Gastric adenocarcinoma (GAC) is the second leading cause of cancer-related deaths worldwide, with more than 22,000 new cases and more than 10,000 deaths in the United States in 2014.1,2 For patients who present with resectable disease, surgery remains the backbone of curative treatment.3,4 However, relapse and disease-specific death are relatively common after gastrectomy alone, making adjunctive therapies an important component of treatment.5,6 The negative impact of postoperative complications (POCs) on survival after cancer surgery has been investigated across many solid tumor types, including head and neck, colorectal, and esophageal malignancies.7–11 In complex operations such as gastrectomy for gastric cancer, morbidity rates remain relatively high, ranging from 35 to 46 % in the Western literature.12–14 Importantly, recent studies in other solid tumors have shown that POCs decrease the ability of patients to receive adjuvant therapy.15,16 The effect of complications on the receipt of adjuvant therapy and the correlation with survival has not been explored in gastric cancer. The present analysis quantifies the association between POCs and long-term survival after gastrectomy for GAC from the U.S. Gastric Cancer Collaborative and explores the combined effect of complications and adjuvant therapy on survival.

Methods

The U.S. Gastric Cancer Collaborative includes 7 academic institutions: Emory University, Johns Hopkins Hospital, Ohio State University, Stanford University, Wake Forest University, Washington University in St. Louis, and the University of Wisconsin. All patients who underwent resection of GAC via an abdominal approach between 2000 and 2012 were included (n = 956). Data on demographic, preoperative, intraoperative, pathologic factors, as well as postoperative therapy and outcomes, recurrence, and survival were collected retrospectively by trained research fellows. Pathology staging was assigned per American Joint Committee on Cancer (AJCC) 7th edition.17 POCs were graded objectively using the modified Clavien–Dindo classification of surgical complications.18 We reviewed all data from the same hospital admission for evidence of postoperative complications, as well as all readmission events related to a surgical complication within 30 days of discharge. Grade V complications resulting in death were not included. Institutional Review Board approval was obtained at each participating institution.

Exclusion criteria included noncurative resection, death within 30 days postoperatively, or incomplete follow-up data. The primary endpoints were overall survival (OS) and recurrence-free survival (RFS). Categorical variables were compared using the Fisher exact test or χ 2 test as appropriate, and continuous variables were compared using the 2-tailed t test. OS was measured from the time of resection to death or last follow-up. RFS was measured from time of resection to recurrence, death, or last follow-up. Survival probabilities were estimated using the method of Kaplan and Meier and compared using the log-rank test.19 Prognostic factors for survival were evaluated using multivariate Cox proportional hazards regression.20 We hypothesized that receipt of adjuvant therapy was related to POCs in a way that could significantly reduce survival. To test this hypothesis, we analyzed for interaction effects between POCs and adjuvant therapy by testing an interaction term between complications and adjuvant therapy was analyzed within the context of the multivariate Cox regression model. A p value <0.05 was considered significant. All statistical analyses were performed using STATA version 13.1 (StataCorp LP, College Station, TX).

Results

Patients and Operations

We analyzed 824 patients who underwent curative surgery for GAC, did not experience 30-day mortality, and had complete follow-up data. Clinicopathologic and operative characteristics are presented in Table 1. The cohort was 44 % female, with a mean age of 65 years (standard deviation 13; range 23–94). A total of 502 patients (61 %) had at least 1 comorbid medical condition (including heart disease, hypertension, pulmonary disease, and peripheral vascular disease). Neoadjuvant chemotherapy was given to 172 patients (21 %). The majority of patients underwent either subtotal (n = 329, 40 %) or total gastrectomy (n = 346, 42 %), with the remainder undergoing distal gastrectomy (n = 137, 17 %), in which <50 % of the stomach was removed, or wedge resection (n = 8, 1 %). Most patients received a D2 lymphadenectomy (n = 487, 60 %), and 85 % of resections (n = 701) were done using an open approach.

Table 1 Clinicopathologic variables associated with complications in 824 patients after gastrectomy for gastric cancer
Full size table

Postoperative Morbidity

POCs occurred in 336 patients (41 %), and a total of 699 events were recorded (Table 2). The most frequent complications were postoperative anemia/bleeding, wound infection, and pneumonia. When stratified according to severity, 47 patients (13 %) had a grade I complication, 172 patients (49 %) had grade II, 74 patients (21 %) had grade III, and 56 patients (16 %) had grade IV.

Table 2 Details of all complications after gastrectomy for gastric cancer in 824 patients by system
Full size table

Clinicopathologic and operative variables associated with POCs are shown in Table 1. Several preoperative (older age, higher ASA class, prior gastrectomy, preoperative GI bleeding), operative (more extensive resection type, concurrent hepatectomy, reconstruction type, jejunostomy tube use, peritoneal drain use, longer operative time, higher estimated blood loss, and need for perioperative transfusion), and tumor-related variables (presence of lymphovascular invasion, perineural invasion, and higher AJCC disease stage) differed significantly between patients who experienced complications and those who did not. Receipt of neoadjuvant chemotherapy rates did not differ significantly between the 2 groups (22 vs. 19 %, p = 0.38). On multivariate logistic regression, age ≥60 years (OR 1.5, 95 % CI 1.1–2.1, p = 0.02), ASA class ≥3 (OR 1.2, 95 % CI 1.01–2.0, p = 0.04), and perioperative transfusion (OR 4.0, 95 % CI 2.7–5.8, p < 0.001) remained independent predictors of POCs. Notably, rate of receipt of adjuvant chemotherapy was significantly lower among patients who experienced complications (47 % compared with 61 %, p < 0.001).

Follow-Up and Univariate Characterization of Prognostic Factors for OS and RFS

The median follow-up period after primary resection was 35 months. The median OS was 23 months in patients who had postoperative complications compared with 45 months in those who did not, while the 5-year OS was 27 % in patients who had postoperative complications compared with 43 % in those who did not (p < 0.0001, Fig. 1a). Among patients who experienced complications, there were no significant differences in survival by complication grade (p = 0.38).

Fig. 1
figure 1

a Kaplan–Meier estimates of OS for 824 patients who did and did not have postoperative complications. b Kaplan–Meier estimates of RFS for 824 patients who did and did not have postoperative complications. Note: Failure events are reported in parentheses in between at-risk time points shown

Full size image

During the study period, 264 patients (32 %) developed recurrence after gastrectomy (114 patients with complications and 148 without). Of these, 79 patients (40 %) recurred with peritoneal disease only, 36 patients (18 %) recurred in the liver only, 21 patients (11 %) recurred with simultaneous involvement of the liver and peritoneum, and 17 patients (9 %) recurred with metastases in the lung. The median and 5-year RFS was 15 months and 23 %, respectively, in patients who had postoperative complications compared with 34 months and 40 %, respectively, in those who did not (p < 0.0001, Fig. 1b). As with OS, RFS did not correlate with complication grade. Univariate analyses of variables associated with OS and RFS are presented in Table 3.

Table 3 Univariate analysis of variables associated with OS and RFS in 824 patients after gastrectomy for gastric cancer
Full size table

Relationship Between Complications and Adjuvant Therapy

On initial multivariate Cox regression of OS and RFS, postoperative complications, age, Asian race, AJCC stage, total gastrectomy, perioperative transfusion, and neoadjuvant chemotherapy emerged as independent predictors of both OS and RFS (Table 4). The presence of postoperative complications had an independently negative impact on both OS (HR 1.3, 95 % CI 1.1–1.6, p = 0.03) and RFS (HR 1.3, 95 % CI 1.01–1.6, p = 0.03).

Table 4 Multivariate Cox regression of factors associated with OS and RFS in 824 patients after gastrectomy for gastric cancer
Full size table

We hypothesized a priori that experiencing postoperative complications might affect the receipt of adjuvant therapy in a way that significantly influences survival. Univariate analysis confirmed a statistically and clinically significant disparity in the rates of adjuvant therapy between patients with complications and those without (Table 2). Furthermore, on multivariate analysis of factors predictive of receipt of adjuvant therapy in this cohort, POCs remained a significant negative predictor (OR 0.5, 95 % CI 0.3–0.7, p < 0.001), when adjusting for age, AJCC disease stage, lymph node status, and perioperative transfusion. When stratifying receipt of adjuvant therapy by Clavien complication grade, there was a slight trend toward higher rates of adjuvant therapy use in patients with less severe complications, although this trend did not reach statistical significance (p = 0.17).

To quantify this effect, we explored the interaction between complications and adjuvant therapy, while controlling for the same confounding variables (Table 5). We found that the interaction was highly significant, as the presence or absence of adjuvant therapy acted as an effect modifier of complications on survival. The combination of experiencing a postoperative complication and not subsequently receiving adjuvant therapy significantly increased the hazard of death by 130 % and the hazard of recurrence by 70 % compared with patients who had neither postoperative complications nor adjuvant therapy (HR 2.3, 95 % CI 1.6–3.2 for OS; HR 1.7, 95 % CI 1.2–2.3 for RFS), even when controlling for disease stage. However, patients who received adjuvant therapy and experienced a complication did not have significantly decreased OS or RFS. Among patients who received adjuvant therapy (using no complications but adjuvant therapy as the reference cohort), the effect of complications becomes nonsignificant (HR 1.1, 95 % CI 0.8–1.5, p = 0.41 for OS; HR 1.2, 95 % CI 0.9–1.5, p = 0.32 for RFS). Additionally, when analyzing only patients with AJCC stage I disease, in which adjuvant therapy is expected to have less impact on survival, the interaction term between complications and lack of adjuvant therapy becomes nonsignificant (HR 1.3, 95 % CI 0.7–2.5, p = 0.44).

Table 5 Multivariate Cox regression of factors associated with OS and RFS taking into account the interaction between complications and adjuvant therapy in 824 patients after gastrectomy for gastric cancer
Full size table

Discussion

Across a large, modern North American cohort, postoperative morbidity had an independently negative impact on long-term oncologic outcomes after gastrectomy for gastric adenocarcinoma, similar to previous findings in several other solid tumor types.2,7,8,21 Importantly, we observed that complications and failure to receive adjuvant therapy interact significantly on their effect on long-term survival. Patients who experienced a postoperative complication were 50 % less likely to receive adjuvant therapy, and the combination of a postoperative complication and not receiving adjuvant therapy increased the risk of death more than 200 % compared with patients who had no complications and no adjuvant therapy.

Gastrectomy for GAC is a complex operation, and complication rates remain relatively high. Our overall complication rate of 41 % is consistent with other Western series, while Asian centers have reported rates of 10–24 %.12–14,22–24 The inconsistency of defining and grading complications and differences between gastric cancer populations in Asia and the West make comparisons of morbidity rates difficult. In the present study, we found that risk factors including patient age, ASA class ≥3, and use of perioperative transfusion were independently associated with higher rates of postoperative morbidity. Similar risk factors have been previously described in surgery for gastric cancer.13,25 Our data did not indicate that the use of preoperative chemotherapy led to increased incidence of perioperative complications, which supports previous results that neoadjuvant therapy is safe for locally advanced GAC.26,27 Notably in our multivariate regression models of survival, receipt of neoadjuvant therapy was consistently associated with lower long-term survival, even when adjusting for pathologically poor factors such as tumor stage. A similar phenomenon has been previously reported in both retrospective gastric cancer data and metastatic colorectal cancer data. Therefore, we suspect that the worse outcomes observed in this subset of patients in our dataset represent an artifact of retrospective analysis, in which we are not fully able to adjust for the adverse pathologic features associated with being selected for preoperative chemotherapy despite adjusting for a robust number of clinicopathologic features. Despite this artifact, our data shows that with complex procedures such as gastrectomy for gastric cancer in which postoperative complications can significantly reduce the receipt of adjuvant therapy and long-term survival, the importance of neoadjuvant chemotherapy cannot be overlooked.

The evidence supporting the negative impact of postoperative morbidity on long-term prognosis after solid tumor resection is mounting. Because we did not include patients with perioperative mortalities, the difference in long-term survival between patients who had complications and those who did not reflects a long-term effect by which postoperative morbidity may decrease long-term survival. Ito et al. studied more than 1000 patients undergoing resection of colorectal liver metastases at Memorial Sloan Kettering Cancer Center and found that patients who experienced complications had a significantly decreased disease-specific survival and disease-free survival at 5 years.21 Similar findings have also been reported in esophageal cancer, colorectal cancer, and head and neck cancers.7–9,11 Kubota et al. recently reported that in nearly 1400 patients from Japan, complications after oncologic gastrectomy resulted in a significant hazard ratio of 1.9 for both overall survival and disease-specific survival.28

Despite numerous reports across cancer types linking postoperative morbidity to decreased survival, the causative mechanism remains to be elucidated. The hypothesis most often discussed is that a prolonged inflammatory state present in patients who experience complications leads to greater degree of host immunosuppression, which may allow residual micrometastatic tumor cells to proliferate, causing disease recurrence.29–31 This hypothesis is strongly supported by the related immunosuppressive effect of perioperative blood transfusions, and the correlation of blood transfusion with earlier cancer recurrence, an effect that is also significant in our own dataset.32,33 Indeed, Kubota et al. showed definitively that gastric cancer patients who experienced a postoperative complication had an increased and prolonged inflammatory state as measured by daily changes in white blood cell count, peak body temperature, and C-reactive protein.28 Though these associations are both strong and suggestive, the causal mechanism between complications, inflammation, and disease progression deserves further elucidation.

This study is the first to investigate effect of decreased receipt of adjuvant therapy in patients with postoperative complications and outcomes of long-term survival. Two recent analyses using ACS-NSQIP data showed that in both pancreatic cancer and stage III colon cancer, the occurrence of postoperative complications was associated with decreased use of or delayed initiation of adjuvant therapy.15,16 We found a similar effect in our dataset. Given this association, we explored the interaction of complications and failure to receive adjuvant therapy on long-term oncologic outcomes. We observed a significant interaction effect between complications and failure to receive adjuvant therapy that strongly decreased overall survival and recurrence-free survival. These results help quantify a logical clinical conclusion that patients who experience a complication may have worse long-term outcomes in part from a failure to receive adjuvant therapy.

Interestingly, Kubota and colleagues discussed their own exploration of adjuvant therapy as a potential mechanism connecting complications and survival. In their single-institution dataset of nearly 1400 patients undergoing resection for GAC, there were no significant differences in receipt of adjuvant therapy between patients who did and did not have complications.28 Therefore, they conclude that adjuvant therapy is not likely to explain the survival difference observed. A few key differences should be noted between these 2 studies. In Kubota et al., the overall complication rate was 15 %, which was significantly lower than our own rate of 41 %.28 Nearly 50 % of the patients in their analysis underwent fully laparoscopic resections, compared with the 7 % in our cohort. Lastly, nearly two-thirds of the patients in their analysis had stage I disease, while the majority of patients in our analysis had stage II or III disease and would have been more likely to be eligible for adjuvant therapy.28 Therefore, both the higher rate of postoperative complications and the greater proportion of patients eligible for adjuvant therapy in our dataset make the failure to receive such therapy more likely to have a significant impact on survival. The inconsistent results between these 2 well-powered studies help emphasize the fundamental differences between the Eastern and Western experiences with GAC treatment.

This study is limited as a retrospective analysis; therefore we are limited in assuming causality in the relationships we observe. However, our results are consistent with the clinical hypothesis specified a priori, and our findings are significant in that they are the first to quantify the interaction between complications and adjuvant therapy and their effect on long-term survival in gastric cancer.

In conclusion, this study demonstrates a strong negative impact of postoperative morbidity on long-term outcomes after gastrectomy for gastric cancer. A significant interaction was observed between complications and failure to receive adjuvant therapy, and this interaction had a strong negative impact on survival. Further efforts aimed at reducing complications may enhance long-term oncologic outcomes in patients with gastric cancer and should have a specific focus on minimizing morbidity to increase rates of adjuvant therapy completion in postoperative patients.