Development of a novel staging classification for Siewert II adenocarcinoma of the esophagogastric junction after neoadjuvant chemotherapy

Abstract

BACKGROUND

Stage classification for Siewert II adenocarcinoma of the esophagogastric junction (AEG) treated with neoadjuvant chemotherapy (NAC) has not been established.

AIM

To investigate the optimal stage classification for Siewert II AEG with NAC.

METHODS

A nomogram was established based on Cox regression model that analyzed variables associated with overall survival (OS) and disease-specific survival (DSS). The nomogram performance in terms of discrimination and calibration ability was evaluated using the likelihood-ratio test, Akaike information criterion, Harrell concordance index, time-receiver operating characteristic curve, and decision curve analysis.

RESULTS

Data from 725 patients with Siewert type II AEG who underwent neoadjuvant therapy and gastrectomy were obtained from the Surveillance, Epidemiology, and End Results database. Univariate and multivariate analyses revealed that sex, marital status, race, ypT stage, and ypN stage were independent prognostic factors of OS, whereas sex, race, ypT stage, and ypN stage were independent prognostic factors for DSS. These factors were incorporated into the OS and DSS nomograms. Our novel nomogram model performed better in terms of OS and DSS prediction compared to the 8^th American Joint Committee of Cancer pathological staging system for esophageal and gastric cancer. Finally, a user-friendly web application was developed for clinical use.

CONCLUSION

The nomogram established specifically for patients with Siewert type II AEG receiving NAC demonstrated good prognostic performance. Validation using external data is warranted before its widespread clinical application.

Key Words: Stage classification, Prognosis, Esophagogastric junction cancer, Neoadjuvant chemotherapy, Siewert type

Core Tip: So far, the ideal staging classification for Siewert II AEG treated with neoadjuvant chemotherapy is lacking. Thus, we established nomogram based on the Cox regression model which incorporating variables associated with overall survival (OS) and disease-specific survival (DSS). The novel nomogram model showed the best performance compared with 8^th American Joint Committee of Cancer pathological staging schemes of esophagus cancer and gastric cancer. The time-receiver operating characteristic curve of the novel nomogram showed an excellent predictive value in terms of 5-year OS [0.665 (0.626-0.704)] and 5-year DSS [0.675 (0.636-0.713)]. Finally, a friendly online web application was developed for clinical use.

INTRODUCTION

In recent decades, the incidence of adenocarcinoma of the esophagogastric junction (AEG) has increased globally. Furthermore, AEG is the leading cause of cancer-related death and accounts for approximately 1.2 million deaths per year worldwide[1]. AEG may be a special type of tumor that is different from esophageal cancer (EC) and gastric cancer (GC) with regards to its location and biological behavior[2-3]. Siewert type II AEG, which arises from the true gastric cardia, differs from Siewert type I and III AEG. Most Siewert type II tumors infiltrate deep into the gastric or esophageal wall and metastasize to the lymph nodes (LNs), therefore, most of the tumors of this type are diagnosed at an advanced stage[4]. Currently, neoadjuvant therapy is mainly offered to patients with clinically advanced Siewert type II AEG, since this treatment option can lead to improved survival compared to patients who first undergo surgery[5].

Owing to the urgent need for precision approaches in cancer care, in 2017, the 8th edition of the American Joint Committee on Cancer (AJCC) proposed separate classification of clinical (cTNM), pathological (pTNM), and postneoadjuvant therapy [y-pathologic (yp)-TNM] stages for AEG[6]. However, the value of ypTNM staging in clinical decision making is currently limited for several reasons. First, the current ypTNM stages are not differentiated based on Siewert type, and the suitability of ypTNM for Siewert type II AEG remains unclear. Second, unlike the cTNM and pTNM classifications, the ypTNM classification does not differentiate adenocarcinoma from squamous cell carcinoma[6]. Third, survival differences are less distinct between the ypTNM classes[6]. Fourth, the prognostic accuracy of the 8th edition AJCC ypTNM staging system for AEG remains uncertain[7].

Therefore, a new staging system based on a treatment strategy specific for Siewert II AEG is warranted. To our knowledge, this is the first study to explore staging classifications for patients with Siewert II AEG treated with neoadjuvant chemotherapy (NAC).

MATERIALS AND METHODS

Patients and data collection

Patient demographics and clinicopathological variables were obtained from the Surveillance, Epidemiology and End Results (SEER) public database. From the 18 SEER registries, the 2021 release of the public-use dataset collected from 2004 to 2015 was queried to identify patients with pathological diagnostic confirmation of Siewert type II AEG. Although the SEER database did not provide detailed information on the Siewert type classification for AEG, research using combined selection terminology of “Primary Site,” [encoded 160 (Cardia)], and “CS site-specific factor 25” [encoded 982 (esophagus, gastroesophageal junction)], facilitated identification of patients with Siewert type II AEG[4]. In the “CS lymph nodes eval” category, a value of six was used to select patients with AEG who received neoadjuvant therapy before surgery in the SEER database, therefore, the patient met criteria for AJCC ypTNM staging. Patients were included if they had pathological confirmation of T1-4aN0-3M0 stage AEG after surgery, along with pathological examination of the LNs. TNM stage was updated according to the 8th edition of the AJCC criteria[8]. In total, 725 patients who underwent NAC were included in the primary cohort. The detailed selection process is shown in Figure 1. The SEER Stat software [Surveillance Research Program, (www.seer.cancer.gov/seerstat), version 8.3.9] from the National Cancer Institute was used to access the database.

Figure 1 Flow chart.

Primary outcomes

Overall survival (OS) and disease-specific survival (DSS) were the primary endpoints. OS was defined as the time from diagnosis to death from any cause. The causes of death were coded using the SEER database according to the data extracted from the death certificate. DSS was defined as the time from diagnosis to AEG-related death.

Statistical analysis

Continuous variables are reported as mean ± SD. Categorical variables were compared using the chi-square test or Fisher’s exact test, and continuous variables were evaluated using the Mann–Whitney U test. Survival curves were estimated using the Kaplan–Meier method, and the log-rank test was used to determine significance.

Model construction

Crude and age-adjusted univariate analyses were performed to identify potential risk factors. After selecting the potential risk factors, we performed multivariate analyses using three selection procedures (forward, backward, and stepwise) to select the best-fit model. A statistical significance level of 0.20 was used for the variables selected for the model. After comparing the models from each procedure, a final model was obtained using backward selection with P < 0.2[9]. For illustration and clinical applicability, online nomograms were created for novel nomograms (OS model: https://huangcmgc.shinyapps.io/ZJAEGOS/; DSS model: https://huangcmgc.shinyapps.io/ZJAEGCSS/).

Model evaluation

The likelihood ratio test was used to assess homogeneity between groups and to evaluate the performance of the predictive models from multiple dimensions. The Akaike information criterion (AIC) was applied to test the goodness of fit[10,11]. The Harrell concordance index (C-index) with bootstrap resampling (n = 1000) and time-dependent area under the receiver operating characteristic (ROC) curve were calculated to assess the prediction accuracy[12,13]. Time-dependent ROC analysis is an extension of the ROC curve, which assesses the discriminatory power of a prognostic model for time-dependent disease outcomes[14]. The area under curve (AUC) was calculated, in addition to an intuitive comparison of the ROC curves[15,16]. For each time point, the AUC estimated the probability that a deceased patient was classified in a higher staging category than a patient who remained alive. Sequential AUCs were compared between the LN ratio (LNR) and the 8^th AJCC TNM staging systems. Decision curve analyses were performed to assess the clinical utility of prediction models by quantifying the net benefits when different threshold probabilities were considered. Generally, the strategy with the highest net benefit at any given risk threshold was considered to have the highest clinical value[17].

All analyses were two-sided, and P < 0.2 indicated statistical significance[9]. All statistical analyses were performed using R version 4.0.1 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Baseline

Overall, 725 patients with AEG who underwent NAC were included in the primary cohort, of which 623 were males (85.9%) and 102 were females (14.1%). The patients diagnosed with ypT1, ypT2, ypT3, and ypT4a AEG stages were 82 (11.3%), 108 (14.9%), 441 (60.8%), and 94 (13.0%), respectively and 218 (30.1%), 223 (30.8%), 186 (25.7%), and 98 (13.5%) patients were diagnosed with ypN0, ypN1, ypN2, and ypN3 stage AEG, respectively (Table 1).

Table 1 Clinicopathological characteristics of Siewert type II adenocarcinoma of the esophagogastric junction underwent radical gastrectomy with neoadjuvant chemotherapy.

Characteristics	n (%)/mean ± SD
Age, (yr)
< 65	468 (64.6)
≥ 65	257 (35.4)
Sex
Male	623 (85.9)
Female	102 (14.1)
Marital status
Married	502 (69.2)
Divorced or separated	79 (10.9)
Single (never married)	91 (12.6)
Widowed	30 (4.1)
Others	23 (3.2)
Race
White	647 (89.2)
Others	78 (10.8)
Year of operation
2004-2009	230 (31.7)
2010-2015	495 (68.3)
Type of surgery
Subtotal gastrectomy	418 (57.7)
Total gastrectomy	156 (21.5)
Unknown	151 (20.8)
Tumor size
< 50 mm	387 (53.4)
≥ 50 mm	200 (27.6)
Unknown	138 (19.0)
The number of examined lymph nodes	17.4 ± 10.9
yp T stage
T1	82 (11.3)
T2	108 (14.9)
T3	441 (60.8)
T4	94 (13.0)
yp N stage
N0	218 (30.1)
N1	223 (30.8)
N2	186 (25.7)
N3	98 (13.5)
Grade
Well differentiated	35 (4.8)
Moderate differentiated	253 (34.9)
Low differentiated	400 (55.2)
The tumor grade cannot be identified	37 (5.1)
Adjuvant chemotherapy
No	530 (73.1)
Yes	139 (19.2)
Unknown	56 (7.7)

Model construction

As shown in Table 2, in univariate analysis, sex, marital status, race, ypT stage, ypN stage, and adjuvant chemotherapy were associated with OS. These factors were incorporated into the multivariate analysis, and an OS nomogram was constructed based on the factors with P < 0.2, including sex, marital status, race, ypT stage, and ypN stage. Only sex, race, ypT stage, and ypN stage were independent prognostic factors for DSS (all P < 0.2), and were included in the DSS nomogram (Table 3). The details of our novel nomogram for predicting OS and DSS are illustrated in Figure 2.

Figure 2 Online web nomograms for predicting 1-, 3-, and 5-year. A: Overall survival for Siewert type II adenocarcinoma of the esophagogastric junction (AEG) underwent radical gastrectomy with neoadjuvant chemotherapy; B: Disease-specific survival for Siewert type II AEG underwent radical gastrectomy with neoadjuvant chemotherapy. OS: Overall survival; DSS: Disease-specific survival; ACT: Adjuvant chemotherapy.

Table 2 Univariate and multivariate analysis of overall survival for Siewert type II adenocarcinoma of the esophagogastric junction underwent radical gastrectomy with neoadjuvant chemotherapy.

	Univariate analysis			Multivariate analysis
	HR	95%CI	P value	HR	95%CI	P value
Age
< 65 yr	1
≥ 65 yr	1.11	0.93-1.33	0.259
Sex
Male	1			1
Female	0.81	0.62-1.06	0.120	0.74	0.57-0.98	0.035
Marital status
Married	1			1
Divorced or separated	1.38	1.05-1.8	0.019	1.51	1.15-1.99	0.003
Single (never married)	1.11	0.85-1.46	0.435	1.07	0.81-1.4	0.642
Widowed	1.40	0.91-2.15	0.130	1.34	0.86-2.08	0.198
Others	0.77	0.43-1.37	0.368	0.74	0.41-1.31	0.302
Race
White	1			1
Others	0.81	0.6-1.1	0.173	0.8	0.59-1.08	0.148
Year of diagnosis
2004-2009	1
2010-2015	1.01	0.84-1.22	0.910
Type of surgery
Subtotal gastrectomy	1
Total gastrectomy	1.04	0.83-1.3	0.762
Tumor size
< 50 mm	1
≥ 50 mm	1.08	0.88-1.32	0.477
The number of examined lymph nodes	1	0.99-1	0.379
yp T stage
T1	1			1
T2	1.42	0.96-2.1	0.082	1.25	0.84-1.87	0.265
T3	1.89	1.36-2.63	< 0.001	1.43	1.01-2.02	0.043
T4a	2.66	1.82-3.88	< 0.001	1.85	1.24-2.75	0.003
yp N stage
N0	1			1
N1	1.40	1.1-1.78	0.006	1.37	1.07-1.76	0.012
N2	1.90	1.49-2.42	< 0.001	1.84	1.42-2.38	< 0.001
N3	3.13	2.37-4.12	< 0.001	3.11	2.31-4.18	< 0.001
Grade
Well differentiated	1
Moderate differentiated	0.87	0.56-1.35	0.539
Low differentiated	1.31	0.86-2	0.211
Gx	1.01	0.58-1.77	0.976
Adjuvant chemotherapy
No	1			1
Yes	0.80	0.63-1.02	0.067	0.67	0.52-0.85	0.001

HR: Hazard ratio.

Table 3 Univariate and multivariate analysis of disease-specific survival for Siewert type II adenocarcinoma of the esophagogastric junction underwent radical gastrectomy with neoadjuvant chemotherapy.

	Univariate analysis			Multivariate analysis
	HR	95%CI	P value	HR	95%CI	P value
Age
< 65 yr old	1
≥ 65 yr old	1.01	0.83-1.24	0.895
Sex
Male	1			1
Female	0.75	0.56-1.01	0.060	0.75	0.55-1.01	0.056
Marital status
Married	1
Divorced or separated	1.23	0.9-1.67	0.197
Single (never married)	1.23	0.93-1.63	0.150
Widowed	1.18	0.71-1.96	0.516
Others	0.84	0.46-1.54	0.579
Race
White	1			1
Others	0.78	0.55-1.09	0.142	0.72	0.51-1.01	0.054
Year of diagnosis
2004-2009	1
2010-2015	0.96	0.78-1.18	0.699
Type of surgery
Subtotal gastrectomy	1
Total gastrectomy	1	0.78-1.28	0.985
Tumor size
< 50 mm	1
≥ 50 mm	1.08	0.86-1.35	0.504
The number of examined lymph nodes	1	0.99-1.01	0.516
yp T stage
T1	1			1
T2	1.44	0.9-2.3	0.125	1.25	0.78-2.01	0.347
T3	2.29	1.55-3.38	< 0.001	1.69	1.13-2.54	0.010
T4a	3.19	2.06-4.94	< 0.001	2.18	1.38-3.44	0.001
yp N stage
N0	1			1
N1	1.56	1.19-2.05	0.001	1.43	1.08-1.88	0.012
N2	2.22	1.7-2.92	< 0.001	1.93	1.45-2.56	< 0.001
N3	3.7	2.73-5.02	< 0.001	3.17	2.31-4.35	< 0.001
Grade
Well differentiated	1
Moderate differentiated	0.93	0.56-1.55	0.795
Low differentiated	1.53	0.93-2.49	0.091
Gx	1.11	0.58-2.1	0.759
Adjuvant chemotherapy
No	1
Yes	0.88	0.68-1.12	0.301

HR: Hazard ratio.

Model evaluation

The performance of the nomograms in predicting the outcomes was evaluated by calculating the C-index, AIC, and log-likelihood. Our novel nomogram performed better than the 8^th AJCC pathological TNM staging system for EC and GC (the nomogram had the highest C-index and log-likelihood and the smallest AIC value) in predicting OS and DSS (all P < 0.001) (Table 4). The nomogram was subjected to 1000 bootstrap resampling for internal validation. The calibration curve also showed good agreement between the 5-year OS and DSS rates predicted by the novel nomogram and the actual 5-year OS and DSS rates (Figure 3A and B). The time-ROC curve of the novel nomogram showed excellent predictive value for 5-year OS [0.665 (0.626–0.704)] and DSS [0.675 (0.636–0.713)] (Figure 3C and D). Decision curves also showed that the novel nomogram provided a higher net benefit than the 8^th AJCC pathological TNM staging system for EC and GC in terms of predicting the 5-year OS and DSS rates (Figure 3E and F).

Figure 3 Comparison of nomogram properties. A and B: Calibration of the nomogram for 5-year overall survival (OS) and disease-specific survival (DSS); C and D: Time- receiver operating characteristic curves of the nomogram for 5-year OS and DSS; E and F: Decision curves of the nomogram for 5-year OS and DSS. EC: Esophageal cancer; GC: Gastric cancer; ypTNM: Y-pathologic-TNM.

Table 4 Model evaluation.

Models	Harrell's C-index	AIC	Log likelihood	P value
OS
Novel nomogram	0.635 (0.609-0.662)	-2923.6	5875.2	Ref.
8th AJCC EC ypTNM staging system	0.610 (0.583-0.636)	-2940.6	5889.1	0.004
8th AJCC GC ypTNM staging system	0.593 (0.568-0.619)	-2948.4	5900.7	< 0.001
DSS
Novel nomogram	0.635 (0.607-0.664)	-2451.9	4927.7	Ref.
8th AJCC EC ypTNM staging system	0.627 (0.600-0.655)	-2458.9	4925.8	0.218
8th AJCC GC ypTNM staging system	0.609 (0.582-0.636)	-2467.2	4938.4	< 0.001

OS: Overall survival; DSS: Disease-specific survival; AIC: Akaike information criterion; AJCC: American Joint Committee on Cancer; EC: Esophageal cancer; GC: Gastric cancer; ypTNM: Y-pathologic-TNM.

Clinical use

An easy-access online web application was developed to intuitively demonstrate the survival probability by inputting the values of four critical variables (age, pathological T stage, LNR, and histological grade). Web applications are available at https://huangcmgc.shinyapps.io/ZJAEGOS/ for the OS (Figure 2A) and https://huangcmgc.shinyapps.io/ZJAEGCSS/ for the DSS (Figure 2B).

DISCUSSION

NAC, which is superior to the surgery-first approach, is increasingly being recognized as an effective treatment option for clinically advanced AEG[18,19]. Hosoda et al[20] found that NAC was associated with improved prognosis in patients with Siewert type II AEG. NAC improved both the 3- and 5-year OS and PFS rates compared to adjuvant chemotherapy and surgery alone in patients with resectable AEG[21]. In addition, NAC can reduce the overall recurrence rate of AEG[22].

Accurate staging is crucial for cancer treatment. The ypTNM cancer staging was added to the 8th edition of the AJCC manual. Prognostication is specific for patients receiving neoadjuvant therapy and is not included in current classifications[23,24]. For the first time, an independent pathological staging system was developed for patients with EC or GC who underwent surgery after neoadjuvant therapy[7]. However, some controversies remain and no global consensus regarding the optimal use of ypTNM exists. Additionally, although the International GC Society Staging Project collected data on more than 25000 GC, patients who received neoadjuvant therapy were excluded from the study[25]. Previous studies demonstrated that ypTN categories are not represented in prognostic stratification. Sisic et al[7] found no significant prognostic difference between ypT0 and ypT0-2 stages in AEG, and only ypT4 showed significantly worse survival rate[7]. Moreover, no clear prognostic distinction was observed between the ypN1- and ypN2-disease stages. Guo et al[26] reported that the TNM staging system demonstrated good prognostic performance in patients with Siewert type II AEG without neoadjuvant therapy, whereas ypT staging was ineffective in stratifying patients who received neoadjuvant radiotherapy.

In addition, although significant biological differences were observed between the Siewert and histological types, the ypTNM staging system for AEG does not consider the effects of these differences[27]. Staging classification, particularly for Siewert type II AEG after NAC, has not been extensively studied and remains unvalidated. Compared with classical TNM staging, individualized nomogram predictions can be more accurate for clinical decision-making for a variety of malignancies[28-30]. Liu et al[4] suggested that a nomogram could be a promising clinical tool for assessing OS in patients with Siewert type II AEG after preoperative radiotherapy. Moreover, factors other than ypT and ypN stages significantly influence patient prognosis, therefore, they should be considered when establishing individual prediction models[31-34]. In this study, data were retrieved from a specific cancer database (SEER) to identify potential prognostic factors, specifically for patients with Siewert II AEG receiving NAC. A nomogram was constructed using independent prognostic factors. Our results clearly demonstrate that neither the esophageal nor gastric schema of ypTNM is perfect for the classification of type II AEG. We also demonstrated the superiority of the specific nomogram in predicting accurately 5-year DSS and OS compared to the ypTNM-EC and ypTNM-GC staging systems. Our findings may positively influence future classification revisions. To the best of our knowledge, the present study is the first to review a large dataset from a national cancer registry to explore individualized staging in patients with Siewert type II AEG undergoing NAC.

This study had some limitations. First, the nomogram is based on a single dataset. Additionally, further investigation in external validation cohorts, preferably from several different regions is required. Therefore, collaborative work with other research groups to build a validated nomogram using multicenter, real-world data is warranted. Another limitation is the lack of information regarding the chemotherapy regimen, drug dose, and duration of NAC. Third, the current SEER database does not include indicators, such as clinical response to NAC, which may affect the prognosis of patients receiving NAC, thus affecting the accuracy of prognosis prediction. Fourth, the nomogram did not contain important information regarding gene mutations, human epidermal growth factor receptor 2 and microsatellite instability status, or other biomarkers. In the era of immunotherapy, both the AJCC-ypTNM staging system and individual nomograms require major revisions to better guide immunotherapy and targeted therapy. Despite this limitation, the SEER remains a valuable national database for the study of cancer treatment[35,36]. Moreover, our nomogram could encourage oncologists to explore staging systems that may predict survival more accurately in patients with Siewert II AEG receiving NAC.

CONCLUSION

To the best of our knowledge, this is the first study to present a novel staging scheme based on a retrospective review of a large, population-based, data registry, that is superior to ypTNM-EC and ypTNM-GC staging systems in terms of prognostic accuracy in patients with Siewert type II AEG after NAC. Our study presents inherent limitations, therefore our results warrant external data validation before clinical application.