- John-Arne Skolbekken
- Accepted 29 May 1998
“When the facts change, I change my mind. What do you do?” asked Professor Michael Oliver at a presentation held in Stockholm on 6 June 1996. This is what readers of the New England Journal of Medicine were told in an advertisement for the cholesterol lowering drug simvastatin during the winter of 1996-7. Oliver's statement is accompanied in the advertisement by the following facts: “In post MI and angina patients with cholesterol levels in the range of 5.5 to 8.0 mmol/l (212-309 mg/dl). Proven to reduce the risk of total mortality by 30% (P=0.0003),1 coronary mortality by 42% (P=0.00001).2 “These facts are followed by another statement from Oliver et al: “Lower patients' cholesterol now.”3 This second statement represents Oliver's change of mind since he, in 1992, stated his “Doubts about preventing coronary heart disease: multiple interventions in middle aged men may do more harm than good,”4 which contributed to the controversy over the importance of lowering serum cholesterol concentrations.5
In a similar advertisement, readers of the Lancet have been informed of the same facts, but unaccompanied by Professor Oliver's statements. Instead, these readers are provided with more facts: “Projected 6-year benefits in 1000 patients with coronary heart disease (CHD). 35 lives saved, 67 MIs prevented, 59 procedures avoided.”
At first sight, a negative response to Oliver's rhetorical question may seem perverse or prejudiced. On second thoughts, given its rhetorical nature and considering that it is being used in a commercial presentation, you might suspect that there is something more to it. Indeed there is, and the aim of this article is to shed some light on the presentation of facts from clinical trials like the Scandinavian simvastatin survival study (4S) and the West of Scotland coronary prevention study (WOSCOPS).6
Facts from WOSCOPS were presented in a press release on 15 November 1995: “People with high cholesterol can rapidly reduce their risk of having a first-time heart attack by 31 per cent and their risk of death by 22 per cent, by taking a widely prescribed drug called pravastatin sodium. This is the conclusion of a landmark study presented today at the annual meeting of the American Heart Association. The results appear in the 16th November edition of the New England Journal of Medicine” (available on URL: http://www.gla.uk/Acad/PathBio/press_release.html).
Before moving on, the reader would do well to note that none of the above statements contains anything that is not true. There are, however, other facts, just as true, that may give a somewhat different impression.
Perceptions of risks and benefits, by both lay people and doctors, are strongly influenced by the way data are presented
In the case of possible benefits of cholesterol lowering drugs on the risk of coronary heart disease, the difference between reporting relative or absolute risk reductions can greatly influence this perception
Accordingly, data from randomised trials of such drugs can be presented in different ways
Doctors should be aware of these perception biases when studying information about drug effectiveness and when making decisions on treatment and communicating with patients
In communications with individual patients, doctors should also acknowledge that effectiveness at the group level may mean uncertainty at the individual level
When the stating of the facts change, you change your mind
Both the advertisements and the press release are based on relative risk estimates. This is an estimate that has been shown repeatedly to give a more favourable impression of the effectiveness of a drug than absolute risk estimates.7–13 These latter estimates can be found in the original articles, but, unfortunately, these are rarely read by the physicians prescribing the drugs. Table 1 gives some examples of possible combinations of absolute and relative risk reductions and shows that the relative risk reduction may remain stable across a set of various trial outcomes. It also remains true in all these examples that the risk in the placebo group is twice that in the treatment group. The real impact of the treatment, however, can only be seen by also reviewing the absolute risk reductions.
Table 2 shows some absolute risk estimates from the 4S and WOSCOPS. If the relative risk numbers in the original WOSCOPS press release were replaced with absolute risk numbers, it would then read: “People with high cholesterol can rapidly reduce their [absolute] risk of having a first time heart attack by 1.9 per cent and their [absolute] risk of death by 0.9 per cent, by taking a widely prescribed drug called pravastatin sodium.” Likewise, the effectiveness of simvastatin could be stated as: “Proven to reduce the [absolute] risk of total mortality by 3.3% (P=0.0003), coronary mortality by 3.5% (P=0.00001).” Yet another way of stating the facts would be to say that patients with angina or after a myocardial infarction may improve their probability of avoiding coronary death from 91.5% to 95% by taking simvastatin, while people without prior coronary disease may improve their probability from 98.3% to 98.8% by taking pravastatin. From earlier studies,7–13 it would seem a fair guess that these statements would leave different impressions than the original ones.
At this point it should also be added that the WOSCOPS group, a year after publishing their original study and releasing the news of their study to the press, published an article wherein they stated their belief in absolute risk estimates as the better estimate of the benefits of treatment.14 This article also contained absolute risk estimates for various subgroups in the study, calculated in an attempt to identify high risk individuals who should be subject to statin treatment. This strategy for selection of high risk patients was attributed to evidence that hypercholesterolaemia alone was not a good indication for statin treatment and that a population based strategy would prove too costly for countries with restricted healthcare budgets. This statement that hypercholesterolaemia alone is an insufficient indication for treatment may lead people to see the press release's primary focus on risk reduction in people with high serum cholesterol as a generous interpretation of the facts.
One for all, all for one
Another point that can be made about the original statements is the phenomenon called pseudocertainty that is, that “outcomes that are merely probable are underweighed in comparison with outcomes that are obtained with certainty.”15
“Proven to reduce the risk…” and “People with high cholesterol can rapidly reduce their risk…” may easily be understood to mean guaranteed improvements rather than improved probabilities. There is, on the contrary, considerable uncertainty with regard to which individuals will benefit from statin treatment. As it is, some people taking statins will suffer a coronary death, just as most people not taking them will live. Apart from taking in these implications of the risk estimates, this uncertainty is further illustrated by the number needed to treat (table 3).16
Prescriptions of cholesterol reducing drugs have risen explosively in recent years. One reason for this may well be messages attributing certainty when uncertainty prevails. To illuminate the mechanisms of the explosion, we have calculated the number of tablets needed to be taken (suggested abbreviation TNT) to save one life (table 4). Based on the number needed to treat and the number of tablets needed to be taken, we are able to state that, based on data from WOSCOPS, “Medicine is not an exact science. Therefore, 200 men without any prior heart disease have to swallow 357 700 tablets over five years to save one of them from dying from coronary heart disease. This is due to the fact that no exact knowledge exists as to whom of these 200 will benefit from the treatment.”
Another mechanism involved here is that of overconfidence in current scientific knowledge.15 An example of this is the statement of the editor of the American Journal of Cardiology that statins are miracle drugs that “are to atherosclerosis what penicillin was to infectious diseases.”17 Despite whatever present argument exists over the long term effects of penicillin and other antibiotics, their effectiveness was never based on treating hundreds of patient daily for half a decade to cure the infection in one of them. Such overconfidence may easily be built on relative risk estimates, but also on a failure to acknowledge that the aetiology of coronary heart disease still has many unknown components. Indeed, if the recent attention to the possible contribution of infections to coronary heart disease proves to be right,18–21 then the statement may be that penicillin is to atherosclerosis what it was to other infectious diseases.
One way of informing patients has been suggested by Hanne Hollnagel, professor of general practice in Copenhagen.22 Her strategy is based on the principles of acknowledging that there is uncertainty involved, avoiding value laden words such as “risk” or “chance,” using absolute estimates, and avoiding relative estimates. For a patient fitting the inclusion criteria for the 4S, her message would then be: “If 100 people like you are given no treatment for five years 92 will live and eight will die. Whether you are one of the 92 or one the eight, I do not know. Then, if 100 people like you take a certain drug every day for five years 95 will live and five will die. Again, I do not know whether you are one of the 95 or one of the five.”
Another issue is the sex of the beneficiaries of statin treatment. Surely, it seems reasonable for readers of the advertisements and the press release to assume that “patients” and “people” include both sexes. However, WOSCOPS was strictly men only, and, although 4S included women, there was no reduction of total mortality for the female patients in the study while their absolute risk of a coronary death was reduced from 4.0% to 3.1%.1 It would, of course, be premature to say that women will not benefit from statin treatment, but it should be acknowledged that such benefits have not been shown by these two studies.
Another question is whether people and patients mean individuals or groups. “People with high cholesterol can rapidly reduce their risk of having a first-time heart attack by 31 per cent,” may be interpreted as all individuals taking the drug will have a 31% reduction of their personal risk. However, this statement is true only for groups of men, a fact that should not be forgotten in consultations with individual patients.
Considering that “the public has a right to information and to be involved in the choice of treatment,”23 doctors have to communicate the risk reduction achieved by cholesterol reducing drugs to their patients. Advocating patient autonomy may not be done with a light heart, however, as a patient's choice may not always be what a doctor would like it to be. Belief in patient autonomy may indeed prove to be a bitter pill for doctors to swallow when they consider that “How people perceive health issues and risk and how they make choices about their own behaviour do not always fall into a rational pattern.”23 If a doctor further believes that a patient's need for statin treatment is a result of irrational lifestyle choices it would seem paradoxical to give the patient the freedom of choice with regard to treatment. On this issue, the doctor has no option, however, as the decision on whether to swallow a pill always lies with the patient.
Then again, what exactly is a rational choice in these matters? A possible answer is that there are several rationalities involved, reflecting various interests. In this case there are at least those of the individual patient, of public health, of the medical profession, and of the pharmaceutical industry. What makes the choice difficult is that it is hard to tell when these interests overlap and when they do not.
In the era of evidence based medicine it is no surprise that measures of clinical effectiveness are focused in drug advertisements. Indeed, this is a most rational strategy given the circumstances. As has been shown, however, we would do well to look beyond the face value of facts stated as relative risk reductions and keep in mind that effectiveness at the group level may mean uncertainty at the individual level.
I thank Olav Helge Førde, Floyd Rudmin, and Steinar Westin for constructive comments on drafts of this manuscript.
Conflict of interest: None.