Have Faster FDA Approvals Caused More Drug Safety Problems? No!

A version of this Health Alert appeared at Forbes.

The media gave some attention to a new study, which suggests that the Food and Drug Administration (FDA) recklessly allows unsafe new prescription drugs onto the market. The research supports a longstanding suspicion that the Prescription Drug User Fee Act (PDUFA), first passed in 1992 and renewed every five years, has caused the FDA to view the research-based pharmaceutical industry as a “partner” and source of revenue, rather than a regulated industry.

Before PDUFA, the FDA was funded by general appropriations. PDUFA has allowed the FDA to increase its revenue by user fees, which drug-makers agree to pay for new drug approvals or facilities inspections. PDUFA was last renewed in 2012. At the time, I endorsed the renewal because there was no likelihood of Congress reducing the FDA’s power: PDUFA was the best way to ensure drug approvals kept moving at the FDA.

These researchers think that’s a bad thing. According to the lead author, Cassie Frank, a physician at Harvard Medical School: “The FDA is under constant pressure to rush new drugs through the pipeline to approval. In its hurry, the FDA is apparently failing to distinguish useful drugs from toxic ones, and more dangerous drugs are slipping through. By the time many drugs receive serious safety warnings, millions of Americans have already been exposed to their side effects, which can sometimes be fatal.”

This conclusion is sensationalist, to put it mildly.

The authors examined drugs approved from 1975 through 2009 and found that drugs approved after PDUFA’s passage were more likely to receive a new black-box warning or be withdrawn than drugs approved before its passage (26.7 per 100 drugs versus 21.2 per 100 drugs at up to sixteen years of follow-up).

That 5.5 percentage point difference got a lot of attention and ginned up renewed accusations that the FDA is in cahoots with the drug industry. This is the result that the authors wanted. However, an objective consideration of their analysis leads to the opposite conclusion.

First, the authors mixed black-box warnings in with withdrawals. The reason for this is that so few drugs are withdrawn that there are no useful conclusions to be made by examining withdrawals alone. As the authors note:

This combined measure has a key advantage over the use of safety-related withdrawals alone, since withdrawals are rare (less than one per year) and hence analyses of this outcome alone have low statistical power.

However, black-box warnings are a different outcome than withdrawals. While withdrawal indicates (to some) that a drug should never have been allowed on the market, a black-box warning indicates exactly the opposite. It is the FDA’s most strict labeling requirement, telling prescribers that the drug has severe side effects. New information has been discovered that increases the likelihood of a physician appropriately prescribing the medicine. This new information would never have been learned if the drugs had not been on the market.

Second, reporting the statistical results as percentages, instead of the raw figures, disguises an important fact: There were actually more black-box warnings and withdrawals in the pre-PDUFA 16-year sample than the post-PDUFA 16-year sample: 80 versus 76. (I had to go through the appendix with a pencil to figure this out.)

In the pre-PDUFA 16-year sample, 42 drugs had black-box warnings. Because some drugs had more than one warning, the total number of warnings was 68. Twelve drugs were withdrawn. For the post-PDUFA sample, there were 48 drugs with warnings, 58 warnings, and 18 withdrawals. (Yes, there were more withdrawals in the post-PDUFA sample, but recall that there are so few withdrawals that they have no statistical power.)

So, patients in the post-PDUFA environment need not fear that more dangerous drugs are getting through the FDA. In fact, quite the opposite is true. While PDUFA improved the speed of FDA approvals after it was passed, normal bureaucratic inertia took over and the FDA returned to its previous, sluggish performance.

Comments (12)

Trackback URL | Comments RSS Feed

  1. Devon Herrick says:

    I encourage you to go to the article and take a look at the co-authors. I quickly recognized three of them as public health (and consumer) advocates who don’t trust the consumer should participate in their own medical decisions.

    There is a rich academic literature on the FDFA. The FDA is naturally risk-averse because a Type I Error (rejecting or stalling a good drug) causes less public scrutiny of the FDA than a Type II Error approving a bad drug. Patients who die of an adverse reaction to a bad drug are much easier to identify than people who die because they couldn’t get access to a good drug not yet approved.

    • Matthew says:

      I suppose the harm being done by not having the drug is better than the harm being done by a bad drug.

      • Devon Herrick says:

        From the viewpoint of the internal politics at the FDA, that is arguably the case. But from a public policy perspective, you want the number of people who die from lack of access to good drug to about equal the number of people who die from approval of a bad drug. As I’ve said in the past, the optimal number of people who are harmed by approved drugs and approved medical devices is a positive number. If you set the goal to have no one harmed by a bad drug or a bad medical device, the number of people helped by drugs and medical devices would be much, much smaller (that is, such a policy will cost more lives than it saves).

        • John R. Graham says:

          I think that’s what we call zero marginal returns. It would be a good principle to apply.

  2. Perry says:


    Here is another problem. A drug is approved and even promoted by the FDA and look what is happening.

    • John R. Graham says:

      I won’t comment on the specific lawsuit. However, new information on a drug that has been widely used for years is not surprising. We never know everything about a drug.

    • Devon Herrick says:

      When I read the headline I thought the same thing…

      “Women who are prescribed Lipitor to control cholesterol may share other risk factors that make them vulnerable to [diabetes], such as high blood pressure or obesity, the company said.”

      Lipitor has long been thought a Wonder Drug. There is little doubt that it provides significant benefits. That said, no cholesterol drug is as good for you as a healthy diet and exercise. If people try to use a drug to mitigate an unhealthy lifestyle and later suffer the consequences, is that the drug maker’s fault?

  3. Thomas says:

    No wonder they reached incorrect conclusions if they determined that black box warnings and withdrawals are the same thing. Black box warnings explicitly define adverse side effects, withdrawals show the drug shouldn’t have been on the market. I don’t see how they can confuse the two.

  4. Buddy says:

    “Second, reporting the statistical results as percentages, instead of the raw figures, disguises an important fact:”

    Well in this case, percentages help defend their claim. It is interesting how easily facts can be presented in such different ways.

  5. Bart I. says:

    Judging by the number of ambulance-chaser class actions being advertised on cable TV, I don’t think it matters what the FDA does.

  6. Greg Scandlen says:

    Conflating unrelated measures is a common tactic of the hysterics — XX million women have been raped, murdered or assaulted in the past X years!!!!! Obviously an “assault”: is wholly different than a rape or a murder but it is handy for beefing up the numbers.

  7. Dennis says:

    Talk about letting your agenda dictate your conclusions! A quick perusal of the appendix reveals many important medications in regular daily use to which a black box warning was attached. Some of these warnings took more than a decade of market availability for them to be added. Some were added after being identified in a similar drug. Others could have been predicted (spinal hematoma in enoxaparin for example). Still these medications are often important tools in patient care despite the warnings.
    How exactly would slowing down the FDA pipeline have identified all of these effects and what would be the impact on patients? As it is the PDUFA decreased the timeline to drug approval from 33 to 16 months. How many drugs were approved sooner as a result that proved helpful, or even life saving? If you had progressive hepatitis C how would you feel if the FDA delayed release of Sovaldi by more than a year?

    This paper should not have survived peer review.