Surgical staging and treatment of early ovarian cancer: long-term analysis from a randomized trial

Abstract

A long-term follow-up analysis of the randomized clinical trial Adjuvant Chemotherapy in Ovarian Neoplasm (ACTION) from the European Organization for Research and Treatment of Cancer was undertaken to determine whether the original results with a median follow-up of 5.5 years could be verified after longer follow-up with more events. In the ACTION trial, 448 patients with early ovarian cancer were randomly assigned, after surgery, to adjuvant chemotherapy or to observation (no further treatment). The original analysis found that adjuvant chemotherapy improved recurrence-free survival but not overall survival and found in a subgroup analysis that completeness of surgical staging was an independent prognostic factor, with better recurrence-free and overall survival among those with complete (optimal) surgical staging. After a median follow-up of 10.1 years, we analyzed the more mature data from the ACTION trial and found support for most of the main conclusions of the original analysis, except that overall survival after optimal surgical staging was improved, even among patients who received adjuvant chemotherapy (hazard ratio of death = 1.89, 95% confidence interval = 0.99 to 3.60; overall two-sided log-rank test P = .05). More cancer-specific deaths were observed among nonoptimally staged patients (40 [27%] of the 147 deaths in the observation arm and 11 [14%] of the 76 deaths in the adjuvant chemotherapy arm) than among optimally staged patients (seven [9%] of the 75 deaths in the observation arm and 11 [14%] of the 76 deaths in the adjuvant chemotherapy arm) (two-sided chi(2) test for heterogeneity, P = .06). Thus, completeness of surgical staging in patients with early ovarian cancer was found to be statistically significantly associated with better outcomes, and the benefit from adjuvant chemotherapy appeared to be restricted to patients with nonoptimal surgical staging.

Figure 1

Kaplan–Meier curves for cancer-specific and recurrence-free survival among patients with early-stage ovarian carcinoma by staging type and treatment arm (observation and adjuvant chemotherapy). All comparisons were between the observational arm and the adjuvant chemotherapy arm. The survival percentage is shown on the y-axis, and time is shown on the x-axis. A) Cancer-specific survival in all 448 patients (hazard ratio [HR] of death = 0.73, 95% confidence interval [CI] = 0.48 to 1.13, P = .16). B) Recurrence-free survival in all 448 patients (HR of death = 0.64, 95% CI = 0.46 to 0.89, P = .007, in favor of adjuvant chemotherapy). C) Cancer-specific survival in optimally staged patients (HR of death = 1.58, 95% CI = 0.61 to 4.08, P = .34). D) Recurrence-free survival in optimally staged patients (HR of death = 0.73, 95% CI = 0.38 to 1.42, P = .35). E) Cancer-specific survival in nonoptimally staged patients (HR of death = 0.58, 95% CI = 0.35 to 0.95, P = .029, in favor of adjuvant chemotherapy). F) Recurrence-free survival in nonoptimally staged patients (HR of death = 0.60, 95% CI = 0.41 to 0.87, P = .007, in favor of adjuvant chemotherapy). G) Cancer-specific survival in patients with a poorly differentiated (grade 3) early-stage ovarian carcinoma (HR or death = 0.62, 95% CI = 0.34 to 1.12, P = .108). The two-sided log-rank test was used to determine P values. All statistical tests were two-sided. N = number of patients; O = number of events observed.

Figure 2

Forest plots of the interaction between the two staging categories (optimal and nonoptimal) vs treatment effect (adjuvant chemotherapy better vs observation better) for cancer-specific survival. Solid squares = hazard ratios (HRs) for cancer-specific survival (with the area of the square being proportional to the variance of the estimated effect); length of the horizontal line through the square = 95% confidence interval (CI); open diamond = HR (middle of the diamond); horizontal points of the diamond = 95% CI for the combined data. CSS = cancer-specific survival; EORTC = European Organization of Research and Treatment of Cancer; O − E = number of events observed minus number of events expected under the null hypothesis; SD = standard deviation; Var. = variance of 1 divided by the logarithm of the HR. Linear trends and heterogeneity of the HRs to detect differences in relative size of treatment effect were assessed by a χ² test for interaction. All statistical tests were two-sided.