In the FDA-regulated world, success is often defined as approval of a new product or indication based on two, well-controlled clinical trials. However, the scrutiny doesn’t end there. FDA’s mission includes determining whether already-approved drugs perform safely and effectively when used by large numbers of patients in routine medical practice.
To understand what happens under these “real world conditions,” FDA has expanded its post-market efforts, including development of a monitoring system (called Sentinel) that will be able to track drug usage and medical history information on tens of millions of patients. Although such information will be useful, it can only provide post-hoc inferences, not proof of causation. Even with this limitation, FDA Matters thinks developing the system is worthwhile and may provide multiple benefits.
There are multiple tools for assessing post-approval safety and efficacy that fit loosely under the rubric of pharmacovigilance. When approving medical products, FDA mostly relies on data that comes from pre-specified hypotheses that are tested through randomized, placebo-controlled, double-blind clinical trials. In contrast, the data that comes from pharmacovigilance is inherently less rigorous; indeed it constitutes a form of “data dredging” that FDA abhors. The heart of the problem is that:
Real world data sets = uncontrolled variables + inconsistent data collection + questionable data accuracy.
When FDA and manufacturers collect adverse events reports, they know there will be underreporting of incidents, as well as limited ability to judge whether problems are drug-related. When FDA looks at the Medicare database, they know that information submitted as part of medical claims is unreliable and subject to systemic bias (e.g. medical coding is designed to support reimbursement, not public health analysis).
The Sentinel database should be superior because it incorporates medical records and patient registry information, along with claims data. Still it provides inferences, not proof.
Active surveillance—continuously monitoring millions of health records—is only worthwhile if these limitations are acknowledged. It can never provide certainty about whether drugs are safe and effective. It can tell you what is worth further examination…but can never tell you the cause of any problem that is identified.
As the FDA mantra goes: association is not causation. No matter how many health records and claims data are reviewed, this is still true.
Clinical trials have limitations, also. Trials don’t tell us how a drug will be used by prescribers. They can never provide complete information about patient outcomes for those individuals with several medical conditions (i.e. multi-morbidity) or who take many medications simultaneously (i.e. poly-pharmacy).
By inference (although not with certainty), pharmacovigilance and active surveillance could bring us closer to addressing potential problems that can’t be resolved by clinical trials. For example, many years ago, I worked on a drug to treat pre-term labor. As I recollect, there were two instances of respiratory problems in a trial of several hundred women. No one could say for sure whether this effect was caused by the drug or occurred at random. A study large enough to find out was infeasible.
Based on the potential respiratory problem, FDA rejected the drug despite the benefits it might have provided to women experiencing pre-term labor. If this same situation were to come up today…maybe FDA would decide differently, knowing it could collect patient outcomes information through pharmacovigilance, particularly active surveillance.
Ideally, FDA would know everything it needs to know about a drug at the time of its approval. Information derived from review of real world data sets can never be as good. But properly understood and carefully analyzed, the inferences derived from pharmacovigilance can add to our understanding about safety, efficacy, drug interactions and side effects.
Instead of just using that capacity to identify post-approval problems, FDA needs to incorporate pharmacovigilance into its thinking about when to approve drugs and with what conditions. FDA’s capacity to do pharmacovigilance and active surveillance should lead to a greater willingness by FDA to approve drugs, particularly those with otherwise solid benefit-risk equations, but burdened by questions that cannot be resolved prospectively or through clinical trials (even in phase 4).
Patients would benefit if FDA made this one of the Sentinel priorities.