RESUMO
Our objective was to evaluate the effect of rifabutin prophylaxis in patients with AIDS and CD4 counts of less than 200 per cubic millimetre using a combination of Q-TWiST (quality-adjusted time without symptoms and toxicity) and multiattribute health utility assessment. The design consisted of a secondary analysis of two previously reported multicentre, randomized, placebo-controlled clinical trials conducted in 78 academic, community and Department of Veterans Affairs HIV centres and private practices. 542 patients with AIDS and CD4 counts of less than 200 per cubic millimetre were assigned to rifabutin 300 mg/day and 562 were assigned to a placebo. A modified Q-TWiST approach was used for comparing treatments based on the occurrence and duration of time with and without severe symptoms and clinical endpoints. Health states were constructed to represent combinations of clinical events experienced by study patients. Five physicians assigned utilities for health states using a six-attribute health classification system. These utilities were used to adjust survival for QOL. The rifabutin and placebo groups were compared using estimated quality-of-life-adjusted days. The incidence of MAC was 9% for the rifabutin group and 18% for the placebo group (p < 0.001). Differences, although not statistically significant, were observed for rates of survival and hospitalization. The rifabutin group experienced less anaemia (p < 0.02), and fever and night sweats (p < 0.02) than the placebo group. Average Q-TWiST days were 325 for the rifabutin group and 309 for the placebo group (p < 0.05). Q-TWiST days were significantly lower for patients with MAC bacteraemia (p < 0.04) and hospitalizations (p < (0.003). Rifabutin prophylaxis resulted in fewer MAC infections and greater quality-of-life-adjusted days of survival compared to no rifabutin. Quality-of-life-adjusted survival, based on a combination of the Q-TWiST and multiattribute health utility index, is a feasible approach for evaluating the outcomes of medical treatment. Future studies should, however, use patient-assigned utility weights to compute Q-TWiST scores, since physician generated utilities may differ significantly from those of patients.