Robert R. Fenichel



Email of 29 December 2000 to David Willman

   (As originally sent, this message include hyperlinks to two AP stories by Lauran Neergaard.  Those links have since broken, so I have added locator information of untested adequacy in blue.)

    I'd like to talk to you some time about the way your series turned out.

    I think I've read it all, but I'm not sure, since in the cluster of stories that appeared on 20 December only one had your byline.  The others I read were on cisapride [Propulsid®], on grepafloxicin [Raxar®], on fenfluramine [Redux®], on Duract®, on mibefradil [Posicor®], on alosetron [Lotronex®], and on troglitazone [Rezulin®].

    I now regret that when you and I spoke, we jumped right into talking about the cases.  That is, we didn't spend nearly enough time talking about the principles underlying contemporary pharmacotherapy, drug regulation, and drug development.  Without this basis, I don't know how to make sense of the story:  Decisions made by reasonable people sometimes become inscrutable, and benign factoids become portentous.

   Perhaps the place to begin is with your lead, where you allude to the ancient physician's creed of "First, do no harm."  This millennia-old saying has changed (or lost) its meaning in the past hundred years. Before that, physicians had cathartics, emetics, leeches, hot baths, cold baths, and the like. They had no tools that favorably altered the natural history of human disease.  The most likely consequence of treatment was harm, but doctors (like any other tool-holders, then as now) were always looking for nails to fit their hammers.  Primum non nocere was the old equivalent of "Don't just do something; stand there."

  Things are different now. Properly applied, modern treatments can have beneficial effects. But because they are not just inactive nostrums any more, they can — and statistically will do some harm too. The only way to do no harm is not to treat patients.

  • Some therapies (surgery for acute appendicitis, antibiotic treatment for bacterial meningitis, etc.) are intended to defer or avert irreversible harm (death, strokes, etc.). We accept that these therapies cause harm and even death (for example, about 1 in 50000 patients die from general anesthesia, and about 10-15% of people who undergo appendectomy turn out to have had not appendicitis, but rather some harmless condition that would have got better on its own, so all they get from the procedure is harm). These therapies are studied in clinical trials to show that the average effect is beneficial. Once physicians are convinced that the average effect of a therapy is beneficial, they recommend the therapy to patients.
  • Some therapies (surgery for cosmesis, acetaminophen (Tylenol®) for headache, ibuprofen (Motrin®) for dysmenorrhea, cisapride for heartburn, etc.) are intended to make people feel better. None of these therapies is likely to improve survival, because the conditions being treated do not reduce survival.  In fact, these therapies all reduce survival: there are anesthetic deaths in cosmetic surgery, acetaminophen sometimes causes liver failure, ibuprofen sometimes causes life-threatening gastrointestinal bleeding, and so on, and there are no lives saved by way of compensation. These therapies are studied in clinical trials (a) to show that they do indeed make people feel better, and (b) to get some sort of estimate of how much harm they do. Therapies that turn out to do lots of harm for a relatively trivial feel-good effect tend not to be used (and not to be approved for marketing, if they are drug regimens), but for the most part the trade-offs cannot be definitively weighed by regulators or even by physicians. I have no way of knowing whether my patient's nose or headache or dysmenorrhea or heartburn is so troubling as to justify a 1 in 10000 or 1 in 100000 risk of death. That is the patient's decision to make. The same patient might the previous day have been skydiving, making another inalienably personal decision about a tradeoff between feeling good and the risk of death.
  With these principles in mind, I suppose you'll see why your series made me uncomfortable, and why your suggested interpretation of some of what I told you is different from what had been my interpretation of what I was saying.

  In a few cases, I believe that your wording is simply inflammatory and misleading.  I don't know of a situation in which a recommendation by an FDA medical reviewer, team leader, division director, advisory-committee member, or advisory-committee majority has been "disregarded" (your word).  The decision pyramid at FDA has certainly reversed recommendations from these sources, just as appellate courts commonly reverse the decisions of subordinate jurisdictions.  To say "disregarded" is to accuse authorities of an obliviousness of which you have no evidence.

  Incidentally, there are examples of overruling in both directions.  For example, an advisory committee in January of 1998 recommended by a vote of 9-2 that the FDA approve tasosartan for the treatment of hypertension, but tasosartan is not yet approved in the United States. (I'm sorry for this disingenuous wording, but as a matter of law (not mere practice), information as to the contents or regulatory fate of applications others than those approved is not public.)

  You are of course correct in pointing out that 6 of the 7 withdrawn drugs that you describe were never shown to save lives; you could have increased the ratio to 8 of 9 if you had expanded your coverage to include astemizole (Hismanal®) and terfenadine (Seldane®), two antihistamines that were withdrawn last year.  But in all of these cases, as with the hundreds of cosmetic or symptomatic therapies still marketed, the expected mortality effect is that mortality will be somewhat increased.  Terfenadine and astemizole are particularly instructive examples.  They were the first non-sedating antihistamines, and they were known to cause occasional lethal arrhythmias.  On the other hand, they didn't cause drivers to fall asleep at the wheel, as older antihistamines do.  Whatever you take for your hay fever, it makes you somewhat more likely to die.  When — and only when — non-sedating antihistamines that were not arrhythmogenic appeared, terfenadine and astemizole were withdrawn. (The story is not quite as slick as I wish it were; there was actually an overlap period, and many within FDA — I among them — thought that the overlap period was overlong.)

  Many other examples come easily to mind.  Isotretinoin (Accutane®) certainly doesn't save lives; it is used for nodular acne, even though as a teratogen, it ranks with thalidomide.  Until a non-teratogenic alternative comes along, it will probably remain available.

  I don't have many specific comments regarding your seven case reports, in part because what I might say is largely deducible from the above, in part because I know little about the specific facts of most of the various cases, and in small part because I have some specific knowledge that remains proprietary.  I do want to say a little about mibefradil and cisapride.

   The approval and subsequent withdrawal of mibefradil taught new lessons to the drug-development and regulatory communities, but I don't think they were the lessons that readers would naturally draw from your story.  Some FDA reviewers and two members of the Cardio-Renal Advisory Committee recommended that mibefradil not be approved, but the issue that concerned them turned out not to be a problem.  Mibefradil was withdrawn for an entirely different reason; I think it's worth going into this in some detail.

   Before approval of mibefradil, FDA (and the sponsor) had long discussions about the possibility that mibefradil might cause the life-threatening ventricular arrhythmia known as torsade de pointes.  No cases of this arrhythmia were seen during the clinical trials, but that fact is only slightly reassuring; even the worst torsade-inducers cause the arrhythmia only rarely.  Suspicions were raised by the fact that mibefradil could be shown in vitro to inhibit an important torsade-associated ion channel (IKr) found in cell membranes, and in volunteers mibefradil appeared to cause dose-related increases in an electrocardiographic measurement called the QT interval.  Substantial prolongation of the QT interval is often associated with torsade and, probably, with other life-threatening arrhythmias.  [The notion that lesser degrees of QT prolongation are associated with non-zero increases in arrhythmic risk is attractive (analogous to the level of blood pressure being coupled to the risk of stroke, both above the normal range and within it), but there are no data to support this notion, and some to refute it.]

   On the other hand, all of the known torsade-inducers have another in vitro property (prolongation of the action-potential duration) that mibefradil doesn't share.  Even more convincing, there are standardized experiments in which dogs or rabbits are by various means brought to the edge of arrhythmia, and then given doses of a suspect drug.  In these experiments, mibefradil actually reduced the likelihood that these animals would develop malignant arrhythmias.  Finally, we learned during the mibefradil evaluation that verapamil and diltiazem, two drugs marketed for over ten years, also caused impressive QT prolongations at high doses, but these drugs were (and are) believed never to induce torsade.

  The picture was mixed, and the Cardio-Renal Advisory Committee recommended approval, as you stated.  There were three dissenting votes, and in hindsight we can see that two members of the Committee wanted to wait for what would turn out to be unhelpful data, but the other dissenter was prophetic.
  •   Drs. Lemuel Moyé and Rob Califf voted to recommend against approval, urging that the Agency instead wait for the results of the MACH-1 trial, then in progress.  MACH-1 was a trial comparing mibefradil to placebo (on top of standard therapy) in patients with congestive heart failure (CHF), looking for an improvement in mortality.  Some effective antihypertensive drugs have turned out to be extremely valuable in the treatment of CHF (e.g., ACE inhibitors), but others (e.g., prazosin) have turned out to be valueless.  Mibefradil is a calcium-channel blocker, and other calcium-channel blockers have turned out to be of neutral or adverse value in CHF, but mibefradil was the first member of its chemical class, and there were mechanistic reasons suggesting that it might be beneficial.
     At the time of the mibefradil Advisory-Committee meeting, the MACH-1 trial was still ongoing, proceeding — as is necessary for credibility — with investigators and patients ignorant of which patients had been assigned to which treatment.  There had been some deaths, just about as many as would be expected in the recruited population of patients with CHF.  As is usually the case with long-running trials, the trial was overseen by a small Data Safety Monitoring Board (DSMB), privy to the treatment assignments and charged with monitoring the deaths to see if they were disproportionately occurring in one treatment arm or the other, and then with stopping the trial if a convincing imbalance arose. 
   "Breaking the blind" (that is, publicly looking at the to-date results) was not seriously considered.  Because the DSMB had not yet stopped the trial, one could be certain that the two groups were doing nearly equally well (or equally poorly).  If among the mibefradil-treated patients there had been a striking preponderance of deaths attributable to torsade, then one might have drawn adverse conclusions about mibefradil's use for the treatment of hypertension and angina, but the possibility of such an imbalance in torsade seemed unlikely when overall mortality was not significantly different between the groups.  And in any event, torsade is always rare, and the addition of the MACH-1 population did not substantially increase the studied human exposure to mibefradil.  A (non-significant) mortality trend favoring (or disfavoring) mibefradil would not have helped; some medications are useful (or not) in CHF, independent of whether they are beneficial in hypertension and angina.  For example, a drug that reduced CHF mortality by 20% would be valuable even if it led to a few drug-induced deaths per thousand patients.  The same rate of drug-induced death in an antihypertensive drug would be unacceptable, because the potential benefit of antihypertensive treatment is so small.  Breaking the blind would have prevented the investigators from continuing MACH-1 to determine whether mibefradil was useful in CHF, and that determination was substantially independent of the question of whether mibefradil should be approved for the treatment of hypertension and angina.

   Moyé and Califf favored deferring action on mibefradil until completion of the MACH-1 trial.  When the MACH-1 trial was completed in the Spring of 1998, the results showed a trend disfavoring mibefradil.  This result (that mibefradil is probably not useful for the treatment of CHF) does not (even in hindsight) help much with the question of whether mibefradil is valuable for hypertension and angina.
  •  Dr. Joanne Lindenfeld voted to recommend against approval, on the grounds that mibefradil was too potent an inhibitor of the liver enzyme called CYP3A4.  That mibefradil inhibited this enzyme was known to the Committee and to the FDA, but such inhibition had never before been the grounds for non-approval or withdrawal of any drug, here or (to my knowledge) abroad.
   At around the same time that mibefradil was approved in the United States, it was approved in several other developed countries, with the sole exception of Sweden, where regulators objected, as had Lindenfeld, to mibefradil's power as an inhibitor of CYP3A4.

   The body depends on the CYP3A4 enzyme for the metabolism of many different drugs, including cisapride and some of the "statin" anti-cholesterol drugs.  When the enzyme is inhibited, these drugs accumulate in the body, so that ordinary doses can effectively become overdoses.  Grapefruit juice and erythromycin are moderate inhibitors of CYP3A4, but neither is as potent an inhibitor as is mibefradil.  The only other inhibitor of similar potency is an anti-fungal compound, ketoconazole (Nizoral®).  When mibefradil was first approved, the labeling described its CYP3A4-inhibiting properties, and the labeling gave specific warnings about a few predictable interactions.  [Drug labels all have the same structure, so the warnings (what to watch out for if you use mibefradil) came in the usual place (after sections telling the physician why he or she might want to use mibefradil at all).  The complaint that the warnings begin on the 278th or 365th line of the label is like a child's complaint that the listing of desserts does not begin until the 114th line of the menu.]

   Over the next few months, adverse reactions associated with mibefradil were reported to FDA.  There were no convincing cases of torsade in patients who were not receiving other drugs known to be associated with that arrhythmia.  There were a few cases of extremely low heart rates, seen when mibefradil was erroneously administered at the same time as one or another drug that tended to lower the heart rate, but most of the adverse reactions were attributable to mibefradil's inhibition of CYP3A4 and consequent intoxication with other drugs.  For example, the anti-cholesterol drug simvastatin (Zocor®) causes dangerous muscle inflammation at high doses, but only very rarely at the doses commonly used.  When ordinary doses of simvastatin were given in combination with mibefradil, several cases of muscle inflammation were reported.  The labeling specifically warned against this interaction, but cases continued to occur.  These cases were for a while perceived as the "fault" of simvastatin, since they could have occurred with simvastatin alone, they could have occurred with other co-therapy (e.g., simvastatin and ketoconazole), and they could not have occurred with mibefradil alone.  They were in any event seen as avoidable, if only physicians could be made to read the labeling.

  Finally, there was a shared sudden understanding that there was no point in allocating the "fault" for these cases to simvastatin or to prescribing physicians.  Even if it were true that all of these cases might in principle have been avoided by physicians' attention to labeling, that observation was not helpful, any more than the observation that almost all auto accidents might be avoided by better driving.  People will drive imperfectly, so a car without seat belts should not be marketed.  Physicians will pay imperfect attention to labeling, so a drug whose safe and effective use requires above-average attention to labeling should also — if there are easier-to-use alternatives — not be marketed.  If it were sufficient that a drug be safe and effective when used as directed in the labeling, then thalidomide could be returned to the market as a sedative, with a label warning that it should not be given to women who might be pregnant.

  The MACH-1 trial was completed right around the time that mibefradil was being withdrawn from use for the treatment of hypertension and angina.  Mibefradil might have remained on the market (although not for the treatment of hypertension or angina) if it had turned out to be beneficial in the treatment of congestive heart failure, but it didn't, so it didn't.

   I'm much less familiar with the history of cisapride than I am with that of mibefradil.  As you point out, cisapride was approved for the treatment of nocturnal heartburn, although most of its usage was for other conditions ("off-label").  Sponsors may not legally promote off-label use, but FDA has no authority to regulate doctors' use of approved drugs.

   In our discussion about cisapride early this month, and then in your series in the Times, the pediatric use of cisapride was prominent.  I remember alluding to a paper that I wrote in 1995 (Combining methylphenidate and clonidine: the role of post-marketing surveillance. Journal of Child and Adolescent Psychopharmacology 5(3): 155-156). I did not have the paper at hand to quote during our conversation, but I gave you a pretty good paraphrase of its text
If any reasonable fraction of the thousands of children now receiving combined clonidine-methylphenidate treatment had initially been enrolled in well-controlled trials, then we could (as we now cannot) know whether this therapy is more effective than alternatives. More to the current point, we would be able to estimate (or at least to provide some sort of upper bounds for) the incidences of potential adverse drug interactions.
In particular, this is the "scandal" that I was referring to when I gave you the quote that you used in the cisapride piece.  The scandal was the absence of adequate trials.  We also talked at some length about how difficult it was to get trials performed for non-first indications, about how pediatric trials are especially difficult, and about the failure of appropriate incentives in this regard for NIH and academia (no money), FDA (no statutory authority), industry (no potential profit), and parents (not MY child).  I was disappointed to see none of these ideas in your article.

   Also, the pediatric use of cisapride that I had heard about while I was at FDA — and that formed the basis of my discussion with you — was all in neonates on ventilators. In these patients, gastric reflux is not merely distressful to child and parents. It can lead to aspiration, chemical pneumonitis, and death.  In such a setting, cisapride might (on average) be beneficial, even if it caused a few deaths.  I assume that there are no adequate trials to show that this benefit exists (if there were, the sponsor would presumably have brought the data to FDA and obtained approval), but the neonatal intensivists who used cisapride presumably all believed that this benefit was present, perhaps so obviously present that rigorous trials would be ethically difficult.  It is not rare that scattered small trials have been sufficient to convince the clinical community, even when they do not meet FDA standards.  Using whatever therapies they think are likely to bring patients unequivocal benefit is what physicians do every day.  In the absence of good trial data they might be wrong, and again, it's a scandal that adequate trials were not done.  But this is a Scandal of the Commons; everyone — see last paragraph — is waiting for the data to be gathered by someone else.  In neonates on ventilators, I doubt that anyone knows whether cisapride is harmful or beneficial.  I didn't know that cisapride had been found Not Approvable for children, but the fact does not speak for itself.  In my experience, being found Not Approvable by FDA is like being found Not Guilty in a criminal proceeding.  Perhaps he didn't do it, but perhaps he probably did do it, yet the evidence was not sufficient to exclude a reasonable doubt.  To interpret a finding of Not Approvable, one needs more information.  Was there evidence of non-efficacy, or merely non-evidence of efficacy?

   To my surprise, all of the pediatric use that you describe seems to have been for symptomatic relief, analogous to the adult use. In that setting, we're back in the world of nose jobs, headaches, and skydiving. Regulators need to reject disproportionate risks, but a risk is not disproportionate just because it is non-zero.  The sponsor and physician have duties to inform, but the patient (here, the parent) ultimately has to choose.

   Incidentally, the absence of data cuts both ways.  You were right not to pay much attention to the fact that some off-label patients are convinced that they were getting great symptomatic benefits from cisapride (see Lauran Neergaard's AP story in the Times on 10 December).  They could be wrong; homeopaths and other charlatans get endorsements all the time, and placebos are often powerful.

  Once again, I wish we'd spent more time laying the conceptual groundwork, because some of the series is first-rate.  When the Prescription Drug User-Fee Act (PDUFA) was passed, I characterized it as a Faustian bargain, and you were right on target to point out how PDUFA has warped FDA's budget to de-emphasize surveillance.  At the Center level and higher, the Agency keeps score by counting approvals.  When the Center computes workload as a metric for allocating funds, reviewers get no credit at all for intra-Agency consultation. Sooner or later, the time pressure on FDA reviewers is going to lead to a public-health disaster, even though — as Neergaard points out in another AP story on 11 December  — few of the drugs recently withdrawn had been approved unusually quickly, and I don't agree that any of them was associated with what could fairly be called disaster.

Letter of 27 November 2005 to Neelam Shah

(The motion of my right index finger is now limited to 15-25º at the distal interphalangeal joint and 30-75º at the proximal interphalangeal joint.  Dr. Shah was my assigned internist at Kaiser Permanente.)

Effective at the end of this year, I will discontinue my Kaiser membership.  

I have been more than satisfied by the service that you personally have provided during our few meetings.  I know it’s tricky to strike the right note as a physician’s physician, and you do it well.  Also, I am sympathetic to the HMO model of medical care, and I know that this is a model that Kaiser can rightly claim to have invented.

Notwithstanding all of that, I no longer have any confidence in Kaiser’s capacity to provide non-maintenance care.  Kaiser’s response to my nerve laceration last year was a series of disappointments, and I have a dysfunctional finger on my dominant hand to show for it.

The history of that fiasco may be evident in my records, but my point of view may not be apparent there.  It started with a simple kitchen accident on Christmas Day, with a 2-cm transverse laceration across the volar surface of the middle phalanx of my right index finger.  Even before I had controlled the bleeding, I determined that the tendons were functionally intact, but that I had lost radial-side sensation distal to the laceration.    I came to the Urgent Care Clinic.  The injury resembled injuries that I had seen many times during my years of emergency-medicine practice.  I expected that a hand surgeon would be called, and that either she would come in to do the repair, or she would advise me that nerve repair this distal was not often successful in patients my age.  That had been the pattern in inner-city emergency departments, and I scarcely expected less of Kaiser.

The physician in the Urgent Care Clinic was an affable endocrinologist, and he quickly determined that my tendons were functionally intact, but that I had lost radial-side sensation distal to the laceration.  Right.  After some prompting from me, he agreed to call for more specialized advice.  He made a call, skillfully closed the skin, and referred me for an appointment with Dr. Cohen in Largo a few days later.

I didn’t enjoy the drive to Largo, but I enjoyed meeting the jovial Dr. Cohen.  He is not a hand surgeon, but rather a general orthopedist.   He drew on his orthopedic experience and found that the tendons of my sutured finger were functionally intact, but that I had lost radial-side sensation distal to the laceration.  Well, yes.  He made another call, and announced that Dr. Gunther, the hand surgeon, could see me right away at Washington Hospital Center.  I drove on.

Dr. Gunther is not a full-time hand surgeon, but he does a fair amount of hand surgery as part of his general orthopedic practice.  He examined my finger, made certain determinations as to the functioning of my tendons and nerve (see above), and scheduled nerve repair for later in the week.

I came back to Washington Hospital Center for the surgery.  I expected to be asked to do a conventional 10‑minute surgical scrub of my own hand, but I was not.  The local surgical field was prepped and draped.

The surgery itself seemed uneventful.  I left with my hand in an elaborate plaster splint, held in place by occlusive tape that covered me distal to the wrist, leaving only my fingertips exposed.  I was given an appointment to return after 14 days.

During the 14 days in the splint and bandage, my palm itched, but it did not hurt, and I remained afebrile.  The splint and bandage smelled sweaty, but not foul.  When I returned to Dr. Gunther, the dressing was removed, and I was not surprised to find a miliary heat-rash-like eruption over my palm.  I was dismayed to see that the surgical wound was full of purulent drainage, but perhaps I should not have been surprised.  I learned 30 years ago that occlusive dressings magnify the risk of infection, and of course it’s still true.  Dr. Gunther irrigated out the wound, removed the skin sutures, sent a swab for C&S, and gave me a prescription for clindamycin.

Thinking of C. difficile enterocolitis, I questioned the choice of clindamycin.  Dr. Gunther did not believe that the flora of Washington Hospital Center made that antibiotic especially indicated, but it was, he said, his usual empiric antibiotic before culture results were available.  (The infecting organism was a pan-sensitive Staph. aureus; a few days later, I saw Gunther’s resident and arranged for a switch to cephalexin).

The rest is anticlimax.  The infection resolved, I came to many sessions of occupational therapy, and I’m left with a bum finger.  My whole right hand grows painfully cold for a few minutes, several times a day, probably as a manifestation of reflex sympathetic dystrophy.  At my own expense, I visited a real hand surgeon a few months ago, and he did not believe that progress was likely or that any intervention was likely to be useful.  I have given up rock climbing (I can’t operate the protective camming devices any more), and I am rather less handy around the house.  I had no intention of returning to any sort of clinical practice, but now that decision has been made for me.  Happily, I make my living by hands-free consulting, and I can type with a different finger.

I’m not sure that any part of my story amounts to negligence.  Kaiser must compete for the health-insurance dollar, and when patients compare one premium to another, they have no way to incorporate the likely time spent in queues, or in driving across metropolitan Washington for a redundant physical exam that could have been reliably performed by a third-year student.  The cost of early post-operative exams must be considered against their low yield.  These are difficult matters.

Still, I have only nine working fingers left, so I am taking them elsewhere.


Page revised: 11/29/2010 21:43