In order to assess the risk and predictors of mefloquine resistance we monitored a cohort of 113 patients in eastern Thailand who had been treated for uncomplicated falciparum malaria with a single dose of 15 mg/kg of the drug and followed up for 42 days. The overall treatment failure rate at day 42 was 59.1% (95% confidence interval (CI) = 50%, 68%) with only 2.7% of the patients being lost to follow-up. There were 6.4% RIII, 20.9% RII, 31.8% RI, and 40.9% sensitive responses, based on a modified WHO classification. A low haemoglobin level on the day of treatment and diarrhoea during the first two days after treatment were independent predictors of treatment failure. These findings remained statistically significant in a Cox proportional hazards model, after controlling for other baseline characteristics and adverse effects. Although a history of digestive disorders prior to treatment was associated with diarrhoea on day 2 (P = 0.024), it was in itself not a predictor of treatment failure (adjusted hazard ratio = 1.16; 95% CI = 0.35, 2.14). A total of 60 patients with an R response were hospitalized for 7 days to receive supervised treatment with quinine-tetracycline. Only three had a positive thick smear for asexual forms of Plasmodium falciparum 14 days later, and quinine-tetracycline therefore remains a good alternative treatment for mefloquine-resistant falciparum malaria.