The inclusion criteria were consecutive patients with CC with curative resection who underwent adjuvant chemotherapy and were histologically diagnosed as stage II/III colon adenocarcinoma at two tertiary medical institutions (Kitakyushu Municipal Medical Center, Kitakyushu, and the Department of Surgery and Oncology at Kyushu University, Higaki, Japan) between September 2009 and March 2016. Patients with two or more cancer lesions or who had been treated with preoperative adjuvant therapy were excluded. Medical records of 1,065 patients with stage II/III CC were retrospectively reviewed and 501 patients with postoperative adjuvant chemotherapy were identified. Three patients treated with irinotecan-containing regimens were excluded. Finally, a total of 349 and 149 patients treated with PO and OX-containing regimens, respectively, were enrolled for subsequent analysis. The median observation period was 1,689 days. The informed consent requirement was waived because of the retrospective nature of the study, in which patient data was kept confidential. All procedures conformed to the ethical guidelines of the Japanese Government and the Declaration of Helsinki. This retrospective study was approved by the ethics committee of Kitakyushu Municipal Medical Center (approval no. 201801055).

Preoperative variables were institution, sex, age, tumor location, preoperative bowel obstruction and preoperative serum levels of CEA and CA19-9. The intra- and postoperative variables were surgical procedure (open or laparoscopic), blood loss volume, operating time, number of harvested lymph nodes, hospital stay length and morbidity. Histological variables were tumor differentiation, T and N stage and lymphatic or vessel invasion. Tumor location was defined as follows: Right colon, including cecum, ascending and transverse colon; and left colon, including descending and sigmoid colon and rectosigmoid junction. The pathological tumor stage was defined according to the Union for International Cancer Control (UICC) TNM classification (19). Tumor differentiation was defined as well-, moderately and poorly differentiated (20). Postoperative complications were evaluated using the Clavien-Dindo classification (21) as follows: Grade I, deviation from the normal course without need for treatment; grade II, complication requiring pharmacological treatment; grade III, complication requiring surgical, endoscopic or radiological intervention; grade IV, life-threatening complication requiring intensive care unit management; and grade V, patient mortality. In the present study, complications greater than grade II were defined as postoperative complications. Postoperative mortality indicated 30-day or in-hospital deaths.

Postoperative adjuvant chemotherapy, regimens and dosages were determined by individual attending physicians, referring to the multi-disciplinary cancer board. Patients who received adjuvant chemotherapy for stage II cancer without high risk factors for recurrence (T4, lymphatic or vascular invasion, preoperative bowel obstruction and high histological grade) were excluded from the present study. All high-risk stage II and stage III patients were considered suitable for adjuvant chemotherapy, except for those intolerant to chemotherapy and those who did not consent to chemotherapy. In most cases, postoperative adjuvant chemotherapy was administered to patients with stage III colorectal cancer (CRC), but it was not recommended for all patients with stage II CRC. OX-containing regimens were recommended primarily for patients with ≥4 lymph node metastases (N2), tumor-invaded adjacent organs or perforated visceral peritoneum (T4), but OX could be initiated at other stages following a decision by either the patient or the treating physician referring to the multi-disciplinary cancer board.

The association between two variables was analyzed using chi-squared or Fisher's exact test as appropriate. For the survival analysis, recurrence-free survival (RFS) was adopted as the primary endpoint. Recurrence data included the presence and date of recurrence and the last follow-up date in patients without recurrence. The last follow-up date was December 2020. The survival analysis was performed using Kaplan-Meier analysis and the curves were compared using the log-rank test. Continuous variables were expressed as the median and the range was assessed using Mann-Whitney U test. The cutoff of the tumor marker was determined according to the upper normal limit at Kitakyushu Municipal Medical Center and Kyushu University Hospital. Multivariate survival analysis was evaluated by Cox's proportional hazard model. In order to minimize the impact of selection bias and other confounding factors, propensity score-matched analysis was performed as previously described (22). Propensity scores were generated using clinicopathological characteristics including institution, age, sex, tumor location and differentiation, pathological T and N stage and lymphatic or vessel invasion. Propensity scores were matched using a caliper width of 0.05 multiplied by the standard deviation of values calculated by a logistic regression model. Each patient with an OX-containing regimen was matched to a patient with a PO regimen using a one-to-one nearest neighbor-matching algorithm without replacement. Following matching, there were 105 patients in each group. All statistical analyses were performed using JMP 15 software (SAS Institute, Inc.). P<0.05 was considered to indicate a statistically significant difference.

Fig. 1 shows the distribution of patients. Among all patients who underwent curative resection and were diagnosed with colon adenocarcinoma, 85 patients who either had ≥2 cancer occurrences or underwent preoperative therapy were excluded. A total of 564 patients received no postoperative adjuvant chemotherapy, 149 patients were treated with OX-containing regimens and 349 patients were treated with PO fluorinated pyrimidine agent. In addition, we also excluded three patients who received irinotecan-containing regimens (folinic acid + FU + irinotecan and irinotecan + S-1).

Table I details the chemotherapy regimens of the PO and OX groups. Patients in the PO group were administered sole UFT and 5'-deoxy-5-fluorouridine regimens at a relatively early stage.

A number of patients without adjuvant chemotherapy were T3N0 patients without other high-risk factors for recurrence. According to UICC TNM stage, among the stage II patients, almost all patients who underwent adjuvant chemotherapy were high-risk stage II (high histological grade, T4 and lymphatic or vessel invasion). There were 20 patients in the OX group and 103 patients in the PO group, which indicates that a large number of regimens for patients with stage II CRC in the adjuvant setting were PO regimens without OX. Among the stage III patients, there were 129 patients in the OX group and 246 patients in the PO group. Two patients without adjuvant chemotherapy died postoperatively; one patient with perforating peritonitis and sepsis due to CC died of multiple organ failure 2 days after emergency surgery; another patient with arteriosclerosis obliterans and poorly controlled diabetes died of multiple organ failure due to anastomotic leakage 4 months after elective surgery. No chemotherapy-associated mortality was identified in the PO and OX groups.

Table II details the characteristics of patients in the OX and PO groups. Compared with the PO group, the age of the OX group patients was younger (P>0.001), poorly differentiated adenocarcinoma was more frequently observed (P=0.023) and T (P=0.033), N (P>0.001), and UICC TNM stage (P>0.0001) were advanced. Vessel or lymphatic invasion was also observed more frequently (P=0.019). The proportion of patients in the OX group differed between the two institutions (P=0.007), reflecting the policy of each institution regarding the use of OX-containing regimens. Tumor location, tumor markers, preoperative bowel obstruction, surgical procedure, blood loss volume, operating time, number of harvested lymph nodes, postoperative hospital stay and morbidity were not different between the two groups. These results suggested physician selection bias as a large proportion of N1 patients were treated with PO chemotherapy, while patients with more aggressive N2 almost exclusively received OX-containing regimens.

In the PO group, 66 of 349 (18.9%) patients had a recurrence of CC, with initial relapse in the liver (n=28) or lung (n=19), tumor dissemination (n=12) and lymph node (n=15), local (n=3) or bone recurrence (n=1) including duplication (data not shown). However, in the OX group, 40 of 149 (26.8%) patients had CC recurrence, with initial relapse in the liver (n=12) or lung (n=13), tumor dissemination (n=15) and lymph node (n=10), local (n=5) or bone recurrence (n=1; data not shown). Patients in the OX group experienced tumor dissemination and local recurrence more frequently than those in the PO group. The 5-year RFS rate was 80.8% in the PO group and 71.9% in the OX group (stage IIA, 85.1 and 83.3; IIB, 58.3 and 50.0; IIC, 66.7 and 71.4; IIIA, 93.3 and 100.0; IIIB, 80.5 and 71.3 and IIIC, 45.7 and 61.9%, respectively; data not shown).

Previous studies have shown that OX administration tends to improve RFS (6-8). In order to minimize the impact of any selection bias in comparing the recurrence risk between the two groups and to elucidate the effects of OX for patients with CC, propensity score matching was performed. Patient characteristics following matching are presented in Table III. Before matching, the prognosis of OX group patients was significantly poorer than that of PO group (P=0.0462; Fig. 2A), potentially due to strong therapeutic selection bias. After matching, the prognosis of the OX group was better than that of the PO group, although this was not significant (P=0.7220; Fig. 2B), which suggests an additional effect of OX for patients with stage II/III CC.

The RFS curve of the PO group is shown in Fig. 3. Univariate analysis showed that T4 (P<0.0001; Fig. 3A), N2 (P=0.0005; Fig. 3B), high CA19-9 (P<0.0001; Fig. 3C) and lymphatic or vessel invasion (P=0.0217; Fig. 3D) were significantly associated with worse RFS. Multivariate analysis indicated that high CA19-9 [hazard ratio (HR), 3.367; 95% CI, 1.773-6.393; P=0.0002], T4 (HR, 2.947; 95% CI, 1.698-5.113; P=0.0001), N2 (HR, 2.704; 95% CI, 1.304-5.609; P=0.0075) and lymphatic or vessel invasion (HR, 1.675; 95% CI, 1.014-2.765; P=0.0437) were independent predictive factors for cancer recurrence (Table IV).

In order to compare the recurrence risk between patients with PO and OX-containing regimens, subgroup analysis was performed using the propensity score-matched cohort to minimize selection bias. Subgroup analysis of patients with T4, N2, high CA19-9 and lymphatic or vessel invasion revealed an additional effect of OX in the high CA19-9 subgroup (P=0.0016; Fig. 4A). This was not demonstrated in the subgroup with lymphatic or vessel invasion (P=0.9722; Fig. 4B) but there was an association with a prognostic benefit of OX in the T4 and N2 subgroups (P=0.6084 and P=0.2184, respectively; Fig. 4C and D), which suggests that patients with T4, N2 and high CA19-9 may benefit from OX-containing regimens such as FOLFOX and CapOX.