The aim of this retrospective investigation was to assess the pattern of failures of 412 patients with recurrent ovarian cancer followed up with different surveillance protocols.Time to recurrence was less than 6 months in 98 women (23.8%), 6 to 12 months in 102 women (24.7%), and more than 12 months in 212 women (51.5%). Symptoms at relapse were referred by 81 women (19.7%). Among the 331 asymptomatic patients, the surveillance procedure that raised the suspect of recurrent disease was clinical examination in 49 (14.8%), imaging technique in 90 (27.2%), serum CA 125 in 77 (23.3%), and both serum CA 125 and imaging technique in 115 (34.7%). At univariate analysis, survival from initial diagnosis was related to stage (P = 0.004), residual disease after initial surgery (P < 0.0001), time to recurrence (P < 0.0001), site of relapse (P = 0.04), and treatment at recurrence (P < 0.0001), and survival after recurrence was related to stage (P = 0.01), residual disease (P < 0.0001), time to recurrence (P < 0.0001), and treatment at recurrence (P < 0.0001). Conversely, symptoms at recurrence had no prognostic relevance. Cox proportional hazards model showed that residual disease and time to recurrence were the only independent prognostic variables for both survival from initial diagnosis (P < 0.0001) and survival after recurrence (P < 0.0001). In conclusion, there was no survival difference between asymptomatic and symptomatic patients at the time of relapse, and therefore, the diagnostic anticipation allowed by a scheduled follow-up protocol did not seem to improve the clinical outcome of patients who ultimately developed recurrent disease.