A novel nomogram and risk stratification system predicting the cancer-specific survival of patients with gastric neuroendocrine carcinoma: a study based on SEER database and external validation

Currently, the low incidence rate and high heterogeneity hindered further investigation of GNEC [15]. Thus, we created and validated a nomogram to predict the 1-, 3-, and 5-year CSS of surgical GNEC patients based on the SEER database. The model demonstrated improved prediction ability compared with the 8th AJCC criteria. We then developed a novel risk stratification system that divided all cases into low-, middle-, and high-risk groups and showed a more remarkable ability to distinguish different risk groups than the conventional AJCC staging system by using X-tile software to determine the cutoff value for the best grouping.

Based on the univariate and multivariate Cox regression analysis, ten parameters (age, gender, grade, T stage, N stage, metastasis, primary site, tumor size, RNE, and chemotherapy) significantly affecting CSS in postoperative GNEC patients were enrolled in the predictive nomogram. By examining the max points of the integrated parameters in the model, grade and TNM stage were regarded as highly significant variables influencing the prognoses. Previous research had shown how these risk variables and GNEC were related. Hu et al. investigated current GNEC epidemiological trends and developed a nomogram to assess these patients’ prognoses. According to the findings, the survival of GNEC was substantially correlated with grade, T, and N staging [12]. Another Chinese study examined the features of 132 Chinese GNEC patients and found that the patients’ survival was independently predicted by the size of the tumor, N stage, mitotic index, radical surgery, and adjuvant treatment [7]. Xu et al. conducted a study comparing intestinal-type GC (IGC) and GNEC patients, finding that GNEC patients had a better prognosis than ICG in the early stages of the tumor. Furthermore, age, gender, tumor size, AJCC stage, T stage, N stage, and surgery were all substantially associated with the overall survival of GNEC patients [16]. Concordantly, the Cox regression analysis proposed in our study also identified old age, gender of male, larger tumor size, and higher TNM stage as independent risk variables for CSS. Larger tumor size and higher TNM stage usually indicated more aggressive tumor behavior with worse survival. Age was also a risk factor, which might be attributed to the factor that elderly patients had worse general conditions and suffered more chronic diseases. The nomogram model included risk factors that were easily gotten and collected from the patient’s hospitalization information. By using the variable score, clinicians could predict the prognosis accurately and evaluate the treatment benefits. This provided a visual means of communication between clinicians and patients. Through the nomogram chart, patients could know their own probability of survival, which would prompt them to make more reasonable treatment choices. Besides, several variables with higher scores, such as older age and gender of male, should be concerned for the relatively poor prognosis.

It was reported that GNEC patients presented irresponsiveness to traditional chemotherapy [17]. But based on our analysis, chemotherapy could improve postoperative survival among GNEC after surgical resection, indicating that adjuvant chemotherapy should still be adopted. In China, patients with GNEC were typically treated with the same chemotherapy agents as those with gastric adenocarcinoma, including fluorouracil, cisplatin, streptomycin, allium annulus, and paclitaxel [18, 19]. Therefore, specific chemotherapeutic drugs and regimens for GNEC should be further explored. In addition, as shown in the model, the primary site of cardia was correlated with poorer prognoses. GC could be divided into cardia and non-cardia according to the tumor location, which presented different epidemiology and tumor behavior [20]. And GC patients with cardia invasion tended to suffer worse survival in previous studies [21]. So, our research first validated that the primary site of cardia was a risk factor in postoperative GNEC patients.

Currently, the AJCC TNM staging system is the main option for cancer prognosis prediction. However, its application in GNEC patients has recently been called into question [7]. It was reported that other clinical parameters, such as gender, grade, and adjuvant therapy, were significantly related to the prognoses of GNEC patients [22,23,24]. So, we created a nomogram by combining various factors (such as demographic and clinicopathologic traits) affecting CSS in patients with GNEC. Additionally, depending on their overall scores, patients were categorized into low-, middle-, and high-risk groups. On this premise, it was possible to compare the power of the nomogram with the conventional AJCC staging system, which was not performed in the previous research. The nomogram demonstrated better predictive ability than tumor staging based solely on AJCC criteria, according to the NRI, IDI, and C-index data. Besides, the effectiveness of the predictive model was tested in both the internal validation and external validation cohorts. According to the KM curves, the stage I and stage II patients couldn’t be well distinguished in the AJCC staging system. The limited ability to discriminate between stage III and stage IV was also presented. Surprisingly, compared to the conventional staging system, the KM analysis showed significantly different CSS among the three risk groups, which can help clinicians customize their management and treatment plans.

Limitation

The current study included various limitations due to the nature of the database, which should be considered when interpreting the results. To begin, because this was a retrospective study, selection bias was unavoidable. Second, the SEER database lacked information on precise data on performance status, comorbidities, varied chemotherapy regimens, treatment for distant metastases, somatostatin analogues, and “neuroendocrine cancer-specific” data. Furthermore, the SEER database’s GNEC classification was not completely the same as the WHO classification. Finally, even though the research was externally validated, the small sample data could only provide limited support for the result. Multicenter data with a larger sample size are required to further evaluate the predictive model.