Value of Radiotherapy in Primary Gastric Cancer and Establishment of a Prognostic Nomogram Model

Background: This study aimed to compare the use of radiotherapy (RT) in gastric cancer (GC) patients from the SEER database and established a nomogram to assess cancer-specic survival (CSS). Methods: Patients from the SEER database between 2004 and 2013 were analyzed. Survival was analyzed by Kaplan-Meier curves and log-rank test. Prognostic factors in multivariate Cox analysis were screened to construct a nomogram. The performance the nomogram was validated via concordance index (C-index), calibration plots, and decision curve analyses (DCAs). Results: 9653 GC patients were analyzed totally. In the entire cohort, patients who received pre/postoperative RT had better survival than those who did not receive RT (P = 0.043 and < 0.001, respectively). Similar results were observed in lymph node-positive patients. However, no signicant survival benet was seen in lymph node-negative patients between postoperative RT group and no RT group (P = 0.057), but patients who received postoperative RT and those who did not receive RT experienced better survival than those who received preoperative RT (P < 0.001 and 0.001, respectively). Prognostic factors of GC analyzed by Cox regression model included age, race, tumor grade, tumor histology type, primary tumor site, T stage, lymph node metastasis ratio, RT status, and chemotherapy information independently (P < 0.001). The nomogram was established and showed excellent prediction performance, and its C-index of 0.725 were signicantly higher than those of nomograms based AJCC system with C-index at 0.643. In addition, the calibration plots performed good consistency between the predicted and actual survival probabilities, and the DCAs indicated better clinical net benets than the traditional AJCC system. Conclusions: RT can CSS in GC those with positive lymph nodes. The construction verication of the outcomes of

According to perioperative adjuvant therapy trials conducted in Asia and the West, adjuvant therapy methods in Asia and the West are different. In the U.S. and Europe, the treatment standard for GC patients with T2 or above or lymph node positivity stipulated is either perioperative chemotherapy based on category 1 evidence or preoperative radiotherapy (RT) and chemotherapy based on category 2b evidence in the NCCN guidelines [4,5]. In Asian countries, compared with Western countries, due to the widespread use of radical gastrectomy and D2 lymph node dissection, more and more studies have focused on investigating postoperative adjuvant chemotherapy without RT or neoadjuvant therapy with D2 lymph node dissection [6,7]. Therefore, the absolute bene t of RT is often overlooked. This article retrospectively analyzed patients with locally advanced nonmetastatic GC through the Surveillance, Epidemiology, and End Results (SEER) database, discussed the value of RT in primary GC, and predicted the patients' cancer-speci c survival (CSS) outcomes by establishing a nomogram prediction model.

Patient Selection
Data were obtained from the largest publicly available cancer database (the SEER database), which almost covered 28% of the U.S. population [1]. SEER*Stat is used to capture patient available information on the o cial network (https://seer.cancer.gov/) through an online program provided by SEER.
Data from the SEER database on GC patients, who diagnosed between 2004 and 2013, and patients who met the strict screening criteria were analyzed. The owchart of detailed selection was shown in Fig. 1.
Patients who received either postoperative RT (Postop RT group), preoperative RT (Preop RT group) or no RT (No RT group) and had AJCC stage IB to IV M0 disease were included in the analysis.

Clinicopathological Data
Patient demographics (age, sex, and race), tumor characteristics (differentiation, site of the primary tumor, clinical T and N stage, tumor histology, 6th edition AJCC stage, and lymph node involvement), treatment (radiotherapy, chemotherapy), and survival data were gained from the SEER database. A median age of 60 years was chosen as the cut-off value. Patient characteristics are demonstrated in Table 1.

Statistical Analysis
The primary endpoint was CSS for GC patients, which was de ned from the time of period of beginning date of diagnosis to ending date of tumor-related speci c death. The Kaplan-Meier method and log-rank test were applied to survival analysis, and the chi-square test was devoted to the comparison of categorical variables. A receiver operating characteristic (ROC) curve to select cut-off values was used for the lymph node metastasis ratio (LNR) and evaluate its predictive value for CSS. Cox regression models were constructed to determine prognostic factors affecting primary nonmetastatic GC. A nomogram was constructed according to the factors that showed statistical signi cance determined by Cox regression analysis.
Using concordance index (C-index), calibration plots, and decision curve analyses (DCAs) to evaluate the accuracy of the nomogram model. The C-index was indicated to the discriminability of the model and constructed to compare the performance of the nomogram model and the traditional AJCC staging system. The calibration plots were created by comparing the actual survival probabilities with the nomogram-predicted probabilities of CSS. The DCAs were conducted to compare the clinical usefulness and net bene t of the predictive model. All statistical processing was accomplished with IBM SPSS Statistics software version 22.0, MedCalc software and R software version 3.6.2. A statistically signi cant cut-off value was set as P < 0.05.

Results
In this study, a total of 64,357 patients were initially enrolled. After strict screening, about 9,653 patients met the criteria. The patients were separated into three groups according to the RT status (No RT, Preop RT and Postop RT groups). The speci c patient information was as follows: 5421 patients did not receive RT, 1002 patients received preoperative RT, and 3230 patients received postoperative RT, accounting for 56.2%, 10.4%, and 33.5%, respectively (see Table 1 for details).

Survival And Lnr Categories
The de nition of LNR was the positive nodes number divided by the total number of lymph nodes detected. The ROC curves were generated for the comparison of the discrimination for the LNR and positive lymph nodes number (no. of positive lymph nodes). As shown in Fig respectively; and the survival predictions of the two were statistically signi cant (P < 0.001). Using LNR to predict the GC patient prognosis was more accurate, objective and effective. Therefore, we grouped the selection of the LNR through the cut-off point of the ROC curve.
As shown in Fig. 2d, the best cut-off point of the LNR was 0.1714, with a sensitivity of 63.1% and a speci city of 67.6% for predicting the prognosis of GC patients. The AUC was 0.696 (95% CI: 0.687 ~ 0.705, P < 0.001). We identi ed the patients with negative lymph nodes as a group (LNR1: LNR = 0%), the rest patients with positive lymph nodes were classi ed into two groups: LNR2: 0% < LNR ≤ 17.14%; and LNR3: LNR > 17.14%.
Kaplan-Meier survival analysis was performed according to LNR grouping (Fig. 3a)

The Analyzation Of Cox Regression Model
In the primary GC cohort, the median survival time was 36 months, and the 1-, 3-, and 5-year CSS rates were 77.6%, 49.8%, and 41.3%, respectively. The hazard ratios for CSS are listed in Table 3 obtained by Cox regression model according to all variables. All variables except for sex (P = 0.280) were revealed as prognostic factors in Cox regression models (i.e., age, race, tumor differentiation, tumor histology type, primary tumor site, T stage, LNR, RT status, and chemotherapy information).  Table 3.

Establishment And Validation Of The Nomograms
The selected prognostic factors were used to construct the nomogram for 1-, 3-and 5-year CSS by using multivariate Cox regression model (Fig. 5). The nomogram included parameters such as age, race, tumor differentiation, tumor histology type, primary tumor site, T stage, LNR, RT status, chemotherapy information.
The performance of the model (i.e., discrimination, calibration and clinical usefulness) was evaluated through C-index, calibration plots, and DCAs. The C-index value of the nomogram for CSS was signi cantly greater than those of nomogram . Moreover, high-quality calibration plots demonstrated excellent consistency between the actual and nomogram-predicted survival probabilities (Fig. 6a-c). The DCAs curves revealed relatively good performance for the model according to clinical application. The threshold probabilities of the new model had excellent net bene ts for predicting the 1-, 3-and 5-year CSS compared with the AJCC system ( Fig. 6d-f). Therefore, we believe that the prediction model has highly accurate prediction ability as a whole and can be used in clinical work.

Discussion
Even adjuvant therapy has been designed to be a tool to improve GC patients' prognosis, it is still the leading cause of tumor-related death. The bene t of RT in GC patients remains controversial [8]. The earliest con rmed application of RT in GC came from the Intergroup 0116 trial. It has been shown that patients who had postoperative chemoradiation therapy had survival bene t than those who underwent surgery alone [9]. However, this conclusion was questioned for its low (10%) utilization in D2 lymphadenectomy and its inclusion of stage IB patients [9].
In our study, the Kaplan-Meier analysis of the primary cohort from the SEER database showed that both preoperative and postoperative RT increased the CSS of GC patients. Similar survival results were obtained in patients with positive lymph nodes. Additionally, Cox regression analysis of lymph nodepositive patients showed preop RT and postop RT groups had lower death risk than no RT groups. It is noted that in patients with negative local lymph nodes, a correlation between preoperative RT and poor survival outcomes was listed (Fig. 3c). Previous literature showed patients who received preoperative chemoradiation with 5-uorouracil had both overall long-term and recurrence-free survival compared to those who received perioperative chemotherapy alone in a large retrospective U.S. multi-institutional study [10]. Besides, patients with N1 disease and those who had tumors with lymphovascular invasion bene ted the most from preoperative chemoradiation in the propensity-adjusted model. However, the result indicated less bene t from preoperative chemoradiation from patients with N0 disease and those with tumors that did not have lymphovascular invasion [10]. In another study, the Adjuvant Chemoradiotherapy in Stomach Tumors (ARTIST) trial, which was aimed to make adjuvant chemotherapy comparison with chemoradiation, failed to demonstrate a difference in disease-free survival and overall survival in the whole cohort study, however, the subgroup analysis demonstrated that patients with intestinal type histology and positive lymph nodes, in particular, may bene t from chemoradiation [11,12]. Therefore, above researches presented in this article suggest that RT in GC patients may bene t those with positive local lymph nodes.
In the subgroup survival analysis based on AJCC stage, we found that with the increase of tumor stage, the effect of RT became more and more signi cant, especially preoperative RT. A rational explanation was preoperative RT could downstage the advanced tumor and increases the chance of curative resection for advanced tumors. Additionally, neoadjuvant chemoradiation has been reported to improve the R0 resection and achieve pathological complete remission (PCR) in advanced GC. A 30% PCR rate and 70% R0 resection rate using preoperative chemoradiation were observed in a multi-institutional single-arm phase I trial in the U.S, which consisted of induction chemotherapy followed by radiation plus concurrent uorouracil [13]. In additionally, a 26% PCR rate and 77% R0 resection rate were also reported in phase II multi-institutional trial in the U.S. (RTOG9904) with two cycles of induction chemotherapy followed by concurrent chemoradiotherapy [14]. Therefore, pathological complete or partial response due to neoadjuvant therapy may be associated with improved overall survival, disease-free survival and D2 dissection [14,15].
The clinical impacts of the LNR in GC patients had been previously reported [16][17][18][19]. In this study, the prognostic value of the LNR and the positive lymph node metastases number in GC patients was compared and analyzed based on the 1-, 3-, and 5-year survival. The results showed that the LNR had a better predictive value in GC patients than the positive lymph node metastases number. In addition, survival analysis showed a correlation between a high LNR and poor CSS. It was consistent with Kim et al. nding that a relation between a high LNR and poor disease-free survival with advanced GC in the ARTIST trial [20]. The multivariate analysis in our study also demonstrated that the LNR was a meaningful predictor of poor CSS. Therefore, using LNR to predict the prognosis of GC patients is accurate, effective and objective, so this study proposes the use of the LNR as a parameter for the establishment of a nomogram.
This study noted that compared with the Postop RT and No RT combined groups, the number of positive lymph nodes, removed lymph nodes and LNR in the Preop RT group were signi cantly reduced (Table 2). However, preoperative RT does not seem to have a similar effect on the primary tumor because the tumor T classi cation distribution was not consistent with that of the Postop RT and No RT combined groups. In addition, a small retrospective series from Spain demonstrated that there was a signi cantly higher probability of achieving a Becker Ia-b response, a favorable pathological response and grade D nodal regression when comparing neoadjuvant chemoradiation to neoadjuvant chemotherapy in resectable advanced GC, Nodal, rather than primary, response was associated with longer 5-year progression-free survival and overall survival. However, no signi cant survival differences were seen in patients with a baseline negative lymph node status regardless of neoadjuvant treatment in either 5-year progressionfree survival (chemotherapy 53%; chemoradiation 80%; P = 0.45) or 5-year overall survival (chemotherapy 58%; chemoradiation 51%, P = 0.92) [21]. This result suggests that preoperative RT may bene t patients with local lymph node metastasis and has little effect on the primary tumor T classi cation.
In our study, the main difference in pathological features between the Preop RT group and the Postop RT and No RT combined groups was local lymph nodes, not the T stage of the tumor. One explanation for this difference is that although the published data reported complete pathological remission, due to limitations of the SEER database itself, patients treated in some trials may not be included and re ected in the SEER database. In addition, lacking information on concurrent or sequential chemotherapy in SEER database, which also affects the remission rate. Another limitation of the SEER database itself is the lack of clinical staging data for GC. Table 1 showed that 31.5%, 27.7% and 16.3% of patients were node negative in the No RT, Preop RT and Postop RT groups, respectively. If preoperative RT was used to transform a patient from a clinically lymph node-positive status to a pathologically lymph node-negative status, then in the negative lymph node preoperative RT group, the mortality rate will be overestimated because this group will include clinical lymph node-positive patients. On the side, the survival bene t of the positive lymph node preoperative RT group may be underestimated because it may have excluded patients with complete clinical remission of locally positive lymph nodes after neoadjuvant therapy.
Several studies have revealed the predictive abilities of nomograms for predicting GC. In this study, the nomogram was used to calculate the CSS outcomes of a single GC patient using Cox regression to screen 9 independent variables. According to the nomogram, the weighting of the LNR was the largest.
The top three factors that affect CSS are the LNR, T stage and tumor grade. The nomogram presented herein showed excellent prediction performance, and its C-index value of 0.725 using signi cant prognosis factors was signi cantly higher than the C-index value of 0.643 using traditional AJCC system.
The calibration plots were close to the ideal 45º dotted line. The DCAs curves with higher cut-off probability levels indicated that the new model had better clinical application value (net bene t) compared with the AJCC system in terms of the 1-, 3-and 5-year CSS. In addition, one of the advantages of the nomogram model is to include the LNR instead of the N stage as an effective parameter. The aforementioned LNR has been shown to be superior to the number of local positive lymph node metastases in predicting patient prognosis, so it can improve the role of prognosis prediction. The prediction model can be used as an important early warning sign for the survival of GC patients for risk estimation. If a patient's risk estimate is high, doctors and nurses can take appropriate measures to promote a clinical prognostic assessment and personalized treatment.
It has been actively prospectively investigated whether chemoradiation provides a survival bene t in addition to or instead of chemotherapy. However, it failed to show a differential survival between patients to receive perioperative chemotherapy vs preoperative chemotherapy and postoperative chemoradiation therapy by randomized CRITICS trial [22]. And the subsequent CRITICS-II trial is optimizing neoadjuvant therapy through comparing only neoadjuvant chemotherapy, only neoadjuvant chemoradiation, and neoadjuvant chemotherapy followed by chemoradiation [23]. Furthermore, another randomized prospective TOPGEAR trial aimed at determining whether preoperative chemoradiation, in addition to perioperative chemotherapy has better outcomes than perioperative chemotherapy alone [24]. the interim results have demonstrated that preoperative chemoradiation can be stabilized delivered to majority of patients with a higher safety [25].