We analyzed factors predictive of relapse risk in patients with stage I invasive epithelial ovarian cancer: 252 patients from the Princess Margaret Hospital provided a data base for hypothesis generation, and data on 267 patients from the Norwegian Radium Hospital were used for hypothesis testing. The outcomes in most analyses in the two series were very similar, validating the following conclusions. Differentiation (grade) was the most powerful predictor of relapse, followed by dense adherence (which resulted in outcomes equivalent to those in stage II) and, finally, large-volume ascites. When the effects of these three factors were accounted for, then none of the following were prognostic: bilaterality (stage Ib), cyst rupture (stage Ic), capsular penetration (stage Ic), tumor size, histologic subtype, patient age, year of diagnosis, and postoperative therapy. These results allow simplification of stage I substaging, as only differentiation, dense adherence, and large-volume ascites (? peritoneal cytology) need be considered. The 5-year relapse-free rate was 98% in patients with grade 1 tumors in whom both dense adherence and large-volume ascites were absent. These patients are adequately treated by operation alone. Although the relapse risk was high enough in the remaining patients to warrant postoperative treatment, a significant benefit could be shown only for a small subset of patients, namely those with densely adherent tumors treated with abdominopelvic radiotherapy. In grades 2 and 3, none of the therapies used in either series was superior to pelvic radiotherapy or operation alone.