A multivariate analysis of prognostic factors was carried out on 301 patients with clinical or pathological stage III/IV Hodgkin's disease treated using the same combination chemotherapy (MVPP) at two centres (Christie Hospital, Manchester, 151 patients, St. Bartholomew's Hospital, London, 150 patients). There were no significant difference in CR or relapse free and overall survival at 5 and 10 years between the two groups. Cox analysis of the Christie data alone produced four significant factors for survival - age, sex, lymphocyte count and stage. The latter three factors showed the same trend for the St. Bartholomew's Hospital patients but failed to reach statistical significance. Analysis of the combined data showed all four factors to be of importance in predicting survival. Three different prognostic groups were identified which separated patients with good, intermediate or poor prognosis in both centres. The good prognostic group included patients aged less than 45 years, lymphocyte count greater than 0.75 x 10(9) l-1 and female patients with stage IIIB disease (5 year survival 85%). The rest were of poorer prognosis with male stage IV patients faring particularly badly (5 year survival 40%). Problems associated with the use of multivariate analysis to produce useful prognostic groupings in patients from different centres, are discussed.