OBJECTIVES: To assess the validity of patient preference assessments of short-term health states associated with endometriosis pain treatments; and to compare individual treatment recommendations generated by a decision-analytic model incorporating patients' rating scale (RS) versus modified time tradeoff (mTTO) values. METHODS: Seventy patients with endometriosis pain valued nine health states associated with 3 equally-effective treatments with different side effect profiles (danazol, GnRH agonist, laparoscopy), using RS and mTTO with sleep as the trading metaphor. With mTTO, subjects chose between spending the next month in each health state versus spending X days "asleep" (an unrestful period of time one would miss) and 30-X days in good health with no symptoms. We examined the internal and across-method consistency of the valuations by checking for violations of logical adherence (valuing states with increasing adverse effects progressively worse) and procedural invariance (consistent rank ordering of states across assessment methods). Finally, we incorporated individual patients' valuations into a decision-analytic model and assessed the concordance of treatment recommendations based on mTTO and RS values (converted into utilities with a power transformation). RESULTS: Mean mTTO values in order of increasing side effects were: danazol - 0.96, 0.80, 0.58, 0.47; GnRH agonist - 0.97, 0.78, 0.59; laparoscopy - 0.77, 0.48. The vast majority of subjects (97% with RS; 99% with mTTO) consistently valued states with increasing morbidity progressively worse; and 97% valued each treatment's health states in a consistent rank order with RS and mTTO. Based on mTTO valuations, individual treatment recommendations were: 42% laparasocpy, 38% danazol, and 20% GnRH agonist. However, recommendations based on transformed RS values differed for 49% of subjects. CONCLUSIONS: Both RS and mTTO demonstrated internal and across-method consistency in assessing the temporary health states associated with endometriosis treatments. However, the choice of assessment method can greatly affect decision-analytically derived treatment recommendations made to individual patients.