Improving the quality of gastric cancer surgery: factors associated with positive resection margins for gastrectomy

Positive margins after gastrectomy have been associated with poor patient prognosis. This study aimed to identify risk factors associated with margin-positive resections. The National Cancer Database was queried from 2004 to 2014 for all patients with gastric adenocarcinoma who underwent resection with curative intent and had known margin status. Univariable and multivariable logistic regression analysis was performed to identify variables associated with positive margins. A total of 32,193 patients were identified who met study inclusion criteria, of which 11.8% (3786 patients) had a margin-positive resection. Tumor size > 6 cm, T3 or T4 tumors, tumor location in the body of stomach or in multiple regions, signet ring cell histology, presence of lymphovascular invasion, positive lymph node involvement, and lack of neoadjuvant therapy were independently associated with an increased risk of positive margins. Advanced disease characteristics, aggressive tumor pathology, and absence of neoadjuvant therapy were associated with margin-positive resections.


Introduction
Gastric cancer is the fifth most commonly occurring cancer and has become the third leading cause of cancer-related deaths worldwide [1,2]. Curative-intent surgery (margin-negative resection with D2 lymph node dissection) in combination with chemotherapy/chemoradiation therapy remains the only hope for cure. As gastric cancer may spread at the level of submucosa, an additional length (5-6 cm) of grossly normal tissue is often resected to maximize the likelihood of achieving a microscopically negative resection margin [3,4]. Despite the aggressive resection, margin-positive rates of up to 20% have been reported in patients after gastrectomy [5][6][7][8]. The occurrence of a margin-positive resection (R1) has been identified as an independent predictor of poor prognosis, particularly among patients with earlystage gastric cancer [7][8][9][10][11][12][13]. Patients with positive resection margins continue to have a high rate of both local and distant recurrences even if they receive aggressive adjuvant multimodality therapy [14].
Currently, the National Comprehensive Cancer Network (NCCN) practice guidelines recommend neoadjuvant treatment for cancers of the esophagus and gastroesophageal junction, and this treatment algorithm is also being explored in other solid tumors, including rectal, pancreas, and breast cancers, to facilitate marginnegative resections- [15][16][17][18][19][20]. Initial investigational studies in patients with locally advanced gastric cancer have also demonstrated promising results for the use of neoadjuvant therapies, with the potential to reduce the incidence of R1 resections as well as improve recurrence-free (RFS) and overall survival [21][22][23]. Neoadjuvant therapy may also enhance local disease control which is the most common pattern of recurrence in gastric cancer patients [19,[24][25][26][27]. Thus, it is important to identify patients with a high risk of R1 resection and recommend them for preoperative chemotherapy or chemoradiation therapy with hopes of improving their overall outcomes.
In this context, the objective of this study was to use the National Cancer Database (NCDB) to identify risk factors associated with R1 resections in gastric cancer patients.

Patient selection
This was a retrospective, population-based study using data provided by NCDB. The NCDB is a collaborative program between the American College of Surgeons Commission on Cancer (CoC) and the American Cancer Society. The NCDB gathers data from over 1500 CoC-accredited cancer hospitals and represents 70% of newly diagnosed cancer cases in the United States [28].
The NCDB was queried between 2004 and 2014 for all patients ≥18 years of age with invasive gastric adenocarcinoma who underwent elective partial or total gastrectomy with curative intent and had available margin status data. Patients were excluded from this study if they had a diagnosis of a second malignancy during their lifetime, metastatic disease, a palliative resection, an emergent resection, or missing/unknown margin status. A resection was defined as emergent if the number of days between diagnosis and surgery was recorded as < 1 day. A palliative resection was defined in NCDB as a surgical procedure, which may include a bypass procedure, performed to alleviate symptoms but not to treat the primary tumor. Disease staging was performed using pathologically-determined tumor size, extension, and nodal involvement in accordance with the American Joint Committee on Cancer (AJCC) Staging System, 7th edition [29]. The flow diagram of the study population selection is depicted in Fig. 1. To ensure that the exclusion of patients with missing/unknown margin status does not introduce potential bias to this study, patient demographics and presenting characteristics were compared between patients with known margin status and those with missing/unknown margin status (Supplemental Table 1).

Statistical analysis
Patient demographics, disease and treatment characteristics, and post-operative and long-term outcomes were abstracted from NCDB for patients meeting study inclusion criteria. Univariable logistic regression analysis using available patient demographics and disease characteristics data was performed to identify variables associated with a margin-positive resection. Significant variables on univariable analysis were selected for inclusion in multivariable analysis. Overall survival (OS) was defined as the time from the date of surgery to the date of death or last contact as recorded in the NCDB. Survival estimates were generated using Kaplan-Meier method and compared using the log-rank test. Univariable Cox proportional hazard regression analysis was performed using available patient demographics and disease characteristics data to identify variables associated with OS. An adjusted Cox proportional hazard regression analysis was performed to evaluate margin-positive resection as an independent prognostic factor. A p-value < 0.05 was considered significant. Statistical analysis were performed using SPSS (Version 22, SPSS Inc., Chicago, IL, USA). This study was approved by the Partners Health Research Committee institutional review board.

Patient characteristics
A total of 32,193 patients were identified from NCDB who met study inclusion criteria. Patient demographics and disease characteristics are summarized in Table 1.
The annual rate of margin-positive resections remained consistent throughout the years captured in this study, ranging from 10.9% to 13.7%. The most of patients (28,407,88.2%) had a margin-negative resection, while 11.8% (3786 patients) had a margin-positive resection.

Univariable and multivariable logistic analysis for positive resection margin
On univariable analysis, patient factors associated with margin-positive resections included younger age (p = 0.002), female sex (p = 0.003), white or black race compared to Asian (p = 0.001-0.004), and a higher Charlson-Deyo comorbidity index (CDCI) (p = 0.004) ( Table 2). Histologically, patients with margin-positive resections were more likely to have tumors of larger size (p < 0.001), poorly or undifferentiated grade (p < 0.001), nonadenocarcinoma histology (p < 0.001), higher T, N, and overall stage (p < 0.001), location in the stomach body, along the greater curvature of the stomach, or in multiple regions (p < 0.001-0.003), and with the presence of lymphovascular invasion (p < 0.001). In addition, patients with margin-positive resections were less likely to have received neoadjuvant therapy (p < 0.001) and more likely to have undergone total gastrectomy or en block/multivisceral resection (both p < 0.001).
On multivariable analysis, tumor size > 6 cm, T3 or T4 tumors, tumor location in the body of the stomach or in multiple regions, signet ring cell histology, presence of lymphovascular invasion, positive lymph node involvement, and lack of neoadjuvant therapy were independently associated with an increased risk of a marginpositive resection ( Table 2).

Discussion
The presence of residual gastric cancer is a strong predictor of poor prognosis after gastrectomy and an independent risk factor for decreased survival [8,10,11]. The association with adverse tumor characteristics suggests that positive margins reflect an underlying aggressive biology of disease. Adjuvant radiation has been shown in clinical trials to provide a survival benefit after gastrectomy; however, patients with positive margins are often excluded from these studies and even aggressive multimodality adjuvant treatment may not be sufficient to overcome the negative prognosis that positive margins portend [14,30,31].
Consistent with previous reports, the present study demonstrated the significantly decreased OS in patients with a margin-positive resection compared to those with a margin-negative resection [8,10,12]. Positive margins remained an independent predictor of poor prognosis after adjusting for stage and biologic factors.
Regarding clinical features associated with positive surgical margin, younger age, female gender, larger tumor size, poorer tumor differentiation, lymphvascular invasion, neoadjuvant chemotherapy, and more advanced tumor stage were identified as risk factors, which was similar to previous studies [32][33][34]. In multivariable analyses, only larger tumor size (> 6 cm), signet ring cell histology, deeper tumor infiltration (T3-4 stage), positive lymph node metastasis, lymphovascular invasion, and without using neoadjuvant therapy remained as independent factors for positive surgical margin. The result indicated that aggressive tumor biology might be the main factors contributing to positive surgical margin after gastrectomy. Particularly, poor differentiated tumor cells, like signet ring cell or diffuse type GC cells, may spread subepithelially beyond the macroscopically detectable boundaries of the lesion [35], thus a wider macroscopically free margin cannot predict microscopically negative margins [36]. Though the use of intraoperative frozen section to determine the extent of surgical margin is applied, the rate of R0 resection remains unsatisfied. For one reason, frozen biopsy is time and resource consuming, and is not feasible to be routinely used in every case. For another, like signet ring cell histology, which is characterized by high mucin content, is difficult to be evaluated on hematoxylin-eosin (HE) staining, resulting to false negative in frozen biopsy [37]. Therefore, surgeons should have more discretion when managing those patients.
In studies of cancers of the esophagus and gastroesophageal junction, neoadjuvant therapy has been shown to reduce the likelihood of a margin-positive resection [16,38]. In the landmark MAGIC trial of patients with resectable gastroesophageal cancer who were randomized to peri-operative chemotherapy and surgery or surgery alone, more patients who received perioperative chemotherapy reached a curative resection [39]. In our study population, not using neoadjuvant therapy was identified as a variable that was associated with the risk of margin-positive resection (OR 2.01, 95% CI 1.84-2.20). As neoadjuvant chemotherapy is a modifiable treatment variable, patients who display several of the other characteristics associated with margin-positive resection can be considered for neoadjuvant treatment to improve their overall outcomes.
Limitations of the present study included the potential selection bias inherent to a retrospective review and the data constraints of using a large, nationwide database. We were not able to differentiate between proximal and distal margin status and a patient was considered to have a margin-positive resection regardless of the location of the positive margin. Detailed information about the neoadjuvant therapy was also not available. However,    despite the loss of some clinical granularity, the NCDB is a representative collection of 70% of diagnosed cancer cases in the United States. The NCDB also does not provide additional information on facility volume, which may impact patient treatment, including use of neoadjuvant therapy, and subsequent outcomes.

Conclusions
Positive margins are an independent predictor of poor survival in gastric cancer. Identification of associated characteristics can aid in the risk stratification of patients who have increased likelihood of a margin-positive resection and allow for appropriate treatment planning including neoadjuvant chemotherapy or chemoradiation therapy with the hopes of improving patients' long-term outcomes.