that unless, from the inception, they adhere to a common standard and leverage the information to model and simulate what they want to do. It might make for much better studies on their part as well, I would think.”
It would also not hurt the partnering prospects of small companies to be a part of an industrywide collaboration or consortium in data transparency. Doing good, being seen doing good, plus doing what you say should be done — not a bad equation for entrepreneurs out to change the world. If nothing else, joining the data transparency movement will help remove the stifling insulation that all too typically surrounds young companies.
Nisen cites the company’s chief executive to summarize why GSK has chosen transparency as the right thing to do. “One of the fundamental pillars that Andrew Witty has articulated from the very beginning was building trust. And one of the key ways to build trust, I would say, is to be transparent, to make our data available — just walking the talk.” The rationale of trust seems sound for any clinical trials sponsor, no matter how large or small.