The inter-relationships between relapse, treatment regimen and compliance in 1009 patients treated for tuberculosis between 1978 and 1987 are reported. Multiple linear logistic regression using relapse as the dependent variable was used because of the complex relationships between treatment, compliance, age and calendar year of treatment. Compliance (P < 0·0001) was the major determinant of relapse, but age (P = 0·047) was also significantly associated. Relapse was not related to sex, site of disease, ethnic group or the presence of multiple disease sites. Although the regimen given, the total duration of treatment and treatment with or without pyrazinamide were not statistically associated with relapse, pyrazinamide containing regimens were associated with better compliance. Compliance was best in those aged 60 years and over and worst in those aged 15–29 years. Some of the non-compliance leading to relapse was only uncovered by close Health Visitor surveillance and not by the physician supervising treatment, emphasizing the important role of field staff in patient follow-up. © 1991, Baillière Tindall All right reserved. All rights reserved.
