Abstract
Background/Aim: The survival benefit of negative resection margins in patients who undergo gastrectomy with positive peritoneal lavage cytology (CY1) is unknown. Patients and Methods: We reviewed the medical records of 128 patients with CY1 but no other distant metastases who had undergone R1 gastrectomy, 21 of whom had positive margins. We compared overall survival (OS) according to margin status. Results: The positive-margin group had poorer performance status scores (p=0.02), higher number of patients had undergone limited lymphadenectomy (p=0.01), had type 4 tumors (p=0.01), and undifferentiated type (p=0.02). Median OS was 19.0 and 16.9 months in the groups with negative and positive margins, respectively (HR=1.26, 95%CI=0.75-2.12, p=0.39). An inverse probability of treatment weighted analysis showed an OS of 13.1 and 11.9 months for the groups with negative and positive margins, respectively (HR=0.83, 95%CI=0.43-1.63, p=0.59). Conclusion: The prognoses of patients with CY1 and negative or positive margins may be equivalent.
A radical gastrectomy that achieves no residual tumor is the most effective treatment for gastric cancer. Thus, adequate resection margins should be ensured when performing gastrectomy with curative intent (1). If adequate resection margins are uncertain, it is advisable to examine the margins by frozen section to ensure that R0 (no residual tumor) resection has been achieved and consider additional resection or re-excision if a cancer-positive resection margin is identified. According to the 2011 annual report of gastric cancer treatment in Japan by the Japanese Gastric Cancer Association nationwide registry, positive proximal margins (PM1) or positive distal margins (DM1) were identified in 2.43% of patients and their 5-year overall survival (OS) rate was 17.6%, whereas that of patients with negative margins was 71.5% (2). Thus, the prognoses were much worse in patients with positive margins than in those with negative margins.
In advanced gastric cancer, free cancer cells are often scattered in the abdominal cavity. Gastric cancer with positive peritoneal lavage cytology, termed CY1 in the Japanese classification of gastric carcinoma (3), is classified as Stage IV and associated with poor prognoses. In the abovementioned Japanese gastric cancer registry in 2011 (2), the 5-year OS for patients with CY1 was only 10.7%.
Surgery with curative intent was not indicated in patients with CY1 according to the treatment algorithm in the Japanese gastric cancer treatment guidelines (4). However, the one-armed CCOG0301 study that explored the efficacy of postoperative S-1 in patients with CY1 status, revealed a 5-year OS rate of 26% after radical gastrectomy (5). Since this study, the tentative standard therapy for CY1 gastric cancer has been radical gastrectomy with S-1 adjuvant therapy (1).
Because most patients with CY1 have infiltrative gastric cancer, it is sometimes difficult to determine the optimal resection line and secure negative margins when resecting the stomach. In some of these patients, additional resection is performed to obtain negative margins when the margins have been judged positive by intraoperative frozen section. However, the prognostic impact of achieving negative margins in patients with CY1 is controversial. Radical gastrectomy for CY1 gastric cancer is classified as R1 (microscopic residual tumor) resection regardless of the margin status and whether that status influences the survival of these patients has not yet been determined.
In the current study, we used propensity scores and inverse probability of treatment weighting (IPTW) to retrospectively compare the survival of patients with CY1 gastric cancer but no other distant metastases who had undergone gastrectomy with positive and negative resection margins.
Patients and Methods
Patients. In this study, the medical records of patients with CY1 gastric cancer who had undergone gastrectomy in the six participating institutions between 2000 and 2019 were retrospectively reviewed. Patients with other distant metastases such as peritoneal, hepatic, or extraregional lymph node metastases, those with R2 (macroscopic residual tumor) resection, and those with other malignancies were excluded, leaving 128 eligible patients for analysis. A flow diagram of selection of participants is shown in Figure 1. The following clinical and pathological characteristics, perioperative treatment, and postoperative outcomes were collected: age, sex, preoperative Eastern Cooperative Oncology Group Performance Status (PS) score (6), operative procedure, extent of lymph node dissection, macroscopic tumor type, histological classification, pathological tumor invasion and lymph node metastasis (pT, pN), pathological status of the distal and proximal margins, postoperative adjuvant chemotherapy, first relapse site, and prognoses. The pathological characteristics of the gastric cancers are described according to the Japanese classification of gastric carcinoma 3rd English edition (3). PM1 and DM1 mean involvement of proximal and distal margins, respectively, whereas PM0 and DM0 mean no involvement. Patients with positive margins on frozen section who underwent further resection to a negative margin were classified as having negative margins. A relapse was confirmed by computed tomography scans.
Operative procedure. Peritoneal lavage cytology was performed routinely immediately after laparotomy, before manipulation of the primary tumor. Any ascites present was aspirated. In patients without ascites, 100 ml of normal saline were instilled and aspirated from the pouch of Douglas. The presence or absence of carcinoma cells was defined as CY1 or CY0, respectively. Suspicious cases were defined as CY0. The results were available during surgery. Gastrectomy was carried out according to the Japanese gastric cancer treatment guidelines (1). None of the patients in the current series had undergone laparoscopic gastrectomy.
Statistical analysis. Continuous variables were compared using the Mann-Whitney U test. Categorical variables were compared using the χ2 test or Fisher's exact probability test. Survival is shown as Kaplan-Meier curves and was compared using the log-rank test. OS was defined as the interval from the date of surgery to the date of death from any cause. Progression-free survival (PFS) was defined as the interval from the date of surgery to the date of relapse or death from any cause. Hazard ratios (HR) were estimated using Cox regression models.
IPTW analysis using propensity scores was conducted to adjust for significant differences in patients' characteristics between the positive- and negative-margin groups. The patients' propensity scores were estimated by a logistic regression model using the following eight variables, which are based on previous univariate analyses to identify prognostic factors for OS (7, 8): age, sex, PS score, extent of lymphadenectomy (standard/limited), macroscopic tumor type {type 4 (diffuse infiltrative)/other}, histological classification (differentiated/undifferentiated), pN, and postoperative adjuvant chemotherapy (presence/absence). IPTW analysis was performed using average treatment effect on the treated: patients with negative margins were weighted by propensity score /(1−propensity score), whereas patients with positive margins were weighted by 1. The balance in covariates between groups before and after IPTW adjustment was assessed using standardized differences. Standardized difference scores of covariates were calculated using the following formulae:
All reported p-values are two-sided. p-values of <0.05 were defined as denoting statistical significance. Analyses were performed using Stata software (version 15.0; StataCorp LLC, College Station, TX, USA).
Disclosure of ethical statement. The protocol for this research project was approved by suitably-constituted Ethics Committees of each participating institution, and conformed with the provisions of the Declaration of Helsinki.
Results
Among the 128 patients with CY1 but no other distant metastases, 66 had undergone total gastrectomy, 57 distal gastrectomy, three proximal gastrectomy, and two remnant gastrectomy after distal gastrectomy. Final pathological examination of the resected specimens resulted in diagnoses of cancer-positive resection margins in 21 patients, including 11 who had undergone total gastrectomy (five PM1 DM0, five PM0 DM1, and one PM1 DM1), nine distal gastrectomy (three PM1 DM0, four PM0 DM1, and two PM1 DM1), and one proximal gastrectomy (one PM1 DM0). The remaining 107 patients had negative resection margins (PM0 DM0).
Table I shows a comparison of the clinical, surgical, and pathological characteristics of the negative- and positive-margin groups. Patients in the positive-margin group had significantly poorer preoperative PS and significantly more of them had undergone limited lymphadenectomy and had type 4 and undifferentiated tumors than those in the negative-margin group.
Adjuvant chemotherapy was applied to 79 (74%) and 13 (62%) patients in the negative- and positive-margin groups, respectively; this difference was not significant (p=0.42). The leading regimen was S-1 monotherapy in both groups, followed by S-1 plus cisplatin (Table II). In total, S-1 was taken by 61 (57%) and nine (43%) patients, in the negative- and positive-margin groups, respectively; this difference was not significant (p=0.25).
Relapse was proven in 63 (59%) and 11 patients (52%) in the negative- and positive-margin groups, respectively (Table III). The leading site of the first relapse was the peritoneum in both groups, having been detected in 43 (40%) and 10 patients (48%) in the negative- and positive-margin groups, respectively; this difference was not significant. The first relapse was local in only one patient in the current series; this patient was in the negative-margin group.
At the time of analysis, 100 patients had died, and the median follow-up period of the 28 surviving patients was 57 months. In the negative-margin group, 83 patients (78%) had died, the causes of death comprising gastric cancer in 52 patients (49%) and surgical complication in two (one anastomotic leakage and one pancreatic fistula) (Table IV). In the positive-margin group, 17 patients (81%) had died, the causes of death comprising gastric cancer in 11 patients (52%) and anastomotic leakage in one. There were no significant differences in cause of death between the two groups.
The OS curves of the patients in the negative- and positive-margin groups are shown in Figure 2. The median OS time was 19.0 [95% confidence interval (CI)=14.4-23.2] and 16.9 (95%CI=6.2-25.2) months, respectively, and the 5-year OS rates were 21.1% (95%CI=13.5-29.9) and 11.5% (95%CI=2.0-30.4), respectively, for the negative- and positive-margin groups; these differences were not significant (HR=1.26, 95%CI=0.75-2.12, p=0.39). The PFS curves of the patients in the negative- and positive-margin groups are shown in Figure 3. The median PFS time was 12.9 (95%CI=9.0-15.4) and 16.9 (95%CI=3.2-20.4) months for the negative- and positive-margin groups, respectively; these differences were not significant (HR=1.05, 95%CI=0.62-1.76, p=0.86).
After IPTW, the absolute standardized differences (ASD) of covariates of the two groups were within 0.1 except for age, sex, and extent of gastrectomy (Table I). Post-IPTW comparisons of OS between the two groups are shown in Figure 4. The median OS time was 13.1 and 11.9 months, respectively, and 5-year OS rates were 9.1% and 11.4%, respectively, for the negative- and positive-margin groups; these differences were not significant (HR=0.83, 95%CI=0.43-1.63, p=0.59). Post-IPTW comparisons of PFS between the two groups are shown in Figure 5. The median PFS was 6.7 and 9.4 months for the negative- and positive-margin groups, respectively; these differences were not significant (HR=0.82, 95%CI=0.42-1.60, p=0.56).
Discussion
The current retrospective study identified no significant prognostic differences between negative and positive resection margins in patients with CY1 gastric cancer although patients with positive margins had poorer PS and more of them had undergone limited lymphadenectomy and had type 4 disease and undifferentiated tumors than those with negative margins. The causes of death and the patterns of the first relapse were similar in the two groups. Even after IPTW, there was no significant survival difference.
Some previous reports have stated that a positive resection margin is an independent risk factor for OS and disease-free survival (9-11). Solaini et al. (12) have reported that, even for conversion surgery for Stage IV gastric cancer, a positive resection margin was the only risk factor for recurrence.
However, local relapses reportedly occur in only 16% of patients with positive margins (13). In our series, no first relapses were local in the positive-margin group. Given the association of a positive margin with other adverse pathologic characteristics, there is often no clinically significant association between positive margins and local relapses because most of these patients develop distant metastases. Postlewait et al. (14) have reported that a positive resection margin is not independently associated with recurrence or survival. Further resection or reoperation for a positive margin often has no significant prognostic impact (15). Interpretation and management of positive margins should be patient-specific and take the clinical scenario and stage of disease into consideration (15). Kim et al. (16) have found that conversion from a positive to a negative margin by re-excision of a positive frozen section margin compared with no re-excision of such a positive margin was associated with improved survival only in patients with fewer than five positive lymph nodes. Bickenbach et al. (13) have reported that margin status is an independent predictor of survival in patients with three or fewer positive nodes or T1-2 disease, but not in patients with more than three positive nodes or T3-4 disease. Similarly, Chen et al. (17) have found a better median survival in patients with N0-N2 disease undergoing re-excision; however, no such relationship was observed in N3 disease. Our CY1 cohort included pN2-3 (three or more positive nodes) in 85%, pN3 (seven or more positive nodes) in 68%, and pT3-4 in 98%. Tu et al. (18) have reported finding no statistically significant difference in 5-year OS between M1 patients with negative and positive margins. Thus, obtaining a negative margin is not significantly associated with OS in the context of other adverse pathologic factors (19).
In our series, the positive-margin group included patients with positive proximal margins in the stomach or esophagus and positive distal margins in the duodenum. Concerning proximal margins, attempts to extend an operation from distal to total gastrectomy or from total gastrectomy to gastroesophagectomy may be associated with increased complications, worse nutritional status, inferior quality of life, and increased mortality. According to the annual report of gastric cancer treatment in Japan (2), the 5-year OS rates after distal gastrectomy, total gastrectomy, and gastroesophagectomy were 76.2%, 55.0%, and 38.5%, respectively. There are few data on the management of distal margins and extending the resection distally to perform a pancreaticoduodenectomy (PD) remains controversial because of the high incidences of complications. PD was performed on only 43 of 23,106 patients (0.19%) who underwent resection in Japan in 2011 (2). No patient in our series had undergone PD.
Because patients with positive margins seem to be in worse condition and have more aggressive disease with poor PS, undergo limited lymphadenectomy, and have type 4 and undifferentiated disease compared to those with negative margins, we performed IPTW analysis using propensity scores to compare patients with negative and positive margins. IPTW analysis is being increasingly used to adjust for confounding factors between groups without reducing sample size (20). In the current series, only 21 patients had positive margins. We therefore opted for IPTW analysis rather than propensity score matching. The patients' propensity scores were estimated on the basis of previous univariate analyses, which showed that PS (1-3 vs. 0), pN (3 vs. 0-2), extent of lymphadenectomy (<D2 vs. ≥D2), histological classification (undifferentiated vs. differentiated), and absence of postoperative adjuvant chemotherapy were significantly correlated with worse OS in patients with CY1 disease (8). To confirm the balance of the distribution of each covariate, we evaluated the ASD. While the p-value depends on the sample size, the ASD does not. When an ASD is less than 0.1, it is judged to be well balanced. In our study, ASDs after IPTW exceeded 0.1 for age, sex, and extent of gastrectomy; however, these are not significant prognostic factors according to our previous univariate analyses (7, 8).
The present study has several potential limitations. First, there were few participants with positive margins despite collecting patients from six institutions. Second, the negative-margin group included cases with re-excision of the positive margins proven by intraoperative frozen section analyses. The survival rates of the initially negative margin cohort and the positive converted to negative margin cohort by re-excision may differ (19). Third, tumor size was not taken into consideration. The positive-margin group may have included larger tumors than the negative-margin group and this may have influenced survival. Finally, this study was limited by its retrospective nature, despite using IPTW. A randomized controlled trial would be preferable to confirm the current results, but would be difficult because of the scarcity of cases.
In conclusion, this retrospective analysis did not show a survival benefit for ensuring negative resection margins in patients with cytology-positive gastric cancer. Given the potential increased morbidity and decreased quality of life with greater extent of resection, it may be preferable to desist from attempting to obtain negative microscopic margins in patients with cytology-positive gastric cancer.
Acknowledgements
This study was funded by the Kawasaki Medical School and the following six institutions participated in it: Higashiosaka City Medical Center, Osaka, Japan; Kawasaki Medical School Hospital, Okayama, Japan; National Hospital Organization Osaka National Hospital, Osaka, Japan; Osaka General Medical Center, Osaka, Japan; Yao Municipal Hospital, Osaka, Japan; and Kawasaki Medical School General Medical Center, Okayama, Japan.
Footnotes
Authors' Contributions
SE and KN wrote the protocol. SE, TY, KN, KF, MI, and JK gathered the data. SE analyzed the data. YF gave statistical advises. All Authors were involved in the drafting, review, and approval of the manuscript and the decision to submit for publication.
Conflicts of Interest
KN received lecture fees from Bristol-Myers Squibb, Chugai, EA Pharma, Eli Lilly, Ono, Taiho, and Yakult, outside the submitted work. The other Authors declare no financial conflicts of interest to disclose.
- Received July 7, 2020.
- Revision received July 20, 2020.
- Accepted July 21, 2020.
- Copyright© 2020, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved