Background

The latest global cancer statistics reveal that gastroesophageal cancer (GEC) is the ninth most common malignancy and the sixth most common cause of cancer-related death [1]. Primary esophageal and GEC can be classified as either adenocarcinoma or squamous cell carcinoma. Caucasians in western countries mainly have adenocarcinoma, whereas Caucasians in eastern countries mostly have squamous cell carcinoma. Presently, GEC is treated by surgery, and although significant progress has been made in the treatment of GEC, the overall 5-year survival rate of patients remains low, which may be due to the lack of an accurate staging system before treatment and a standard treatment protocol [2, 3].

Different stages of GEC require different treatments. Patients with early stages of the disease can be treated with surgery, those with locally advanced stages can be treated with preoperative radiotherapy and chemotherapy, and those with distant metastases can receive palliative care to improve symptoms and quality of life. The prognosis of patients with GEC at different stages varies greatly [4, 5]. Cunningham et al. reported that among patients with resectable GEC, those receiving neoadjuvant chemotherapy had better overall survival (OS) (hazard ratio [HR] = 0.75; 95% CI 0.60–0.93; P < 0.01) and progression-free survival (PFS) (HR = 0.66; 95% CI 0.53–0.81; P < 0.01) compared to those treated with surgery alone [6]. As neoadjuvant chemoradiation replaces EGC resection, the pathological staging of advanced cancer is losing its clinical relevance, although it can still be used as a reference for staging early disease and predicting patient survival. Moreover, neoadjuvant chemoradiation associates with a decreased number of examined lymph nodes (ELNs). The eighth edition of NCCN includes TNM staging for GEC patients who received neoadjuvant chemoradiation; however, there are disadvantages to using this staging system because of the decreased number of ELNs [7]. In addition, it is unclear whether this population can benefit from adjuvant chemotherapy (ACT). Therefore, we propose a new TNM staging system based on the logarithmic odds of positive lymph nodes (LODDS), which provides more accurate pathological staging of patients with GEC who have received neoadjuvant treatment and selects the population who can truly benefit from ACT. The currently recognized comprehensive treatment for GEC still recommends ACT as the recommended treatment method. Nano-activated carbon can play the role of lymph node tracer in the operation of gastric cancer, thyroid cancer, pancreatic cancer and other malignant tumors. Nano-activated carbon also has the same adsorption characteristics as carbon, has a larger specific surface area. Therefore, it can be used as a carrier to adsorb chemotherapy drugs for lymphatic system targeted chemotherapy. Green nanomaterials are the direction of future treatment [8,9,10].

This study included data from the SEER database of patients with resectable GEC who received neoadjuvant chemoradiation. The patients were divided into five groups according to the LODDS in order to evaluate the accuracy and relevance of the new staging system compared to the old staging system in predicting the prognosis of patients. The benefits of the new staging system were validated through external data, and the population that could benefit from ACT was confirmed.

Methods

Patient cohort

SEER*Stat (version 8.4.0) software was used to identify 1 444 patients with resectable GEC confirmed between 2004 and 2019. The inclusion criteria were as follows: (1) patients with pathologically confirmed GEC (ICD-O-3: C16.0), (2) patients with complete follow-up and survival data, (3) patients who received neoadjuvant chemotherapy or radiotherapy, (4) patients who underwent radical surgery, and (5) patients with resectable primary GEC. The variables were age, gender, race, grade, tumor size, tumor stage, ELN, positive examined lymph nodes (PLN), and survival information. Patients were excluded if data were missing. The externally validated data consisted of 914 patients with resectable GEC who received neoadjuvant chemoradiation and underwent surgery in our department from 2011 to 2018. The staging of all tumors was according to the TNM staging system (eighth edition).

Statistical analysis

The distribution of pathological factors was examined, and the TNM staging system was replaced with the LODDS staging system, which was defined as log [(PLN + 0.5)/(ELN-PLN + 0.5)]. LODDS values ranged from − 2.1 to 1.8, with an inter-group spacing of 0.2. Patients were divided into 17 groups, and the HR (range, − 2.1 ≤ LODDS ≤ − 2.0) was defined as 1 and sorted from low to high. The OS of the adjacent two groups was subjected to log rank test, and 4 chi square values with significant differences were selected from 16 chi square values. The 17 groups were divided into five stages. LODDS1 (− 2.1–1.8) consisted of 56 individuals, LODDS2 (− 1.7–1.4) consisted of 395 individuals, LODDS3 (− 1.3–0.2) consisted of 505 individuals, LODDS4 (− 0.1–0.2) consisted of 82 individuals, and LODDS5 (0.3–1.8) consisted of 92 individuals. The survival curves were compared among the five groups, and an improved TLODDSM staging system (I-IVA) was established by replacing the N staging system in the AJCC staging system (eight edition).

When evaluating the discriminative ability of the two staging systems for prognosis, we used the Akaike information criterion (AIC) and Bayesian information criterion (BIC). We used decision curve analysis (DCA) to evaluate the net clinical benefit of the new staging system and compared it with the AJCC staging system (eight edition). We used chi-square test to evaluate the prognostic homogeneity of the two staging systems (the larger the likelihood ratio, the better the prognostic homogeneity of the staging system) and survival curve and COX analysis to analyze the prognosis of patients at different stages. All statistical analyses were conducted using SPSS (version 24.0) and R language (version 4.0.0) software packages with P < 0.05 indicating statistically significant differences.

Results

Patient demographics

A total of 1130 patients with resectable GEC who received neoadjuvant chemoradiation were included (Fig. 1). Specifically, 516 (45.7%) patients were 65 years of age or older, 168 (14.9%) patients were female, 290 (25.7%) patients had poorly differentiated tumors, 410 (36.3%) patients had tumors greater than 5 cm in diameter, and 88 (7.8%) patients had T4 tumors. The median ELN was 14, whereas the median PLN was 1 (Table 1). The median survival period was 27 months (range, 0–190 months), and the number of deaths was 590 (52.2%).

Fig. 1
figure 1

Flowchart of the selection process of included patients

Full size image
Table 1 Characteristics of patients
Full size table

New lymph node staging system

As the number of ELNs increased, the patient’s prognosis improved (Fig. 2A); however, as the LODDS increased, the patient’s prognosis worsened (Fig. 2B). The range of LODDS was − 2.1–1.8, and we divided LODDS into 17 groups at intervals of 0.2. The HR of − 2.1 ≤ LODDS ≤ − 2.0 was defined as 1, and the HR values for each stage were calculated and sorted from low to high, with the highest as 1.0 < LODDS ≤ 1.8 (HR = 7.875; 95% CI 1.672–37.095; P = 0.009). Survival analysis was performed on 17 consecutive stages, and chi square values were calculated. Four high chi square values were identified as cutoff points (4.207, 1.265, 0.392, 0.265), and patients were divided into five stages as follows (Table 2). The 5-year survival rates for LODDS1–5 were 71.87, 57.80, 35.14, 24.14, and 16.46%, respectively, with statistically significant differences, indicating that our staging system has a high discriminative ability (Fig. 2C). According to the eighth edition of staging, we used the LODDS staging system instead of the traditional staging system to divide patients into 35 groups. The HR of T0LODDS1M0 was defined as 1, and the HR values for each stage were calculated and sorted from low to high, with the highest as T4bLODDS5M0 (HR = 398.007; 95% CI 24.479–6471.206; P < 0.001; Table 3). Survival analysis was performed on 35 consecutive stages (Fig. 3A), and chi square values were calculated. Four high chi square values were identified as cutoff points (1.958, 1.007, 1.332, 0.636), and patients were divided into five stages as follows (Fig. 3B): stage I (T0LODDS1), stage II (TisLODDS1), stage IIIA (T1LODDS1, T2LODDS1), IIIB (T3LODDS1, T0LODDS2, TisLODDS2, T1LODDS2, T2LODDS2, T0LODDS3, TisLODDS3, T0LODDS4, T1LODDS3, T2LODDS3, T3LODDS2, T4aLODDS1, TisLODDS4, T0LODDS5, T3LODDS, TisLODDS5, T1LODDS4), and stage IVA (T4aLODDS2, T1LODDS5, T4bLODDS1, T2LODDS4, T4aLODDS3, T3LODDS4, T4bLODDS2, T2LODDS5, T3LODDS5, T4bLODDS3, T4aLODDS4, T4aLODDS5, T4bLODDS4, T4bLODDS5). The HR of stage I was 1. The 5-year survival rates of stage II (HR = 11.118; 95% CI 1.005–123.042; P < 0.05), stage IIIA (HR = 13.058; 95% CI 1.722–99.014; P = 0.013), stage IIIB (HR = 22.675; 95% CI 3.186–161.352; P = 0.002), and stage IVA (HR = 63.741; 95% CI 8.872–457.934; P < 0.001) were 100.00, 100.00, 67.53, 44.58, and 9.60%, respectively, with statistically significant differences (P < 0.05; Fig. 4A). However, the 5-year survival rates of stage I to stage IVA in the traditional TNM staging system were 53.98, 57.42, 41.89, 39.99, and 9.98%, respectively, with no statistically significant differences (P > 0.05; Table 4).

Fig. 2
figure 2

A hazard estimates of death from ELN; B hazard estimates of death from LODDS; C The Kaplan–Meier curves of OS for patients in our new LODDS staging system

Full size image
Table 2 The COX analyses of LODDS groups in the training set
Full size table
Table 3 The COX analyses of TLODDSM groups in the training set
Full size table
Fig. 3
figure 3

A The Kaplan–Meier curves of OS for patients in our new LODDS staging system; B The new TLODDSM staging system

Full size image
Fig. 4
figure 4

A The Kaplan–Meier curves of OS for TLODDSM staging system; B The Kaplan–Meier curves of OS for traditional TNM staging system; C TLODDSM stage were compared to the TNM stage in terms of 3-year OS in our DCA; D TLODDSM stage were compared to the TNM stage in terms of 5-year OS in our DCA

Full size image
Table 4 The COX analyses of different stage groups in the training set
Full size table

Comparison of prognostic effectiveness between the two staging systems

The AIC (7356.862) and BIC (7361.242) of the TLODDSM staging system were lower than the AIC (7385.357) and BIC (7389.737) of the traditional TNM staging system, and the likelihood of the TLODDSM staging system was higher than that of the traditional TNM staging system (chi square test, 113.091 versus 95.953, Table 5). The DCA showed that the TLODDSM staging system had a higher net benefit compared to the traditional TNM staging system (Fig. 4C, D). Taken collectively, these results indicate that the TLODDSM staging system is superior to the AJCC staging system.

Table 5 Performance of the different staging systems for predicting prognosis
Full size table

Selection of a population that can benefit from adjuvant chemotherapy

Patients in stage I of the TLODDSM staging system did not benefit from ACT (P > 0.05; Fig. 5A). Furthermore, stage-II patients did not benefit from ACT (HR = 0.33; 95% CI 0.01–8.18; P > 0.05; Fig. 5B), stage-IIIA patients did not benefit from ACT (HR = 0.54; 95% CI 0.16–1.90; P > 0.05; Fig. 5C), stage-IIIB patients did benefit from ACT (HR = 0.55; 95% CI 0.42–0.71; P < 0.01; Fig. 5D), and stage-IVA patients did benefit from ACT (HR = 0.53; 95% CI 0.31–0.90; P < 0.01; Fig. 5E).

Fig. 5
figure 5

A OS for patients with or without ACT in TLODDSM stage I group in our training set; B OS for patients with or without ACT in TLODDSM stage II group in our training set; C OS for patients with or without ACT in TLODDSM stage IIIA group in our training set; D OS for patients with or without ACT in TLODDSM stage IIIB group in our training set; E OS for patients with or without ACT in TLODDSM stage IVA group in our training set; F OS for patients with or without ACT in TNM stage I group in our training set; G OS for patients with or without ACT in TNM stage II group in our training set; H OS for patients with or without ACT in TNM stage IIIA group in our training set; I OS for patients with or without ACT in TNM stage IIIB group in our training set; J OS for patients with or without ACT in TNM stage IVA group in our training set

Full size image

Patients in stage I of the traditional TNM staging system did benefit from ACT (HR = 0.61; 95% CI 0.36–1.02; P < 0.05; Fig. 5F). Furthermore, stage-II patients did not benefit from ACT (HR = 2.07; 95% CI 0.44–9.74; P > 0.05; Fig. 5G), stage-IIIA patients did benefit from ACT (HR = 0.51; 95% CI 0.37–0.70; P < 0.01; Fig. 5H), stage-IIIB patients did not benefit from ACT (HR = 0.75; 95% CI 0.47–1.21; P > 0.05; Fig. 5I), and stage-IVA patients did benefit from ACT (HR = 0.39; 95% CI 0.15–1.04; P < 0.01; Fig. 5J).

T LODDS M staging system in the external validation group

After considering the inclusion criteria, 914 patients with resectable GEC who received neoadjuvant chemoradiation in our hospital were included. Specifically, 297 (32.5%) patients were 65 years of age or older, 268 (29.3%) patients were female, 413 (45.2%) patients had poorly differentiated tumors, 377 (41.2%) patients had tumors with dimeters greater than 5 cm, and 472 (51.6%) patients had T4 tumors. The median number of ELNs was 18, and the median number of PLNs was 2 (Table 1). The median survival period was 35 months (range, 0–80), and the number of deaths was 433 (47.4%). According to the above criteria, the patients were further divided into 35 groups and classified into five stages based on the TLODDSM staging system (Fig. 6A). Stages I, II, IIIA, IIIB, and IVA had 5-year survival rates of 100.00, 97.78, 88.91, 32.33, and 17.44%, respectively, with statistically significant differences (P < 0.05, Fig. 6B). The 5-year survival rates of stage I to IVA in the traditional TNM staging system were 55.70, 0, 73.54, 55.04, and 20.99%, respectively, with no statistically significant differences (P > 0.05; Fig. 6C).

Fig. 6
figure 6

A The Kaplan–Meier curves of OS for patients in our new LODDS staging system in our external set; B The Kaplan–Meier curves of OS for TLODDSM staging system in our external set; C The Kaplan–Meier curves of OS for traditional TNM staging system in our external set; D TLODDSM stage were compared to the TNM stage in terms of 3-year OS in our DCA in our external set; E TLODDSM stage were compared to the TNM stage in terms of 5-year OS in our DCA in our external set

Full size image

The AIC (5229.220) and BIC (5245.503) of the TLODDSM staging system were lower than the AIC (5285.257) and BIC (5301.54) of the traditional TNM staging system, and the likelihood of the new staging system was higher than that of the old staging system (chi-square test, 275.798 versus 229.595, Table 5). The DCA showed that the TLODDSM staging system had a higher net benefit compared to the traditional TNM staging (Fig. 6D, E). The externally validated data from our center supported this conclusion.

The TLODDSM staging system showed that stage-I patients did not benefit from ACT (P > 0.05; Fig. 7A), stage-II patients did not benefit from ACT (P > 0.05; Fig. 7B), stage-IIIA patients did benefit from ACT (HR = 0.15; 95% CI 0.03–0.77; P < 0.05; Fig. 7C), stage-IIIB patients did benefit from ACT (HR = 0.67; 95% CI 0.47–0.96; P < 0.05; Fig. 7D), and stage-IVA patients did benefit from ACT (HR = 0.73; 95% CI 0.57–0.93; P < 0.01; Fig. 7E).

Fig. 7
figure 7

A OS for patients with or without ACT in TLODDSM stage I group in our external set; B OS for patients with or without ACT in TLODDSM stage II group in our external set; C OS for patients with or without ACT in TLODDSM stage IIIA group in our external set; D OS for patients with or without ACT in TLODDSM stage IIIB group in our external set; E OS for patients with or without ACT in TLODDSM stage IVA group in our external set; F OS for patients with or without ACT in TNM stage I group in our external set; G OS for patients with or without ACT in TNM stage II group in our external set; H OS for patients with or without ACT in TNM stage IIIA group in our external set; I OS for patients with or without ACT in TNM stage IIIB group in our external set; J OS for patients with or without ACT in TNM stage IVA group in our external set

Full size image

The traditional TNM staging system showed that stage-I patients did not benefit from ACT (HR = 1.16; 95% CI 0.57–2.37; P > 0.05; Fig. 7F), stage-II patients did not benefit from ACT (HR = 1.20; 95% CI 0.40–3.55; P > 0.05; Fig. 7G), stage-IIIA patients did benefit from ACT (HR = 0.15; 95% CI 0.03–0.74; P < 0.05; Fig. 7H), stage-IIIB patients did not benefit from ACT (HR = 0.62; 95% CI 0.38–1.03; P > 0.05; Fig. 7I), and stage-IVA patients did benefit from ACT (HR = 0.68; 95% CI 0.54–0.87; P < 0.01; Fig. 7J).

Discussion

The eighth edition of the TNM Staging System for Esophageal Cancer and GEC distinguishes clinical staging (cTNM), pathological staging (pTNM), and neoadjuvant treatment staging (ypTNM), and no longer applies one staging model. Unlike cTNM and pTNM staging, ypTNM staging for the two pathological types, adenocarcinoma and squamous cell carcinoma, is identical [11,12,13]. For the staging of GEC, the seventh edition continues to be used, which emphasizes PLN as the node staging standard, without reflecting the impact of the lymph node station number on staging [14, 15]. Although many studies support the correlation between the lymph node stage number and GEC prognosis, some lymph nodes may fuse due to the invasiveness of the tumor, or some PLNs may rupture during surgical dissection, resulting in inaccurate node staging based on the number. Furthermore, neoadjuvant chemoradiation may result in a decreased number of ELNs [16,17,18]. Therefore, a new node staging system is needed to more accurately evaluate the prognosis of patients. New staging systems for lymph nodes, such as the lymph node ratio (LNR), LODDS, negative examined lymph node (NLN) number, and lymph node micro-metastases, can better predict the prognosis of patients [19,20,21]. According to Yang et al., in the context of a new TNM staging system, LODDS has better predictive value than node number, LNR, and NLN number in patients with PLNs [11].

LODDS is defined as the logarithm of the probability ratio of positive or negative lymph nodes when one lymph node is detected, which effectively balances the number of ELNs and the number of PLNs [22]. Spolverato et al. reported that the LODDS system (C index: 0.636; AIC: 4304.0) was more sensitive than the node score system (C index: 0.632; AIC: 4308.4) and the LNR system (C index: 0.631; AIC: 4225.8) in 804 patients who underwent radical gastrectomy for gastric cancer [23]. Among Siewert type II gastroesophageal adenocarcinoma patients who underwent surgery after neoadjuvant radiotherapy, the LNR was more sensitive than the PLN number, and LODDS was an independent prognostic factor (HR = 1.16; 95% CI 1.08–1.25; P < 0.01) [24]. Zhou et al. demonstrated that LODDS (C index: 0.675; AIC: 6243.740) and LNR (C index: 0.686; AIC: 6261.027) staging systems were superior to the traditional node staging system (C index: 0.658; AIC: 6355.077) in predicting the prognosis of 735 patients with gastroesophageal adenocarcinoma who completed surgery, and externally validated data also confirmed this conclusion [25]. To reduce the stage migration caused by a decreased ELN number [26,27,28], we replaced the traditional TNM staging system with the LODDS staging system (LODDS1–5). The overall survival of patients was significantly improved with the LODDS1–5 staging system, and stages I, II, IIIA, IIIB, and IVA showed statistical differences in the 5-year survival rate, whereas the traditional TNM staging system showed no statistical differences, indicating that the new staging system can better predict the prognosis of patients. Furthermore, we compared the AIC, BIC, and DC of the two staging systems and observed that the heterogeneity in the TLODDSM staging system was relatively low, making it the best staging system for evaluating patient prognosis.

Different stage combinations can lead to completely different prognoses, and it is unclear whether patients with GEC who receive neoadjuvant chemoradiation can benefit from ACT. Rahman et al. reported that patients with GEC who received neoadjuvant chemotherapy benefited from ACT (HR = 0.84; 95% CI 0.77–0.94; P = 0.001) [29]. Mokdad et al. [30] studied 10 086 patients with gastroesophageal adenocarcinoma who received neoadjuvant chemotherapy and radiotherapy in 2006 and 2013 and reported that ACT improved the OS (HR = 0.79; 95% CI 0.72–0.88; P < 0.001) and median survival (40 versus 34 months). On the other hand, Sisic et al. observed no improvement in OS (P = 0.331) and relapse-free survival (P = 0.118) in patients with GEC receiving ACT [31], suggesting not all patients require ACT. Our study shows that stage-I, -II, and -IIIA patients did not benefit from ACT, whereas stage-IIIB and -IVA patients did benefit from treatment. In terms of the traditional TNM staging system, stage- I, -IIIA, and -IVA patients did benefit from ACT, whereas stage-II and -IIIB patients did not benefit from treatment. Our study also shows that the poorer the staging, the more likely the patient is to benefit from ACT. However, traditional TNM stages do not imply that patients with poorer prognosis are more likely to benefit from ACT, indicating that our staging system can guide the postoperative monitoring, follow-up, and prognosis of patients and avoid the consequences of over-treatment and under-treatment.

This study has several limitations. First, the inconsistency of neoadjuvant radiation dosage and chemotherapy regimen can lead to certain biases. Second, different centers, different pathologists, and the lack of unified standards for lymph node detection can also lead to bias. Third, this study was a retrospective study. The optimal cut-off value for the new staging system is still difficult to reach consensus, and when referring to LODDS for staging, the calculation is not fast and convenient. Except for patients with a higher risk of lymph node staging migration, it may not significantly improve the predictive prognosis. Presently, there is no standard protocol for the staging and treatment of GEC after receiving neoadjuvant chemoradiation. However, we established the LODDS staging system to re-stage patients, which can guide treatment and predict prognosis.

Conclusion

The TLODDSM staging system is superior in staging and predicting the prognosis of patients with GEC after receiving neoadjuvant chemoradiation (radiation first). We recommend that stage-IIIB and -IVA patients should receive ACT.