There is a suggestion that drug resistance rates decreased in developing countries over the period 1962-85, while recent data suggest that resistance may be increasing. The initial decrease in resistance appears to be associated with well-functioning National Tuberculosis Control Programmes (NTP), while the recently observed increase may be due either to understaffed, resource-poor programmes or to the effect of the HIV epidemic, or to both. It is possible that the HIV epidemic may overwhelm the NTP, resulting in decreased programme efficiency and ultimately increased drug resistance. Resistance surveillance appears to be a good measure of programme efficiency. For research purposes, primary drug resistance surveys should be done on a sample of relevant patients which includes and distinguishes between HIV-positive and HIV-negative patients. At this time, there is not enough information to warrant a recommendation regarding HIV testing of TB patients for surveillance purposes. In order for resistance surveys to be relevant from the public health perspective, one must know the proportion of patients presenting for treatment having previously received treatment. The meaningful denominator for drug resistance surveys from the programme evaluation perspective should be the number of patients presenting for treatment. For initial drug resistance surveys the measurement should be the number of people never having received prior TB treatment with resistant bacilli, divided by the number of new patients presenting for treatment. For acquired resistance, one should look at all patients who begin treatment with susceptible bacilli who become resistant 6 months later.