BEWARE THE LAW OF AVERAGES: EVALUATING TASTE HETEROGENEITY IN BENEFIT-RISK TRADEOFF PREFERENCES

Purpose: Understanding patients' willingness to accept treatment-related risks of serious adverse events is important in selecting optimal treatments.  However, common practice of reporting average estimates of risk tolerance could lead to incorrect inferences for many patients.  This study assessed heterogeneity in patients' risk tolerance for serious infection, cancer, and surgery with Crohn's disease (CD) medication treatments.

Methods: We administered a discrete-choice experiment survey to patients enrolled in an online panel of CD patients. Participants completed 8 choice questions that required comparing two constructed medical therapies. Each treatment option was characterized by differing levels of time with active disease symptoms, severity of symptoms, duration of therapy with corticosteroids, and risks of serious infection, cancer, and requiring surgery. Scale-adjusted latent-class choice models identified groups of patients with similar treatment-outcome preferences and risk tolerances.   

Results: The analysis identified three distinct subgroups in the sample of 811 participants characterized by: 1) those disproportionately concerned with duration of active symptoms (61%), 2) those disproportionately concerned with steroids (25%), or 3) those disproportionately concerned with treatment-related risks (14%). (See Figure.) Class membership was significantly correlated with severity of current and most-recent relapse symptoms, experience with CD complications, and having a family member with cancer. The more risk-tolerant respondents more often were under age 44 (p<0.01) and were more likely to be male (p<0.01).   For a treatment that would avoid 12 months of moderate CD symptoms, patients in the 3 classes would accept serious-infection risks of more than the maximum 30% evaluated in the study design, 12%, and 16%, respectively.  Conversely, for a treatment with a 5% increased risk of requiring surgery, patients in the 3 classes would require reductions in moderate symptoms of 1.2 months, 4.5 months, and more than the maximum 12 months evaluated in the survey design.

Conclusions: There is substantial preference heterogeneity among patients with CD for medications that vary in efficacy and potential harms. Avoiding the ecological fallacy of assuming all patients have mean-value preferences could result in significant differences in the conclusions of studies intended to support clinical and regulatory decision making.