Summary
Background and aims: Diagnosis, treatment and follow up of risk conditions such as hypercholesterolemia and osteoporosis are prominent tasks of contemporary medical practice. The aim is to prevent, or at least postpone the onset of adverse health outcomes such as angina pectoris, heart attacks, strokes and fractures. Dealing with risks involves decision making under uncertainty. For patients to be able to engage meaningfully in shared decision making, benefits of risk reducing interventions must be communicated in easily comprehensible ways. From randomised controlled trials effectiveness of such interventions may be estimated and conveyed in traditional formats such as relative risk reduction, absolute risk reduction or number needed to treat (NNT). Alternatively, to account for the time dimension, prolongation of (disease free) life or, equivalently, postponement of adverse events may be used. There is ample evidence that the different formats for risk reductions yield different decisions, i.e. framing effects. The most consistent finding is that decision makers are more inclined to accept interventions when risk reductions are explained in relative rather than absolute terms. To some extent decisions on hypothetical drug therapies by NNT and postponement of adverse outcomes have been studied empirically. It appears that lay people are insensitive to effect size in terms of NNT but sensitive to the length of postponement. The aim of this Ph.D. study was to explore how physicians and lay people understand and respond to the concepts of NNT and postponement when making decisions about risk reducing interventions against cardiovascular diseases and osteoporosis. The thesis encompasses four scientific papers covering the following research questions:
When considering risk reducing drug therapies, are lay people sensitive to effect size in terms of NNT for different diseases, treatment costs and interpretations of NNT?
Are medical doctors sensitive to the magnitude of NNT when they consider recommending a cardioprotective drug therapy?
Are lay people affected by personal risk information when considering how long they expect to live?
Do lay people respond differently when risk reductions are explained in terms of NNT rather than postponement of adverse events?
Materials and methods: Attendees to a health study (n=2754), medical doctors (n=1616) and a sample of the general population (n=2000) were approached in three different surveys, which shared a common design: Respondents were presented with hypothetical clinical scenarios or vignettes regarding long term preventive drug therapies. Important aspects of the therapies, in particular effect measures such as NNT or postponement of adverse outcomes, were varied across the scenarios. The respondents were randomly allocated to one version of the scenario and asked about their preferences for the intervention. Differences in responses to the different scenarios were analyzed using statistical methods as appropriate. In the fourth project a subset of the attendees to the health study (n=1748) were asked a simple question about their anticipated longevity; did they expect to live shorter, longer or about as long as the mean for Norwegians? Differences in anticipated longevity between respondents with high and low cardiovascular risk respectively, were analyzed.
Results:
When lay people considered risk reducing drug therapies, the proportion consenting to therapy was fairly constant over a broad range of NNTs (50 to 1600). This pattern was consistent across different diseases to be prevented, different treatment costs and different interpretations of NNT. Among medical doctors the proportion recommending a cardioprotective drug therapy dropped by 20% when NNT increased from 50 to 200. Furthermore, lay people were more inclined to accept drug therapies explained as preventing interventions in terms of NNT compared to drugs conceptualized as postponing interventions. For example, in the context of heart attack prevention, 93% consented when informed in terms of NNT, 82% consented when informed in terms of a long postponement (8 months) of heart attack for one out of four patients, whereas 69% consented when informed in terms of a short postponement (2 months) for all patients.
Finally, lay people’s anticipated longevity was moderately associated with personal cardiovascular risk information; odds ratio for high risk versus low risk individuals was 2.4 (95% CI 1.7 – 3.3) per level of anticipated longevity (shorter, about as long as, and longer than the mean, respectively).
Conclusion and implications:
The major finding of this project was that lay people were more inclined to accept prophylactic drug therapies when risk reductions were explained in terms of number needed to treat to prevent one unfavourable outcome rather than postponement of adverse events. Second, it was confirmed that lay people are insensitive to the magnitude of NNT in complex decisions. Medical doctors, on the other hand, were sensitive to effect size in terms of NNT. Finally, there was a statistically significant but modest association between personal risk information and anticipated longevity. For clinical practice implications are that NNT as well as postponement should be used with caution when explaining risk reductions to patients, but that NNT may be suitable for communication between medical doctors. For further research the findings pose questions about how NNT and postponement would affect real life decisions. Second, if effect size does not really matter, what goals are important and what do patients expect to achieve when considering a risk reducing drug therapy? Finally, whether the link between personal risk information and anticipated longevity is emotional or cognitive in nature might be explored.
Olaf G. Aasland, tidligere leder av Legeforskningsintituttet, har vært medveileder, og er medforfatter på en av de fire artiklene som danner bakgrunnen for Halvorsens doktorgradsarbeid (Medical doctors` perception of the "number needed to treat" (NNT). A survey of doctors´ recommendations for two therapies with different NNT. Halvorsen PA, Kristiansen IS, Aasland OG, Førde OH / Forde OH. Scandinavian Journal of Primary Health Care 2003; 21:162-6).
Hovedveileder har vært Ivar Sønbø Kristiansen, tidligere professor ved Universitetet i Tromsø.