Different Ways to Describe the Benefits of Risk-Reducing Treatments Peder A. Halvorsen, MD; Randi Selmer, PhD; and Ivar Sonbo Kristiansen, MD, PhD, MPH. Ann Intern Med, June 19, 2007; 146(12): 848-856
Knowing Number Needed to Treat May Help Patients Consent to Treatment Interventions News Author: Laurie Barclay, MD; CME Author: Charles Vega, MD; Medscape: June 20, 2007. CME
The Study
Excerpted from abstract:
The survey presented scenarios regarding a hypothetical drug therapy to reduce the risk for heart attacks (1754 respondents) or hip fractures (1000 respondents). The data sources for both scenarios were clinical trials. Respondents were randomly assigned to a scenario with 1 of 3 outcomes after 5 years of treatment. For the drug to prevent heart attacks, the outcomes were postponement by 2 months for all patients, postponement by 8 months for 1 of 4 patients, or an NNT of 13 patients to prevent 1 heart attack. For the drug to prevent hip fractures, the outcomes were postponement by 16 days for all patients, postponement by 16 months for 3 of 100 patients, or an NNT of 57 patients to prevent 1 fracture. … The overall rate of response to the survey was 81%. In the heart attack scenarios, 93% of respondents who were presented with the NNT outcome consented to drug therapy, 82% who were presented with the outcome of large postponement for some patients consented to therapy, and 69% who were presented with the outcome of short postponement for all patients consented to therapy (chi-square, 89.6; P less than 0.001). Corresponding consent rates for the hip fracture scenarios were 74%, 56%, and 34%, respectively (chi-square, 91.5, P less than 0.001). Respondents who said that they understood the treatment effect were more likely to consent to therapy.
Commentary
Summarizing their results, the study’s authors write
Treatment effects expressed in terms of NNT yielded higher consent rates than did those expressed as equivalent postponements. This result suggests that the description of the anticipated outcome may influence the patient’s willingness to accept a recommended intervention.
While the difficulty of communicating statistically valid information about possible outcomes to medical professionals as well as patients has long been recognized, this study is valuable for its emphasis on the impact this process has on treatment decisions made by the patient.
On the other hand, I am not convinced that the specific results (e.g., that using the number needed to treat results in higher consent rates than postponement of events) has been proven by this study’s use of hypothetical scenarios rather than actual clinical situations, especially when the patients’ understanding of the scenarios is unclear.
I do agree with the editorial by Harold C. Sox, MD, in the same issue of Annals of Internal Medicine,
Halvorsen and colleagues’ findings are welcome because they remind us about framing effects and show how they could affect decisions about preventative interventions. They show that it is hard to communicate the risk for events realistically and without biasing our patients. Hopefully, this reminder will stimulate researchers, medical students, and practicing physicians.
Because it is beyond the legitimate scope of the article, an important issue on which this editorial statement touches but does not elaborate is the ethics of framing information. It is by no means certain, for example, that increasing the proportion of patients agreeing to treatment is necessarily the “good outcome.”
Footnotes
__________
