Hilton P. Terrell, MD, PhD
On November 6, 2000, a prepublication version of an article concerning a relationship between phenylpropanolamine and stroke was released for Internet posting.(1) The reason for the early release of the study was a judgment that the information was so critical that waiting even a few weeks would endanger the public health. Triggered by this report, all products containing the drug were withdrawn from store shelves. These actions are a nationwide equivalent of evacuating a shopping mall because a suspicious package has been found that might be a bomb. In the case of a shopping mall, however, the public would be readmitted after the package has been removed and examined. If the package were an innocent Christmas gift, efforts would be made to return it to its owner. In the case of phenylpropanolamine (PPA), however, past experience suggests that the drug has been permanently removed from the shelves, whether or not the equivalent of a bomb is established.
What kind and quality of information is required for this degree of alarm? The PPA article displays an instance in which the bomb squads are over eager, out of bounds, and risking the public's welfare. The article and the action based upon it will be pressed into service as examplar of like decisions which are not adequately founded. The error alleged is that a group holding too narrow a view of reality coupled with too much power interferes with the public's judgment, liberty, and economy.
A core error is that mere data are permitted to rule as if data could somehow speak for themselves. Facts never speak for themselves; they must be fitted into an extrinsic framework to make sense.
The framework is derived from information and judgments from other areas. Medicine must function in relation to other fields of knowledge and values. Law, social structures, economies, human aspirations and the like all must help determine how medical data are utilized.
A detailed review of the study design is not the focus of this review; rather, it is the lack of an appropriate interpretive context that is our concern. That is not to say that the study design and execution are above reproach. The design was case-control. Oddly, the stroke subjects were significantly different from the controls with whom they were compared.
The patients who suffered the hemorrhagic strokes were more likely to
be black, to be less educated, to smoke, to have a history of hypertension,
to have used cocaine recently, to use alcohol regularly, and to have a family
history of hemorrhagic stroke. These defects were answered only by statistical
evaluation. The authors tackle some other problems of temporal-precedence
bias, recall bias, and selection bias with variable persuasive power. Interpretive
possibilities different from the guilt of PPA are not offered. The only
statistically significant association of hemorrhagic stroke was with diet
pill use, not the other formulations of the drug. That finding offers
the possibility that the strokes were caused by other dieting activities,
such as abrupt increase in exercise, ketotic states induced by caloric deprivation,
or even an increase in cigarettes smoked. In other words, PPA's connection
may have been merely as a marker for the actual culprit, not as a cause.
Further, though hemorrhagic strokes can be very bad events with death or
permanent sequelae, they are not invariably so. The study does not establish
that the PPA-associated strokes were similar and serious in outcome. The
study was multi-center and captured data over a recent five and a half year
Thus, while conscientiously performed, the study has not escaped the weaknesses of case-control studies. A larger issue remains, however, even if the study is completely accurate. What does it mean? Statistical significance and practical significance are not at all the same thing. The authors touch upon this matter when they offer an estimate of the risk that one woman who takes a PPA diet pill will suffer a hemorrhagic stroke due to that action within a three day period. Their estimate is from one in 107,000 to one in 3.26 million. The relative risk had previously been estimated from the study as about 16-fold. What matters in patient care is absolute risk. If I travel to Brazil, my relative risk of being bitten by piranhas might rise 16 fold or more. My absolute risk in Brazil, however, would likely remain negligible. I would not factor piranha risk into my Brazilian travel plans, unless I had rather unusual travel plans. I would find it humorous if some well-intentioned person asserted piranha risk as worthy of serious consideration. I would find it obnoxious if my trip were banned because of the risk from the fish.
The extreme range of the absolute risk from PPA carries its own message as to the uncertainty of this whole matter of risk assessment. From top to bottom the variation is 30-fold. To achieve their estimate of absolute risk the authors used a daily incidence rate for hemorrhagic stroke for persons 35 to 54 years of age. It should be noted that this is an older group than the study's age range of 18 to 49 years of age. If the hemorrhagic stroke rate is higher in the reference group compared to the study group, then the estimate of absolute risk would also be too high.
If a mid-point risk in their range of estimates is selected there would
be one hemorrhagic stroke added by diet pill PPA per 1.5 million women in
every three day window of their use. What activities, other than taking
a diet pill, entail a similar risk? If one and a half million women drive
a car 16 miles round trip to babysit their niece while their sister receives
chemotherapy, what is the risk? Twenty-four million passenger miles doubtlessly
would result in a death or serious injury. What if they make the same trip
just to have their nails manicured? Dying is a rather high price for pretty
fingernails. What misogynistic agency is failing in its duty to protect
these women from their lethal vanity?
Liberty as Interpretive Context
Life is risk. Liberty entails risk. Risk estimates necessarily include an element of fancy. The categories upon which the risks are predicated are never a perfect fit. The risk that I will suffer neurological injury if struck from behind in a motor vehicle accident could be estimated based on data from such accidents. I have cervical arthritis. Might not that increase my risk? On the other hand, I drive a politically unpopular heavy vehicle. Might not that decrease my risk? Depending upon the categories selected or combined, the estimates would vary widely. As fallible as I am, might not I make better estimates of my chances? At exceptionally low levels of risk, is anyone able to make a useful guess?
To the more nearly ultimate point, who actually has legitimate authority
to make an estimate of risk and bind another's behavior based on their
guess? On what basis is a level of risk determined to be too great? Data
by themselves are impotent to answer this question. Parts per billion surveillance
tools are now available in medicine. We have polymerase chain reaction able
to find a single DNA fragment in a pint of tissue and radioimmunoassay able
to find a few cells with certain surface characteristics out of millions.
Centralized data bases and reporting systems can enumerate rare events in
this nation of more than a quarter billion people. We can find events never
before noted in the haystack of life. When we have found one, however, we
don't know what to do with it. The action taken on the basis of the study
at issue is an example of a misfire. We have fired up our national bulldozer
and plowed it under. Net risk may not be there or may be so small as to
fall well within the irreducible background risks of life. We might plow
under something else which we have not noted, such as liberty and self-determination.
We might back our dozer over something behind us that is totally out of
our consideration. But, we have a powerful dozer and an altruistic determination
to use it for the public good, the public's opinion, liberty, and wallet
The PPA study does not have a design adequate to fully answer the concern about PPA danger. Its implementation contained serious identifiable flaws:
1. The action taken on the basis of the outcome was drastic and premature, and indicates more about those taking the action than about PPA.
2. These events illustrate a paradigm for the operation of modern monocular, myopic medicine, tyrannized by data manipulators yet orphaned by understanding.
3. The harm in such actions to liberty, comfort, and economy, inter alia, is cumulative.
4. Liberty and economy are both related to health values indirectly and
have value in themselves which may outweigh health, in the view of many.
Summary and Conclusion
For those occupying seats of power, extraordinarily small risks now regularly impel draconian actions. Why should a government agency debar a citizen from making and acting on risk estimates, especially when they are probably very small risks? One response could be that a product marketed to the public constitutes the vendor making risk estimates for someone else. That is unavoidable. My next exhalation might carry a microbe lethal to someone else in the room. A spark from my fireplace could ignite my neighbor's roof. The venetian blind I manufacture could choke a toddler caught in its cords. Holding me accountable on a case-by-case basis, as cumbersome as it may seem, is preferable to swarms of Lilliputians affixing me to the ground with a tens of thousands of regulatory threads. A free marketplace is a ruthless corrections officer for both buyers and sellers. It is far superior to well-intended but speculative estimates of small risks.
PPA diet pills were associated with hemorrhagic stroke in women; the authoritarian response is carpet bombing, scorched earth. All PPA, in whatever formulation, is pulled from sale. Men, who were not affected, are denied its use. The possibility of unknown countervailing benefits of PPA are not even considered. The known benefits of PPA as a decongestant are decided to be unworthy of the tiny calculated risk. It is all done for the public, which need not concern itself with thinking. Organization on High has fixed the problem directly or by intimidation of its industrial servants. That Organization on High may have also imagined the problem is too complicated for the hoi polloi.
While it is a largely empirical endeavor, medicine has too exclusively
concerned itself with medical data and facts, and too little with how they
properly fit into wider arenas of human existence. Annoyed and oppressed
now by "third parties" both public and private, we have earned
some of our grief through disregard of other sources of knowledge and frames
of reference. Corporately, we were spoiled through the middle decades of
the last century into acting as though we should be supplied with whatever
power of money we wanted for our noble task. We were supplied with substantial
quantities of both, and became more specialized and sequestered from concepts
such as liberty, individual responsibility, and self-determination. The
promotions of our profession were not irrational and produced some good
outcomes, but simultaneously nourished attitudes of libertinism, irresponsibility,
and totalitarianism which are now coming to fruition. "Third party"
thuggery can be viewed as one of the ways these results are being manifested.
It is time to "out" pallid data which are pastiched into a predetermined
1. Kernan WN, et. al. Phenylpropanolamine and the risk of hemorrhagic stroke, NEJM Home, posted at http://www.nejm.org/content/kernan/1.asp. The study was published in print as: Kernan W, et. al. Phenylpropanolamine and the risk of hemorrhagic stroke, NEJM, 343, 12/21/2000, pp. 1826-1832.
Dr. Terrell practices family medicine in Florence, South Carolina
and is a member of the AAPS Board of Directors and a member of the Editorial
Board of the Medical Sentinel. His address is McLeod Family Medicine
Center, 555 East Cheves Street, Florence, SC 29506. (843) 777-2800, Fax
Originally published in the Medical Sentinel 2002;7(1):22-24. Copyright©2002Association of American Physicians and Surgeons (AAPS)