Purpose: To examine how sensitive physicians are to information format (gain in disease free survival or the equivalent number needed to treat (NNT)) when asked whether they would recommend a lipid lowering drug treatment.
Methods: Random samples of general practitioners (n=450) and internists (n=450) were mailed a vignette presenting a male patient with an unfavourable cardiovascular risk profile and a hypothetical drug, “Neostatin”. The benefit of “Neostatin” was described in terms of the NNT to observe one less patient with cardiovascular disease or, alternatively, the average gain in survival without cardiovascular disease. We used a statistical model of cardiovascular disease prevention to estimate three levels of effectiveness in terms of NNT or corresponding disease-free survival. To ensure clinically relevant numbers we used simvastatin as our reference drug in the model. This procedure yielded 6 versions of the vignette, in which NNT after 24 years of therapy was set at 9, 17 (reference) and 34 respectively, whereas the average gain in disease free survival was 32, 17 (reference) and 9 months, respectively. Each physician was randomly allocated to one version of the scenario. We asked them to evaluate “Neostatin” on a Likert scale from zero (a very poor choice) to ten (a very good choice) and whether they would recommend “Neostatin” for the patient. We used logistic regression and the non-parametric Jonckheere-Terpstra test to analyse trends in proportions and ratings, respectively.
Results: The response rate was 53 %. With NNT set at 9, 17 and 34, 43%, 31% and 26% would recommend “Neostatin” (p for trend=0.02), respectively. For corresponding gains in disease free survival (32, 17 and 9 months), 52%, 49% and 40% respectively would recommend the drug (p for trend=0.10). On the rating scale mean values were 5.5, 5.0 and 4.7 across the respective NNT scenarios and 6.1, 6.2 and 5.2 across the scenarios presenting survival gains (p for trend=0.05 for NNT as well as survival gains). At an aggregate level 34% recommended “Neostatin” when presented with NNT compared to 47 % of those presented with gain in disease free survival (chi-square=8.8, p=0.003).
Conclusion: Physicians presented with NNT were less likely to recommend a hypothetical therapy than those presented with gain in disease free survival. Physicians were sensitive to effect size in terms of NNT.
Candidate for the Lee B. Lusted Student Prize Competition