We collected the clinical data of patients with stage III epithelial ovarian, tubal, and peritoneal cancer between January 2002 and December 2012 who were treated in the Institutional Review Board of Taipei Veterans General Hospital. Patients in the control cohort were treated according to the protocol described in the GOG 114 trial [1]; patients in the experimental cohort were treated according to the protocol described in the GOG 172 trial [2]. A flow chart describing the screening and matching process is shown in Fig. 1.
We used a propensity score-matching technique to compare two intraperitoneally treated cohorts of patients. The technique of propensity score, which is a covariate summary score, was used for the analysis in the present study [4, 5].
The matching procedure involved a two-step analysis. In step 1, we used 10 conditioning variables to develop propensity scores. In step 2, we used an algorithm of the nearest neighbor matching within a specified caliper distance (0.25 σ in the current study, where σ was the standard deviation of logit of propensity score) without replacement to create matched samples [6, 7]. Thus, for a given patient in the experimental group, we identified all of the patients in the control cohort whose propensity scores lay within a specified distance of that of the patient in the experimental group. From this restricted set of control patients, we matched the patient in the control group whose propensity score was closest to that of the patient in the experimental group. Eventually, this created two samples of equal size (1:1 matching).
We analyzed progression-free survival and overall survival in this study. All analyses were performed using STATA SE software, version 12 (Stata Corp., College Station, Texas, USA). P values less than 0.05 were considered statistically significant.