OBJECTIVE: To determine if it matters whether we use community or patient preferences in cost-utility analysis. METHODS: Patients were randomized within 6 weeks of acute myocardial infarction to a 2-month cardiac rehabilitation intervention (n=99) or to usual care (n=102). Data were collected at baseline and at 2, 4, 8, and 12 months. Community-based preferences for patients' health states were measured using the Quality of Well-Being (QWB). Patients' preferences for their subjectively-defined health states were measured using the Time Trade-off (TTO) technique. Agreement between QWB and TTO measures was assessed using intra-class correlation coefficient (ICC). Responsiveness of each measure was calculated as the standardized response mean (SRM). Quality-adjusted life years (QALYs) experienced by each patient were estimated separately using both QWB and TTO measurements. Costs, measured from the societal perspective for each individual patient, included those borne by the health-care system and the rehabilitation program and the patients. Incremental cost-utility ratios were estimated using mean costs and QALYs for the study groups. QWB-based and TTO-based cost-utility results were compared. RESULTS: Agreement between QWB and TTO scores varies from negligible (ICC=0.069) at baseline to strong (ICC=0.607) at 12-month assessment. TTO scores are higher than QWB scores (p<0.01). QWB and TTO scores for both groups of patients improved (p<0.05) between baseline and 12 months. SRMs are 0.64 for QWB and 0.34 for TTO. QALYs gained by rehabilitation are 0.011 using QWB and 0.040 using TTO, at a cost (US $ 2001) of $702 per patient. The cost-utility of rehabilitation is $62,000 per QWB-based QALY gained and $17,500 per TTO-based QALY gained. CONCLUSIONS: The QWB and TTO results are different. This may not be generalizable but is cause for concern because it suggests that the cost-effectiveness of an intervention may differ depending upon whether community or patient preferences are used.