Pre-Approval Drug Studies Adequate
When John Jenkins (FDA Director of Office of New Drugs) testified to the Institute of Medicine this past week that: “I don’t know that we’re ready at this point to say we are going to completely change the paradigm and start asking for databases 10 times bigger than they have been,” he wasn’t suggesting that current safety database sizes were “adequate”. What he was suggesting is that: “There could be a lot of really unintended consequences to what sounds like an easy, good idea.” In other words, he is justifiably concerned that a knee-jerk reaction to highly publicized safety recalls of new drugs is going to create unintended health hazards, chief of which is delayed access to important new medicines.
So, is this a justified concern, or is it kowtowing to industry, as industry opponents are sure to claim? Here are some facts to consider. Safety database size is determined by several factors that are described in detail in ICH Guidance E1A and FDA Guidance on Pre-Marketing Risk Assessment. The most important factor is the minimum exposure size needed to detect an adverse event “signal”. A signal is usually regarded as a single event. A simple rule of thumb is that a single adverse event occurring with a frequency of 3/X, where X is the drug-exposed population, will be evidenced with chronic dosing 95% of the time. So, for instance, the minimum recommended chronic dosing (i.e. 6 months) exposure of 300 subjects in ICH E1A will essentially rule out a safety signal occurring with a frequency of 1/100 (1%) if said signal is not evidenced during the studies. This is the minimum requirement. Most safety databases resulting in approval of the most widely used drugs are much larger. In the Vioxx case, the original NDA contained exposure to rofecoxib of 1396 subjects treated for at least six months at any dose. So, this size was sufficient to rule out a safety signal occurring with a frequency of about 0.2% with 95% confidence.
I think you’ll immediately see the limitation of this way of considering adequate exposure. When adverse events are rare, e.g. those occurring at a frequency of 0.2% would certainly qualify, a single case or no cases would be expected during the pre-marketing studies. A single case can easily (and usually is) dismissed as not drug-related, unless prior experience raises a red flag (e.g. signals of acute organ failure, certain allergic reactions). At the other end of the spectrum, common adverse events, such as myocardial infarctions in an elderly population, can easily (and usually are) dismissed as due to background disease, because often the safety database is not large enough to determine whether the adverse event occurs with a frequency greater than the expected background rate (or greater than the comparator rate).
In response, some experts have proposed large, controlled safety studies to provide adequate exposure to see rare signals and also to allow comparisons between therapies prior to marketing approval. The downside, as Dr. Jenkins points out, is that large studies designed to detect safety signals are very expensive and take very long to enroll. Yes, such studies are the gold standard, but are they the correct standard for most drugs? What is the appropriate degree of confidence in a new drug’s safety that should exist prior to marketing approval? The IOM is addressing such questions now. The general industry consensus is that current requirements are adequate. As for me, on most days I agree with the industry consensus. But then some days I arrive at the office and read that a newly approved drug causes a serious, perhaps fatal, adverse reaction. On those days, I wonder if we’re doing all we should prior to marketing approval.
