BACKGROUND The treatment of pulmonary disease caused by opportunist mycobacteria is controversial. It is uncertain whether in vitro sensitivity testing predicts clinical response in the way it does forMycobacterium tuberculosis. The literature suggests that the combination of rifampicin (R) and ethambutol (E) is important whereas isoniazid (H) may not be, but to date there have been no published reports of randomised controlled trials in the treatment of these conditions. The British Thoracic Society has conducted the first such trial, a randomised study of two regimens in HIV negative patients with pulmonary disease caused by M avium intracellulare (MAC),M malmoense, and M xenopi.

RESULTS From October 1987 to December 1992, 141 physicians entered 223 patients (106 with M malmoense, 75 with MAC, 42 withM xenopi). At entry the RE and REH groups were comparable over a range of demographic and clinical features. For each species there was no significant difference between RE and REH in the number of deaths, but when the three species were combined there were fewer deaths from the mycobacterial disease with RE (1%v 8%, p=0.018, odds ratio 0.10, exact 95% CI 0.00 to 0.76). For M malmoense the failure of treatment/relapse rates did not differ appreciably between the regimens, but for MAC there were fewer failures of treatment/relapses with REH (16% v 41%, p=0.033) With M xenopi there was a non-significant trend in the same direction (5%v 18%, p=0.41) and when all three species were combined there was a significant difference in favour of REH (11%v 22%, p=0.033). There was no correlation between failure of treatment/relapse and in vitro resistance.M xenopi was associated with the greatest mortality (57% at 5 years), MAC was the most difficult to eradicate, and M malmoense had the most favourable outlook (42% known to be alive and cured at 5 years).

CONCLUSIONS The results of susceptibility tests performed by the modal resistance method do not correlate with the patient's response to chemotherapy. RE and REH are tolerated better than previous regimens containing second or third line anti-mycobacterial drugs. Treatment ofM malmoense with RE for 2 years is preferable to REH. The addition of H reduces the failure of treatment/relapse rates for MAC and has a tendency to do so also forM xenopi, but there is a suggestion that REH is associated with higher death rates overall. Better regimens are required.