Big Data and Pharmacovigilance, Part II

By Dov Fox

Yesterday, I wrote about Ryan Abbott‘s new article that proposes to use health information exchanges to improve the collection of data about the safety of approved drugs. Today, I want to address his recommendation that the government pay private parties to help the Food and Drug Administration conduct post-market drug surveillance.

The article argues that health information exchanges will make far more data available for observational research, but the current drug regulatory system is not structured in a way that would allow it to use the information effectively. This is because pharmaceutical companies have powerful incentives to emphasize data favorable to their products. (The FDA’s own research is dwarfed by comparison to what industry does and could do, while insurance companies and academics have weaker incentives to investigate.)

So Abbott suggests encouraging private parties to submit evidence that an approved drug is ineffective or unsafe. Under his proposal, if the FDA determines the evidence warrants a change to labeling or approval status, the submitting party would receive a financial prize. Its size would depend on how much the government saved by avoiding the costs of ineffective and unsafe medicines. Abbott suggests that the prize come from taxpayer dollars, unless it’s determined that the pharmaceutical acted negligently (or worse), in which case the company itself would be responsible for paying it.

It might make more sense to put the burden on pharmaceutical companies, even when they haven’t been negligent, given that it’s their products that are found to cause harm. Not only do they have the most opportunities to seek out problems early in testing and development; they are also the parties deriving the most financial benefit from medicine sales. Claims by pharmaceutical companies that they can’t operate with additional costs or regulation sound exaggerated, at least for many of those companies, given that the pharmaceutical industry has some of the highest profit margins of any business sector.

The bounty system nevertheless deserves serious consideration as an innovative proposal for harnessing market-based incentives to improve the regulatory surveillance of approved drugs. And it need not be restricted to health information exchanges. Abbott suggests applying this system to the way agencies deal with scientific questions generally. We might worry competitors might trump up data hostile to approved drugs. The provocative question that Abbott’s article invites us to ask is whether we are better off evaluating medicines under an inquisitorial system or an adversarial system.

Big Data and Pharmacovigilance, Part I

By Dov Fox

So much new data are created every day that 90 percent of all data worldwide has emerged in just the last two years. This information revolution has the potential, argues Bill of Health guest blogger Ryan Abbott, to transform how we develop new drugs, set clinical practices, and finance health care. His interesting new article paints an alluring “vision of a drug regulatory system powered by big data”:

“When the Food and Drug Administration (FDA) approved the cholesterol-lowering drug simvastatin in 1991, it was based on pre-marketing controlled clinical studies that included a total of 2,423 patients. In 2011 alone, just in the United States, almost a hundred million prescriptions were written for the drug. Imagine the impact of being able to analyze data from every one of those patients to evaluate whether simvastatin is safe and effective.”

The surveillance of pharmaceuticals after they’ve gone to market will matter more and more, Abbott argues, as personalized medicines become more difficult – and perhaps less necessary – to regulate before they’re released. He proposes a new plan for the post-market regulatory system that relies on the health information exchanges (HIE) created by the HITECH and Affordable Care Acts. These exchanges are slated to amass a vast repository of data on individual patients. Their large size and inclusive nature will enable more accurate analyses in observational research, Abbott suggests, and in ways that minimize the bias and selectivity problems associated with current data sets.

There are at least three obstacles to the integration of these exchanges in drug regulation. First, HIEs will be expensive. While the federal government provided considerable funding to get these exchanges off the ground, Abbott recognizes that in order to remain viable, they will probably have to sustain themselves financially. Second, their meaningful impact on post-marketing surveillance will require consistent reporting standards and information-sharing mechanisms. Third are important patient concerns about the privacy of their personal health information. States are experimenting with different patient participation models to address privacy concerns. For example, Abbott notes that in some states HIEs are free to exchange information without patient consent, while in others patients can opt-out of information exchange altogether. Either is permitted by HIPAA, so long as the information is de-identified so it can’t be used to identify individual patients.

Abbott argues that it’s worth tackling such concerns that the adoption of HIEs pose for citizens, policy makers, health care providers, and the health care industry, so we don’t squander the opportunity to use health information exchanges to their full benefit. Public support for data collection isn’t enough. That data must be translated into a format that regulators can use—something I’ll address tomorrow in my next post on the subject.