A total of 60 patients underwent injection of prostaglandin (PG)F2 alpha into the affected tube and corpus luteum. Two serum samples, with a mean interval of 2.1 days, were taken pre-operatively and human chorionic gonadotrophin (HCG) and progesterone were determined by fluoroimmunoassay retrospectively. Cases requiring a second intervention (n = 8) were compared with uncomplicated cases (n = 52). The receiver-operator characteristic curve was used to determine optimal discriminatory levels of each diagnostic test. Among the complicated cases, 75% had progesterone values > 20 nmol/l and 100% had a daily HCG increase > 50 IU/l. The corresponding figures for the uncomplicated cases were 17% and 23% respectively. A 'high-risk' group representing 12% of the patient material was identified, and was biochemically profiled by serum progesterone values > 20 nmol/l, together with daily HCG changes > +50 IU/l/day. These patients had an 86% risk for a second intervention. Conversely, 88% of patients had only a 4% risk for such an intervention. We conclude that the combined use of two sequential serum HCG samples and one serum progesterone sample, the last of each taken not more than 24 h pre-operatively, can predict the risk for a complicated course following local injection of 15-methyl-PGF2 alpha for small tubal pregnancies.