De-escalating chemotherapy for stage I–II gastric neuroendocrine carcinoma? A real-world competing risk analysis

Gastric cancer (GC) is a highly heterogeneous malignancy, including various pathologies which present significantly different molecular patterns, tumor behavior, and prognoses [1]. Gastric neuroendocrine carcinoma (GNEC) is a rare histology type of GC, which occupies 0.1 to 0.6% of whole patients, with an increasing trend over the past few decades [2]. GNEC is a poorly differentiated, high-grade malignancy, which presents dismal prognoses and is prone to distant metastasis [3]. Compared with gastric adenocarcinoma, many studies have observed worse survival in GNEC patients [4, 5].

Based on the Japanese Classification of Gastric Carcinoma, a frequent treatment plan for GNEC is a combination of radical surgery and adjuvant chemotherapy [6]. Xie et al. retrospectively analyzed clinical data from a single center, indicating that stage I-III GNEC patients could benefit from adjuvant chemotherapy (median survival time 43 months vs. 13 months, p = 0.026) [7]. However, a Chinese study recruited 804 GNEC patients from 21 centers between 2004 and 2016, indicating that stage I–II patients could not obtain improved prognoses after adjuvant chemotherapy based on platinum or 5-fluorouracil [8]. There are still no clear clinical guidelines or consensuses that focus on this issue. With the popularity of gastroscopy and other early screening methods, the incidence of early-stage GNEC is increasing. Whether these patients can benefit from adjuvant chemotherapy remain controversial.

A nomogram, also known as a visual risk regression model, is a tool that predicts a specific clinical result, or the probability of a certain event based on the values of multiple clinical indicators and biological data. Despite the development of computer numerical calculation, nomograms demonstrate the advantages of convenience, simplicity, and practicality. They have been used widely to predict prognosis in various tumors, such as colon cancer, breast cancer, thyroid cancer, and so on [9‐11]. However, to the best of our knowledge, a nomogram to predict the survival of patients with stage I–II GNEC patients has not been reported.

Based on the Surveillance, Epidemiology, and End Results (SEER) database, the retrospective research was designed to investigate the potential effect of adjuvant chemotherapy in stage I–II GNEC patients and construct a predictive nomogram.

PSM was then performed to reduce the discrepancy of baseline data between groups. Figure S1 showed that SD in most parameters was less than 0.1, which indicated good balancing performance. Finally, 186 patients were divided into two groups: chemotherapy (n = 93) and no-chemotherapy (n = 93). Table 1 shows the baseline characteristics before and after PSM.

In the research, we identified 424 GNEC patients from the SEER database and performed KM survival analysis and competing risk analysis. The results indicated that adjuvant chemotherapy could improve OS, but it could not improve CSS in these early-stage patients. When evaluating the actual effect of chemotherapy, cancer-specific survival was reported as a more reliable outcome than OS because OS can be diluted by OCD [16]. Furthermore, the presence of OCD might impede the detection of CSD, particularly in stage I–II patients who had a longer lifetime than advanced stage patients [17]. To investigate the potential competing risk of death, we further conducted competing risk analyses. After 1:1 PSM, a similar 5-year cumulative incidence of CSD was observed between the two cohorts (35.4% vs. 31.4%, p = 0.731). And there was no significant difference of cumulative incidence of OCD was found (9.7% vs. 19.7%, p = 0.133). Then the subgroup analyses and multivariate competing risks regression analysis confirmed no significant relationship between chemotherapy and CSD. Patients in the chemotherapy cohort presented higher proportion of male, grade III/IV, primary site of cardia, and tumor size over 5 cm, which had been confirmed as high-risk factors in previous studies [18]. Therefore, a lower percentage of high-risk variables in the no-chemotherapy cohorts and other probable comorbidities not recorded in the SEER database might help partially explain the difference in OCD between the two groups. Relatively long survival time also increases the risk of OCD, such as other malignancies and cardiovascular diseases.

GNEC is characterized by poor biological behavior and resistance to traditional chemotherapy [19]. For GNEC patients without distant metastasis, primary tumor resection remains the first choice. The Japanese Classification of Gastric Carcinoma recommended sequential adjuvant chemotherapy after radical surgery [6]. Several research supported the view of aggressive adjuvant chemotherapy. Xie et al. retrospectively analyzed clinical data from a single center, indicating that stage I-III GNEC patients could benefit from adjuvant chemotherapy (median survival time 43 months vs. 13 months, p = 0.026) [7]. Liu et al. investigated the prognosis of 43 GNEC patients, concluding that postoperative medical therapy was required. The median survival time was 44 months in the chemotherapy group and 15 months in the no-chemotherapy group. Besides, the research recommended the chemotherapy regimen of cisplatin plus Pacditaxel [20]. However, other research presented opposing views. A Chinese study recruited 804 GNEC patients from 21 centers between 2004 and 2016, indicating that stage I-II patients could not obtain improved prognosis after adjuvant chemotherapy based on platinum or 5-fluorouracil [8]. Therefore, there is still controversy about whether early-stage GNEC patients could benefit from adjuvant chemotherapy after radical surgery. Compared with the advanced stage, stage I–II GNEC patients tended to obtain more prolonged survival. However, no research has been conducted into the role of chemotherapy in these specific patients. Based on the competing risk analysis, this is the first study to look at the effect of chemotherapy in stage I–II GNEC patients and find that these patients cannot benefit from adjuvant chemotherapy.

Based on the results of competing risk analysis, a nomogram was proposed to individually predict the survival of stage I–II GNEC patients. Age, gender, grade, T stage, N stage, and tumor size were incorporated into the predictive model. N stage and T stage were considered high-risk factors by examining the max points of the integrated parameters. Previous research showed how these risk variables and GNEC were related. Hu et al. indicated a gradually rising trend of GNEC incidence in the past 40 years and constructed predictive nomograms. Age, grade, TNM stage, and primary tumor resection were significantly correlated with 3-, 5-, and 10-year OS [21]. Xu et al. conducted a comparative study between intestinal-type GC (IGC) and GNEC patients, demonstrating that GNEC patients presented a longer survival to ICG in the early-stage tumor. And age, gender, tumor size, AJCC stage, T stage, N stage, and surgery were considered the risk factors with the OS of GNEC patients [18]. As in our study, the multivariate competing risk analysis found older age, male gender, greater tumor diameter, and higher TNM stage as independent risk factors for CSD. Compared with the traditional multivariate regression model, the nomogram could visually present the individual probability of 1-, 3-, and 5-year CSD. The model was also verified in the validation cohort, illustrating the good predictive potential along with the high credibility of our nomogram.

There are several limitations in the research. To begin with, because this is a retrospective study, selection bias is unavoidable. Second, there is no “neuroendocrine cancer-specific” data in the SEER database. As a result, key critical characteristics for precise stagings, such as the Ki67 value and chemotherapy regimens, cannot be collected. Also, the external validation cohort is relatively small and further external multicenter prospective validations are required.

By using competing risk analysis and 1:1 PSM analysis, we demonstrated that stage I–II GNEC patients could not benefit from adjuvant chemotherapy after surgery. De-escalation of chemotherapy should be considered for stage I–II GNEC patients. The nomogram could individually predict the 1-, 3-, and 5-year CSD in these patients and presented excellent prediction ability. Further research and RCTs are required to validate the conclusion.