Prognostic nomogram for previously untreated patients with esophageal squamous cell carcinoma after esophagectomy followed by adjuvant chemotherapy

Abstract

Objective

The aim of the study was to establish an effective prognostic nomogram for esophageal squamous cell carcinoma after radical esophagectomy followed by adjuvant chemotherapy in those previously untreated patients.

Methods

The clinicopathological data from 328 patients who underwent radical esophagectomy followed by adjuvant chemotherapy or not at the Tianjin Medical University Cancer Institute and Hospital between 2006 and 2010 were retrospectively studied. Nomograms which predicted survival of esophageal squamous cell carcinoma were established based on the Cox proportional hazards regression model. To determine its predictive accuracy and discriminatory capacity, the concordance index and calibration curve were calculated after bootstrapping in the internal validation. An external validation of 76 patients in 2011 was prospectively studied at the same institution. To verify the performance of the nomogram, the comparison between the nomogram and Tumor-Node-Metastasis staging system was conducted.

Results

The 5-year overall survival was 43.1% in the primary cohort. Based on multivariate analyses, five independent prognostic variables including gender, tumor length, T stage, N stage and chemotherapy cycles were selected to build the nomograms to predict disease-free survival and overall survival. The concordance index of the nomogram to predict overall survival was 0.71 (95% confidence interval, 0.63–0.79), which was superior to the predictive power of Tumor-Node-Metastasis staging system (0.64) in the primary cohort. Meanwhile, the calibration curve showed good accuracy between predictive and actual overall survival. In the validation cohort, the concordance index (0.77) and calibration plot displayed favorable performances. The other nomogram to predict disease-free survival also performed well.

Conclusions

The prognostic nomogram provided individualized risk estimate of survival in patients after esophagectomy followed by adjuvant chemotherapy.

Introduction

Esophageal cancer ranks the eighth in the most common malignancy worldwide and is the sixth leading cause of cancer mortality (1,2). Although esophageal adenocarcinoma is currently predominant in the western countries, esophageal squamous cell carcinoma (ESCC) accounts for the bulk of cancer incidence and mortality globally (3–5) and is the fourth most lethal cancer in China (6). Despite the progress in multidisciplinary treatment in recent years, the prognosis of esophageal cancer is still poor due to its aggressive biological property and rather late stage at diagnosis, with the 5-year overall survival (OS) rate ranges from 15 to 25% (2). Esophagectomy is by far the best curative option for patients with local or locoregional esophageal cancer and keeps to play an important role in achieving local control. With the improvement of the surgical techniques and perioperative nursing strategy, surgery-related complications and mortality have declined (7). Even though comprehensive and effective procedures have been used in selecting appropriate patient for curative surgery, many patients relapse within 2 years and the 5-year OS rates rarely surpass 40% (8). To improve survival after esophagectomy, adjuvant therapies including chemotherapy and radiotherapy have been introduced. Although no consensus has been reached for the effectiveness of such therapies, an increasing number of evidence has shown advantages of adjuvant chemotherapy since JCOG 9204 (9–11).

The establishment of an appropriate cancer staging system is valuable in providing prognostic information and guidance for patient follow-up and subsequent treatment. As with the other cancers, prognostic information for ESCC after esophagectomy is also extremely important. By far, the most widely used staging system for ESCC is the TNM classification system of the seventh edition of the American Joint Committee on Cancer (AJCC). However, a large amount of studies have illustrated that many other clinicopathological factors are also associated with prognosis, such as age (12), gender (13), differentiation grade (14), resection margin status, positive circumferential resection margin (15), lymph node ratio and the presence of extracapsular lymph node involvement (8), which are not currently taken into account in the TNM classification system and instead some other prognostic staging systems based on TNM have been put forward (16–20). In addition, the AJCC TNM staging system is applicable to the general population rather than individual patients. Thereafter, it arises urgent need to propose a more comprehensive prognostic system especially for individual patients.

Nomogram, a statistical predictive model of cancer specific survival, by scoring every single prognostic variable for each single patient, has the power to solve the above-mentioned problems. There have been different nomograms developed for diverse kinds of malignancies, including breast cancer (21–23), extranodal NK/T-cell lymphoma (24), intrahepatic cholangiocarcinoma (25,26), gastric carcinoma (27), pancreatic adenocarcinoma (28). When compared with the traditional systems, these nomograms have shown better accuracy of prediction. With regard to esophageal cancer, several nomograms have also been developed (8,29,30). However, those are mainly for adenocarcinoma. Our present study pays close attention to the prognosis of ESCC, in order to identify critical prognostic factors for OS estimation and to propose an easy and applicable prognostic nomogram approach for patients with esophagectomy followed or not by chemotherapy. To our knowledge, this is the first attempt to establish an ESCC nomogram based on clinicopathological data of a relatively larger patient cohort underwent radical esophagectomy and adjuvant chemotherapy. This study also evaluated whether the model is more effective in survival prediction in comparison to the TNM staging system.

Patients and methods

Patients and study design

The retrospective study included 328 previously untreated patients who underwent radical esophagectomy in the Tianjin Medical University Cancer Institute and Hospital between January 2006 and December 2010. The inclusion criteria were as follows: no neoadjuvant therapy; no history of other malignancies; radical esophagectomy with microscopically negative margin (R0); no irresectable tumors or distant metastases; histopathologically verified ESCC; adjuvant chemotherapy received or not and with complete clinicopathological and follow-up information. The exclusion criteria included the following: perioperative mortality; received adjuvant chemoradiotherapy or radiotherapy; received biotherapy and developed a second primary cancer during follow-up.

For the prospective study, an independent validation cohort of 76 ESCC patients from January 2011 to December 2011 was recruited using the same inclusion and exclusion criteria. The study was approved by the institutional ethics committee and was censored on April 2015. The requirement for informed consent from each patient was waived.

Diagnosis and treatment

Patients were conducted by pretreatment evaluations to ensure clinical stage and rule out surgery contraindications, including a history and physical examination; upper endoscopy with biopsy; routine hematologic examinations; contrast CT scans of chest and abdomen and bronchoscopy and pulmonary function test. Surgery was then performed by experienced esophageal surgeons. Transthoracic esophagectomy with two or three-field lymph node resection was chosen based on the location of tumor and clinical stage. When the tumor was located in the lower esophagus, surgeons mostly chose two-field (thoracic and abdominal) lymph node resection. Furthermore, the upper esophageal cancer, including both the cervical and the upper thoracic esophageal cancer, almost performed three-field (cervical, thoracic and abdominal) node resection. While in the middle esophageal cancer, three-field resection was chosen only if clinically indicated enlargement of cervical lymph nodes, or two-field lymph node resection was enough. The esophagus was dissected en bloc along with its adjacent mediastinal tissues. All patients were categorized according to the AJCC Seventh TNM staging system. It is necessary to note that all T4 patients we included were diagnosed with T4a tumors according to the pathological reports. In addition, the tumor length was defined as the length measured by pathologic sampling. Patients with T3, T4a tumors or had lymph node metastasis were suggested adjuvant chemotherapy in our institution. The major recommended chemotherapy regimen were six courses of 5-fluorouracil (5-FU) plus platinum, and a small portion of patients received 5-FU plus paclitaxel. Those patients with poor tolerance and poor compliance received less than six cycles of the combination regimen, or single agent regimen.

Follow-up

After surgery, all patients were followed up regularly. Patients were observed after surgery at 3–6 month intervals during the first 2 years and every 6–12 months thereafter. At each follow-up, a complete physical examination was conducted, together with abdominal ultrasound and chest X-ray; contrast-enhanced CT, upper GI endoscopy and biopsy and other further examinations as clinically indicated. The follow-up was conducted by telephone to patients or their family members subsequently. Cancer recurrence or metastasis was diagnosed when a new tumor appeared according to imaging or biopsy. The primary endpoint of the study was OS. Disease-free survival (DFS) was calculated from the date of surgery to the time of recurrence or metastasis. Furthermore, OS was defined as the period from esophagectomy to the date of death, or until the last follow-up. Those who were died from other causes or still alive on April 2015 were considered to be censored data.

Statistical analysis

Statistical analyses to identify risk characteristics were performed using IBM SPSS Statistics, Version 20.0. Demographic, clinicopathologic and treatment data were described using percentages or median values. Categorical variables were compared using the χ²-test or Fisher's exact test. When the variables were ordinal, the non-parametric test was conducted. Continuous variables were managed using the t-test. Survival curves according to prognostic factors were estimated with the Kaplan–Meier method and compared using the log-rank test. The prognostic variables used to establish the nomogram were selected based on the Cox proportional hazards regression model using backward stepwise selection with the Akaike information criterion. Hazard ratios (HRs) and 95% confidence intervals (CIs) of these factors were estimated to quantify the strength of these associations. All statistical tests were two-tailed.

The nomogram that can visualize the prognostic strength of these different risk factors in a single figure was constructed using the package of rms in R, version 3.1.3 (http://www.r-project.org/). In the validation of the nomogram, internal validation by bootstraps with 1000 resamples was used. Then, the concordance index (C-index) and calibration curve were calculated to determine its predictive accuracy and discriminatory capacity. Compared with the TNM staging system, the C-index was calculated. The larger C-index, the more favorable accuracy of the prognostic model. Finally, we performed external validation, in which total points of each patient in validation cohort were calculated according to the novel nomogram. Then, the Cox regression was conducted using patients' total points as a factor to evaluate the value of the established nomogram. The regression analysis was used to derive the C-index and the calibration plot.

Results

Clinicopathological characteristics and survival of patients

The baseline characteristics of the two cohorts are listed in Table 1. In the primary cohort, the median age was 61 years (range, 36–92 years). There were more men than women in this study (ratio, 4.56:1). Most patients presented with local advanced stage (III stage, 57.6%). The same as before, the middle esophagus (61.6%) was predominant site of ESCC. Nearly half of patients (163/328) received chemotherapy in the primary cohort. Furthermore 57.1% (93/163) of patients received more than four cycles, the rest of patients received only one to three cycles. The regimen of 5-FU plus platinum was chosen for 82.2% (135/163) of patients. The remaining patients received the regimen of 5-FU plus paclitaxel (16/163, 9.8%), or irregular regimens (12/163, 7.4%).

Table 1.

The baseline characteristics of all patients with esophageal squamous cell carcinoma

Characteristics	Primary cohort (n = 328), number of patients (%)	Validation cohort (n = 76), number of patients (%)	P value
Sex			0.379
Male	269 (82.0)	59 (77.6)
Female	59 (18.0)	17 (22.4)
Age (years)			0.434
<60	152 (46.3)	39 (51.3)
≥60	176 (53.7)	37 (48.7)
History of smoking			0.751
Yes	222 (67.7)	50 (65.8)
No	106 (32.3)	26 (34.2)
History of drinking			0.342
Yes	140 (42.7)	37 (48.7)
No	188 (57.3)	39 (51.3)
Surgical complications			0.823
Yes	33 (10.1)	7 (9.2)
No	295 (89.9)	69 (90.8)
Location			0.172
Upper	24 (7.3)	8 (10.5)
Middle	202 (61.6)	38 (50.0)
Lower	102 (31.1)	30 (30.9)
Length (cm)			0.003
Median	4.2	3.7
Range	0.3–14.0	1.0–6.0
Differentiation			0.556
Good	14 (4.3)	4 (5.3)
Moderate	249 (75.9)	58 (77.4)
Poor	65 (19.8)	13 (17.3)
T stage			0.070
T1	19 (5.8)	10 (13.1)
T2	85 (25.9)	23 (30.3)
T3	63 (19.2)	11 (14.5)
T4a	161 (49.1)	32 (42.1)
N stage			0.375
N0	189 (57.6)	40 (52.6)
N1	96 (29.3)	23 (30.3)
N2	32 (9.7)	11 (14.5)
N3	11 (3.4)	2 (2.6)
TNM stage			0.598
IA + IB	16 (4.9)	8 (10.5)
IIA + IIB	123 (37.5)	25 (32.9)
IIIA + IIIB + IIIC	189 (57.6)	43 (56.6)
Adjuvant chemotherapy			0.001
No	165 (50.3)	22 (28.9)
Yes	163 (49.7)	54 (71.1)

The median follow-up time was 44.9 months (range, 3.4–107.6 months) of all patients (Fig. 1A) in the primary cohort. The 1-, 3- and 5-year survival was 86.6, 55.1 and 43.1%, respectively. The 5-year OS of patients with adjuvant chemotherapy versus surgery alone was 50.5 versus 36.1%, respectively (Fig. 1B).

Figure 1.

Kaplan–Meier overall survival (OS) curves of esophageal squamous cell carcinoma (ESCC) patients in the primary cohort. (A) The survival curve for all patients in the primary cohort (n = 328). (B) Patients with adjuvant chemotherapy displayed more favorable survival than those with surgery alone (n = 328).

Independent prognostic factors of patients

The results of the univariate analysis are listed in Table 2. The risk factors that predicted OS were selected into a Cox proportional hazards regression model to conduct multivariate analysis. In addition, considering that the potential benefit of chemotherapy was possibly associated with cycles of the therapy, this variable was put into the model. Multivariate analysis identified four independent prognostic variables that were most associated with OS: gender, tumor length, N stage and adjuvant chemotherapy cycles (Table 3). Furthermore, in view of the higher hazard ratio of T stage and its prognostic value of T stage has been verified in the previous study, T stage was also used here to construct the prognostic nomogram.

Table 2.

Univariate analysis of the primary cohort

Prognostic factor	DFS			OS
	P	HR	95% CI	P	HR	95% CI
Gender (female versus male)	0.016	0.611	0.409–0.912	0.005	0.544	0.354–0.834
Age (≥60 versus <60)	0.830	0.970	0.733–1.283	0.400	1.130	0.850–1.501
History of smoking (yes versus no)	0.245	1.198	0.883–1.626	0.149	1.259	0.921–1.721
History of drinking (yes versus no)	0.078	1.287	0.972–1.704	0.315	1.157	0.871–1.536
Surgical complications (yes versus no)	0.507	1.169	0.736–1.857	0.350	1.241	0.789–1.953
Length	0.000	1.137	1.068–1.210	0.000	1.150	1.083–1.222
Location
Middle versus upper	0.661	1.132	0.650–1.971	0.381	1.304	0.720–2.362
Lower versus upper	0.449	1.250	0.702–2.225	0.329	1.360	0.733–2.522
Differentiation
Moderate versus good	0.616	1.214	0.569–2.593	0.333	1.454	0.681–3.105
Poor versus good	0.221	1.644	0.741–3.647	0.331	1.491	0.667–3.336
T stage
T2 versus T1	0.034	3.561	1.102–11.515	0.065	3.029	0.933–9.840
T3 versus T1	0.006	5.215	1.605–16.942	0.009	4.838	1.489–15.720
T4a versus T1	0.000	8.244	2.618–25.963	0.001	7.336	2.330–23.101
N stage
N1 versus N0	0.000	2.216	1.625–3.021	0.000	2.153	1.569–2.954
N2 versus N0	0.000	2.475	1.551–3.949	0.000	2.925	1.868–4.582
N3 versus N0	0.000	4.025	2.085–7.769	0.000	4.156	2.208–7.826
TNM stage
II versus I	0.026	4.982	1.217–20.397	0.044	4.256	1.037–17.462
III versus I	0.001	10.200	2.522–41.249	0.002	9.166	2.267–37.061
Chemotherapy cycles
1–3 versus 0	0.246	0.799	0.547–1.167	0.394	0.855	0.596–1.226
≥4 versus 0	0.052	0.725	0.525–1.002	0.004	0.602	0.425–0.851

DFS, disease-free survival; OS, overall survival; HR, hazard ratios; CI, confidence interval.

Table 3.

Multivariate analysis of the primary cohort

Prognostic factor	DFS			OS
	P	HR	95% CI	P	HR	95%CI
Gender (female versus male)	0.027	0.627	0.413–0.949	0.007	0.544	0.350–0.846
Length	0.033	1.082	1.006–1.164	0.002	1.117	1.040–1.199
T stage
T2 versus T1	0.281	2.538	0.466–13.819	0.453	1.981	0.332–11.802
T3 versus T1	0.103	4.209	0.746–23.737	0.163	3.642	0.594–22.347
T4a versus T1	0.022	7.843	1.344–45.777	0.048	6.462	1.017–41.052
N stage
N1 versus N0	0.000	2.141	1.513–3.031	0.000	2.055	1.447–2.918
N2 versus N0	0.000	2.963	1.727–5.084	0.000	3.515	2.075–5.953
N3 versus N0	0.000	4.611	2.263–9.395	0.000	4.469	2.280–8.760
Chemotherapy cycles
1–3 versus 0	0.078	0.706	0.480–1.039	0.032	0.667	0.461–0.965
≥4 versus 0	0.000	0.474	0.334–0.672	0.000	0.376	0.259–0.546

Nomogram development and internal validation

Nomograms for DFS and OS prediction were established using the five prognostic factors (Fig. 2A and B). In the internal validation, the C-indices for predicting DFS and OS were the same after bootstrapping, which was 0.71 (95% CI, 0.63–0.79). The calibration curve for the probability of 5-year DFS and OS showed an optimal agreement between the actual observed outcome and the prediction by the novel nomogram (Fig. 3A and C).

Figure 2.

The nomograms for predicting DFS (A) and OS (B) in patients with ESCC after radical esophagectomy and adjuvant chemotherapy. To estimate the survival of an individual patient, the value of each prognostic factor was acquired on each variable axis, followed by a line drawn upward to determine the number of points. Then, the sum of these five numbers is located on the total points axis, and a straight line is drawn downward to the survival axes to determine the likelihood of 1-, 3- or 5-year survival.

Figure 3.

The calibration curves for both nomograms. The calibration curve for predicting patient DFS at (A) 5 years in the primary cohort and at (B) 3 years in the validation cohort. The calibration curve for predicting patient OS at (C) 5 years in the primary cohort and at (D) 3 years in the validation cohort. The nomogram-predicted probability of survival is plotted on the x-axis; and the actual survival is plotted on the y-axis.

External validation of nomogram for DFS and OS

The median follow-up was 42.3 months (range, 3.3–50.9 months) in the validation cohort. The 1- and 3-year survival was 90.8 and 61.5%, respectively. Further, 71% (54/76) of the patients received adjuvant chemotherapy, and 46.3% (25/54) performed more than four cycles, 94.4% (51/54) chose the regimen of 5-FU plus platinum.

In the external validation, the C-index to evaluate the predictive accuracy for DFS was 0.65 (95% CI, 0.46–0.85). The calibration curve showed good agreement between the actual survival and the prediction by the novel nomogram (Fig. 3B).

The predictive accuracy for OS as measured by the C-index was 0.77 (95% CI, 0.60–0.93) in the external validation, demonstrating that it is a model with good discriminative capability. The calibration plot suggests that the nomogram has good performance, based on the good agreement between the prediction and the actual observation in the probability of 3-year OS (Fig. 3D).

Comparison of predictive accuracy between the nomogram and TNM stage system

To evaluate the prognostic value of the nomogram, the comparison between the nomogram and the AJCC Seventh Tumor-Node-Metastasis (TNM) Staging system was conducted in both cohorts. The C-indices for DFS prediction were 0.63 and 0.62 by TNM staging system in the primary and validation cohorts, respectively. The C-indices for OS prediction were 0.64 and 0.73 by TNM in the primary and validation cohorts, respectively, all were inferior to the nomogram. These results suggested that the nomogram possesses an improved predictive accuracy for clinical outcomes in patients with ESCC undergoing esophagectomy and adjuvant chemotherapy.

Discussion

With the incidence number of esophageal adenocarcinoma arising in the western countries, ESCC is still predominant in China. It has been demonstrated that the histologic tumor type is an independent prognostic parameter in esophageal cancer, for that esophageal adenocarcinoma has a better long-term prognosis than ESCC (31). Given the differences in genetic background, epidemiology, pathophysiology and pathogenesis between the two types (2,3), prognostic system specific for ESCC needs to be attended. The AJCC TNM staging system has taken an initiative role in stratifying this type of patients. The prognosis of this staging system is based on pathological assessments, including the invasion depth of the primary tumor, the number of positive lymph nodes and the status of distant metastasis. However, results from a number of studies have questioned the adequacy of the TNM system in both esophageal adenocarcinoma and ESCC, and reversions have been proposed (16–20). Some researchers have even combined TNM classification with molecular biology data to assess the prognosis of patients with ESCC (32). Nomogram is considered to be a useful tool capable of combining variety of prognostic factors perfectly for individual prediction, and prognostic nomogram for esophageal adenocarcinoma has been developed earlier (8,29,30). This study established a novel nomogram to predict survival targeted ESCC, especially for those with adjuvant chemotherapy.

This nomogram is aimed to estimate the probability of 1-, 3- and 5-year OS based on a multivariate Cox proportional hazards model. Several clinicopathologic features and molecular characteristics have emerged as important prognostic factors in ESCC over the past years. Our nomogram utilizes five variables which are either significantly associated with clinical outcomes for ESCC, such as gender, tumor length, T stage and N stage, or factors that we believed to be potentially relevant to the benefit of therapy which is chemotherapy cycles. It is previously widely known that sex differences affect esophageal cancer incidence, and it has caught people's attention as a prognostic factor nowadays. Regardless of different sex-related tumor-specific environmental exposures (e.g. alcohol, tobacco), growing number of evidence suggest that hormone levels influence this phenomenon (33,34). Furthermore, Bohanes et al. demonstrated that women of 55 years or older with locoregional squamous cell or younger than 55 but with metastatic squamous cell have a significantly better outcome, and these associations are possibly related to both estrogen and androgen exposition (13). Our results showed that tumor length was an independent prognostic factor, similar to previous studies (35). Due to the longitudinal growth of ESCC in the lymphatic-rich submucosa, micrometastases or even regional lymph node metastasis already have occurred beyond a certain length. Mohamad et al. (35) proposed a modified TNM classification system using a 3-cm tumor length cut-off point, which indeed matches to our tumor length cut-off point according to the ROC curve (data not shown). Increasing depth of invasion (defined as T stage) has also been verified as a predictor of regional lymph node metastasis and poor survival. With regard to N stage, it undoubtedly is an independent prognostic variable. Some authors make attempts on improving the N stage, such as redistributing the numbers of positive lymph nodes (19), redefining regional lymph nodes and nonregional lymph nodes (16) and ruling out the number of dissected lymph nodes (8). However, these standpoints need more investigation.

It is advocated that multidisciplinary therapy is the best choice for ESCC especially when the disease is locoregional, whereas surgery is still the principal and radical therapy in majority of patients. Hence, the prognosis of patients after esophagectomy is more important to evaluate, which may play a very important role in scheduling the individual follow-up strategy. This study pays attention to the value of adjuvant chemotherapy. Several data have confirmed that adjuvant chemotherapy can improve survival, especially in patients with lymph node metastasis (11). Our results demonstrated that the effects were associated with the chemotherapy cycles. However, it is not recommended in regard to adjuvant radiotherapy in ESCC. A multicenter controlled trial has proved that postoperative radiation therapy does not increase survival after radical resection for squamous cell carcinoma of the middle and lower esophagus (36,37). ESCC occurs mainly in the middle, and the emphasis is very much on the complications of the middle esophageal cancer radiotherapy, such as dysphagia caused by esophagitis or even pulmonary complications that may influence the patient's quality of life severely. While the value of adjuvant radiotherapy is controversial, it has a potential prospect in adjuvant chemotherapy. Furthermore, evidence such as JCOG9907 (38) had concluded that patients with preoperative adjuvant chemotherapy had more favorable survival than that with postoperative chemotherapy. However, due to some conditions of the restrictions, neoadjuvant treatment in China was not widespread. The data about neoadjuvant chemotherapy were limited, and therefore this study only discussed the benefit of adjuvant chemotherapy.

This nomogram was based on a Cox proportional hazards regression model and performed well in both internal and external validation. The effects of several separate and independent clinical variables are integrated by a visual nomogram to give an individualized risk assessment for each patient. Undoubtedly, the advantage of this methodology is apparent from a number of previous clinical settings (23–28). Similar to those studies, our nomogram also showed more favorable ability in predicting OS compared with TNM staging system. Although a plethora of biomarkers were associated with prognosis, the prominent status of the AJCC TNM staging system cannot be challenged. However, the TNM staging system only considered the invasion depth of the primary tumor, the status of regional lymph nodes and distant metastases. Our nomogram put into other important prognostic factors, such as gender, tumor length and chemotherapy cycles, with which the nomogram could stratify patients more effectively. Then, we conducted comparisons between our nomograms and TNM staging system. As the C-indices showed, the predictive power of the two nomograms were superior to the TNM staging system in both the primary and validation cohorts, which demonstrated that our nomograms had more effective individualized risk estimates of survival and more favorable application prospect than TNM staging system in future.

Although the novel nomogram demonstrated good levels of accuracy for the prediction of survival, there are some limitations in this study. First, this nomogram was developed based on data obtained from a single institution, its performance waits to be tested in larger cohorts including multiple geographic regions. Second, the therapeutic strategies may appear distinct between various medical institutions. The effectiveness of the nomogram remains to be evaluated. Third, the prognostic variables we used in this study were rather restricted to common clinicopathological features. In the future, with more and more biological and genetic features (32,39–44) become more and more relevant, as well as hematological information such as plasma fibrinogen level (45), platelet count (45) and lymphocyte–monocyte ratio (46), and finally, with the optimization of clinical treatment strategy, sustained refinement of the nomogram is undoubtedly necessary, in which more and different prognostic factors might be included in the model.

In conclusion, we have incorporated gender, tumor length, T stage, N stage and chemotherapy cycles into a prognostic nomogram to predict OS of ESCC. In comparison to TNM staging system, this nomogram outperforms in the prediction of prognostic information in patients with esophagectomy and adjuvant chemotherapy, both in the training and validation groups. It might also help to tailor individual follow-up schedules. Further studies may help to expend the validation of the method and improve the modeling by parameter optimization.

Authors' contributions

J.D. collected and analyzed data. T.D. wrote the manuscript. L.Z., M.B., H.L., H.Y., and Y.Q. contributed to analyze the data. Y.B., designed the project and edited the manuscript. Y.B. is the guarantor of this work and, as had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

This work was supported by grants from the National Natural Science Foundation of China (Nos. 81201946, 81372394, 30772484, 81401257, 81171120) and National Research Platform of Clinical Evaluation Technology for New Anticancer Drugs (No. 2013ZX09303001).

Conflict of interest statement

The authors declare that there is no conflict of interests regarding the publication of this article.