Skip navigation

Pills, Purchasing & Presentation

Readers will be well aware that Bandolier has a bee in its bonnet about the way that information is presented. The number-needed-to-treat (NNT) has been used frequently to convey the numerical power of clinically relevant end-point.

If the manner of presentation affects the way in which the information is used, then presentation becomes as important or even more important than the information itself.

Perhaps this is self-evident. It explains the clever advertisements which appear in our papers, on our TV screens, and in many of the medical and scientific journals we read.

Pills


The "Dear Doctor" letter from the Committee on the Safety of Medicines of October 18 stated that "combined oral contraceptives containing desogestrel and gestodene are associated with around a two-fold increase in the risk of thromboembolism". It is not easy to advise an anxious patient on the basis of this statement.

The Sunday Times of October 22 gave some numerical estimates of the increased risk. With the suspect pill the risk "appeared to be double that of other, older brands - at a rate of 30 incidents per 100,000 women". It also helpfully stated that "the risk for a healthy woman not on the pill is 5 per 100,000, while the risk of thrombosis during pregnancy is 60 per 100,000".

Bandolier has no knowledge that these numbers form the the evidence upon which the CSM based its decision. They can, however, form the basis on which to calculate NNTs for this intervention (though perhaps here we should be using the NNH - numbers-needed-to-harm!). In any event, using these numbers Bandolier calculates that:
  • Compared with other pill brands there is an increased risk of a thrombosis (odds ratio 1.95, 95%CI 1.1 - 3.5). One woman in 6667 on the suspect brands would have a thrombosis who would not have had a thrombosis if she had been taking other brands (NNT 6667, 95%CI 3553 - 53950).
  • Compared with not being on the pill there is an increased risk of a thrombosis with the suspect brands (odds ratio 4.2, 95%CI 2.2 - 8.1). One woman in 4000 on the suspect brands would have a thrombosis who would not have had a thrombosis if she had not been on the pill (NNT 4000, 95%CI 2733 to 7459).
  • Compared with being pregnant there is a significantly lower risk of thrombosis with the suspect brands (odds ratio 0.5, 95%CI 0.34 - 0.78).

Bandolier believes that these clinically interpretable numbers, one in 6667 versus other pill brands and one in 4000 versus no pill, would have made the 'Dear Doctor' letter more useful (above and beyond all the arguments about the process of letting everyone know).

Many other questions are, of course, left unanswered until the data from the trials on which the CSM based its conclusions are published. Was there, for instance, any evidence of some benefit from the use of the suspect pills which could be set against the increased risk of thromboembolism?

Purchasers and presentation


The importance of the way in which information is presented is emphasised by Fahey et al [1], who gave 182 health authority members results from a randomised trial on breast cancer screening and results from a systematic review on cardiac rehabilitation. The results were presented to them (shown in the table below) in four different ways:
  • relative risk reduction
  • absolute risk reduction
  • proportion of event-free patients
  • numbers of patients treated to prevent one death

From the 140 questionnaires returned the willingness to fund either programme was influenced significantly by the way in which results were presented.

Relative risk reduction produced significantly higher inclination to purchase, followed by NNT. Intriguingly only three respondents, "all non-executive members claiming no training in epidemiology" said that they realised that all four sets of data summarised the same results.

Doctors and presentation


It is not only members of health authorities who are susceptible to altered perceptions of effectiveness according to the way in which the results of studies are presented to them. Two studies have looked at the effects of presentation on decisions by doctors in teaching hospitals in Canada [2] and on GPs in Italy [3]. Both used data from the Helsinki heart study.

Hospital doctors


In the first of these studies, David Naylor and colleagues [2] compared clinicians' ratings of therapeutic effectiveness by looking at different end-points presented as percent reductions in relative risk, absolute risk, and numbers-needed-to-treat. The study was conducted using random allocation of questionnaires which used relative data or absolute data, each with NNT data, among doctors of various grades at Toronto teaching hospitals. They used an 11-point scale anchored at "no effect" and running from -5 "harmful" to +5 "very effective".

Relative presentation consistently showed a tendency to higher scores - that is the intervention was interpreted as being more effective. Where data from a single end point, that for any myocardial infarction, was examined, both relative and absolute comparison was scored consistently higher than NNT presentation of the same data.

The results are shown in the figure above. NNT reporting of the same information produced a reduction of about 2 points in the effectiveness scale, reducing the judgement from quite effective to one of only slight effect.

General practitioners


This second study [3] presented information to 148 GPs using information from the trial as if it referred to five different drugs. The presentations were:-
  • relative risk reduction
  • absolute risk reduction
  • difference in event-free patients
  • NNT to prevent one event
  • events reduction and mortality

For each statement about effects, the GPs were asked to mark a 10 cm line labelled "I would definitely not prescribe this drug" on the left and "I would definitely prescribe this drug" on the right. The statements were presented in random sequences.
The results are shown in the figure. Presentation as relative risk reduction produced a very large tendency towards prescribing with a mean score of 7.7 out of 10. All other presentations produced scores of between about 2.5 and 3.5.

Comment


This is all very interesting, but what does it mean? One simple message comes from all three studies. For those who want to influence others, use the relative risk reduction as your means of presenting data. For those who are likely to be influenced by data presentation, never, ever, accept information on the basis of relative risk reduction alone.

There is a more complicated message, for which, to some extent, we need to answer the question about what should be the "true" or "appropriate" response to these various forms of data presentation? This is not answerable, however, at least not in any simple way.

A secondary response is then to ask whether you actually understand the result - is the difference in event free patients or absolute risk reduction immediately useful to you? Would you rather have a few words which conveyed the message - something like "X patients have to be treated for one to benefit"? Perhaps that is why NNT is becoming the presentation method of choice.

In any event, armed with this information it will be an interesting experience to examine articles, talks and advertisements to see how the information is being presented.

References:

  1. T Fahey, S Griffiths, TJ Peters. Evidence-based purchasing: understanding results of clinical trials and systematic reviews. British Medical Journal 1995 311: 1056-60.
  2. CD Naylor, E Chen, B Strauss. Measured enthusiasm: does the method of reporting trial results alter perceptions of therapeutic effectiveness? Annals of Internal Medicine 1992 117: 916-21.
  3. M Bobbio, B Demichelis, G Giustetto. Completeness of reporting trial results: effect on physicians' willingness to prescribe. Lancet 1994 343: 1209-11.



previous or next story in this issue