The objective of the present study was to improve the definitions of optimal and suboptimal surgery in ovarian cancer. A retrospective prognostic factor analysis was done in a series of 433 patients with advanced ovarian cancer (stages III and IV) treated during the cisplatine era from 1980 to 1997 with assessment of postoperative residual disease by number and ranges of size of peritoneal nodules. Mean age of patients was 56.7 years. Median follow-up of patients alive was 138 months (range: 6-266 months). Median overall survival was 23 months. Significant prognostic factors for overall survival in univariate analysis were age, the presence of ascites, FIGO stage, treatment period, histological subtype, grade, results of surgery as defined by size and number of residual nodules. In multivariate analysis, quality of surgery defined by no versus few versus numerous residual nodules showed to remain an independent prognostic factor for outcome (P = 0.003), whereas size of residual nodules did not achieve significance. In conclusion, only complete surgery (no residual nodules) confers a real survival advantage. Cytoreduction to few and small nodules is associated with favorable outcome and could be qualified as optimal. Otherwise, cytoreduction leaving numerous nodules, whatever their size, remains suboptimal. Such patients should be considered for neo-adjuvant chemotherapy.