Background. Acquired resistance to standard chemotherapy for tuberculosis (TB) is an increasing problem worldwide. Vietnam has one of the highest incidences of TB and also has a large population of potential migrants to other countries. Since 1979 the international Organisation for Migration (IOM) has been running a supervised programme of TB treatment for intending migrants from Vietnam where few facilities for bacteriological culture and sensitivity testing exist, This study aimed to assess the most important factors for predicting non-response to first-line treatment as treatment starts and whether any further indicators occur during the course of treatment which may enable more accurate prediction of non-response.Methods. In all, 130 subjects failing to respond to first-line therapy (cases) between 1990 and 1995 were compared with 673 subjects who responded to therapy (controls) on various demographic and clinical characteristics using logistic regression to create a prognostic index. Variables analysed included the patient history of past TB treatment, weight, age, sex and radiological and bacteriological findings, All subjects also tested negative for HIV status.Results. The chief markers of successful response were x-ray signs and degree of sputum smear positivity, These markers provided a prognostic index with an optimal cutoff providing about 70% sensitivity and 80% specificity. Incorporating further measures obtained through the first 3 months of treatment improved the sensitivity to 80%.Conclusion. While this study enabled prediction of the majority of subjects failing to respond to first-line therapy, other factors need to be assessed before recommendations for altering treatment regimens can be made, The prognostic index could be useful in assessing subjects for closer supervision.