Inflammation-based prognostic scores predict the prognosis of locally advanced cervical esophageal squamous cell carcinoma patients receiving curative concurrent chemoradiotherapy: a propensity score-matched analysis

Patient population

Study approval was obtained from the Chang Gung Medical Foundation Institutional Review Board (201800845B0), and written informed consent from the patients or their families was not judged necessary for this kind of retrospective study. We retrospectively reviewed ESCC patients with available medical records who underwent treatment between January 2005 and December 2015 at Kaohsiung Chang Gung Memorial Hospital. The eligibility criteria were as follows: (1) squamous cell carcinoma in histology; (2) locally advanced status, stage III, without distant metastasis or neck/celiac lymph node metastasis; (3) complete CCRT with curative intent; (4) survive more than 3 months after completing CCRT; (5) no history of second primary malignancy, such as head and neck cancers; (6) no form of any acute or chronic infection/inflammatory disease; and (7) Eastern Cooperative Oncology Group performance status 0–1. Ultimately, there were 411 ESCC patients who met the criteria for further analysis, including 63 patients who had tumors located in the cervical esophagus and 348 patients with thoracic esophageal tumors.

In order to prevent selection bias for better comparison, the propensity score matching method was used among the 348 thoracic ESCC patients. First, we used binary logistic regression to calculate a propensity score, with covariates including age, gender, tumor T status, tumor N status, tumor stage, and tumor grade being entered into the propensity model. Subsequently, a one-to-one match with the closest matching scores between the 63 cervical thoracic patients and 63 thoracic ESCC patients was determined. The algorithm used is shown in Fig. 1.

Definition of inflammatory biomarkers and clinical tumor stage

In our study, chest computed tomography (CT), endoscopic ultrasonography (EUS), and positron emission tomography (PET) scans were performed for each patient, and the clinical tumor stage was determined according to the 7th American Joint Committee on Cancer (AJCC) staging system (Edge et al., 2010). The definition of cervical esophageal cancer is that the tumor lies in the neck and is bordered superiorly by the hypopharynx and inferiorly by the thoracic inlet (sternal notch), approximately 15–20 cm from the incisors (Edge et al., 2010).

Blood samples were obtained before treatment to measure the biomarkers of interest, which included the white blood cell count, platelet count, neutrophil count, lymphocyte count, albumin level, and CRP level. The NLR was calculated by dividing the neutrophil count by the lymphocyte count, and the PLR was defined as the platelet count divided by the lymphocyte count; the cut-off values for the NLR and PLR were 2.5 and 103, respectively (Dutta et al., 2011; Xie et al., 2016). The cut-off levels for CRP and albumin in this study were 10 mg/l and 3.5 g/dl, respectively, with these levels being based on those used in previous studies (Forrest et al., 2003; McMillan, 2008; McMillan, 2013; McMillan et al., 2001). The mGPS was calculated using the CRP and albumin values and the scoring system was as follows: (1) patients with a normal CRP value (≤10 mg/l) were allocated a score of 0, regardless of the albumin level; (2) patients with a CRP level >10 mg/l combined with an albumin level ≥3.5 g/dl were allocated a score of 1; (3) patients with a CRP >10 mg/l and an albumin <3.5 g/dl were allocated a score of 2 (McMillan, 2008). The optimal cut-off level for the CRP/albumin ratio was defined as 9.5 in the subsequent analysis (Wei et al., 2015). All the indicators involved in the calculation of the inflammation-based prognostic scores were examined before the patients underwent chemotherapy and radiotherapy.

CCRT setting

Curative intent radiotherapy was planned for each patient, and the details are described below. First, a customized thermoplastic immobilization device was designed. Subsequently, CT-simulation for image acquisition was performed. Inverse plan intensity-modulated radiotherapy (IMRT) was then used to deliver 6- or 10-MV photons to cover the treatment field, including the neck and mediastinum. The gross target volume (GTV) was defined as the gross tumor and gross lymph nodes (LNs) according to chest CT scan and/or PET-CT images. The clinical target volume (CTV) comprehensively covered the esophagus, the mediastinal LNs, and the supraclavicular LNs. The planning target volume (PTV) was defined as the CTV expanded by 0.5–1.0 cm margins in all directions. The total doses prescribed to the PTV were 66–70 Gy in 33–35 daily fractions for cervical ESCC and 50–50.4 Gy in 25–28 daily fractions for thoracic ESCC, followed by a boost dose to the gross neck LNs for an additional 10–16 Gy in 5–8 daily fractions.

Chemotherapy was performed concurrently with radiotherapy and consisted of cisplatin (75 mg/m²; 4-hour infusion) on day 1 and 5-fluorouracil (1,000 mg/m²; continuous infusion) on days 1–4 every 4 weeks. For patients with creatinine clearance <60 mL/min, carboplatin was used instead of cisplatin (Chen et al., 2017a; Chen et al., 2017b; Chen et al., 2018).

Statistical analysis

Comparisons between the groups were performed using the chi-square test for categorical variable data. A Cox regression model was used for univariate and multivariable analyses, and the hazard ratio (HR) and 95 confidence interval (CI) were computed with the Cox proportional hazards model. Overall survival (OS) was calculated from the date of diagnosis of the esophageal cancer to the date of death or last contact. The Kaplan–Meier method was used to estimate OS, and the log rank test was performed to evaluate the differences between the groups for univariate analysis. The statistical analyses were performed with the SPSS 19 software package (IBM, Armonk, NY, USA). All of the tests were two-sided tests, and P < 0.05 was considered statistically significant.

Patient characteristics

There were 411 locally advanced inoperable ESCC patients who completed CCRT with a curative intent at Kaohsiung Chang Gung Memorial Hospital who were retrospectively investigated in this study, including 63 cervical ESCC patients. All 63 cervical ESCC patients had an Eastern Cooperative Oncology Group performance status ≤1, and these patients consisted of 61 male patients and two female patients with a median age of 58 years (range, 37–80 years). The tumor T status was revealed to be T2 in 1 patient (1.6%), T3 in 13 patients (20.6%), and T4 in 49 patients (77.8%). Meanwhile, five patients (7.9%) were diagnosed as having N0 status, 24 patients (38.1%) were diagnosed as having as having N1 status, 23 patients (36.5%) were diagnosed as having as having N2 status, and 11 patients (17.5%) were diagnosed as having as having N3 status. In terms of tumor stage, seven patients (11.1%) had stage IIIA, five patients (7.9%) had stage IIIB, and 51 patients (81.0%) had stage IIIC. Among the 63 patients, 14 patients (22.3%) were classified as grade 1, 36 patients (57.1%) as grade 2, and 13 patients (20.6%) as grade 3.

Using the propensity score matching method, 63 matched patients of the 348 thoracic ESCC patients were identified to compare to the 63 cervical ESCC patients. Parameters between these two groups were all matched without statistical difference, including tumor age, gender, T status, N status, tumor stage and tumor grade. The clinicopathological characteristics of these cervical and thoracic ESCC patients are shown in Table 1.

Inflammation-based prognostic scores and clinical outcomes

The inflammation-based prognostic scores used in our study included NLR, PLR, albumin value, CRP value, mGPS, and CRP/albumin ratio which were all well investigated in previous studies (Dutta et al., 2011; Feng, Huang & Chen, 2014; Forrest et al., 2003; McMillan, 2013; Miyata et al., 2011; Wei et al., 2015; Xie et al., 2016; Yodying et al., 2016). A total of 63 thoracic ESCC patients were identified to completely match the 63 cervical ESCC patients using the propensity score matching method. There were no significant differences in the baseline characteristics of these two groups, except for PLR (P = 0.047). The comparison of inflammatory biomarkers in these cervical and thoracic ESCC patients are shown in Table 2.

With respect to cervical ESCC, NLR ≥ 2.5 (P = 0.019), PLR ≥ 103 (P = 0.013), CRP value >10 (P = 0.037), mGPS ≥ 1 (P = 0.037), and CRP/albumin ratio ≥ 9.5 (P = 0.033) were significant predictors of worse OS in the univariate analysis. Patients with NLR ≥ 2.5 had worse OS compared to others with NLR <2.5 (12.0 months versus 32.6 months, P = 0.016, Fig. 2A); and worse OS (11.6 months versus 25.3 months, P = 0.010, Fig. 2B) was also found in the 21 patients with PLR ≥ 103 than the other 42 patients with PLR <103. The 35 patients with mGPS ≥ 1 were found to have worse OS in comparison with the 28 patients with mGPS of 0 (12.0 months versus 25.3 months, P = 0.037, Fig. 2C); the total of 28 patients who had CRP/albumin ratio ≥ 9.5 had worse OS compared to the other 35 patients who had CRP/albumin ratio <9.5 (12.0 months versus 25.3 months, P = 0.030, Fig. 2D). In a multivariable analysis, PLR ≥ 103 (P = 0.010, HR: 2.66, 95% CI [1.27–5.58]) and mGPS ≥ 1 (P = 0.030, HR: 2.03, 95% CI [1.07–3.86]) were the independent prognostic parameters of worse OS.

With respect to thoracic ESCC, the univariate analysis showed that NLR ≥ 2.5 (P = 0.041), PLR ≥ 103 (P = 0.024), CRP value >10 (P = 0.001), mGPS ≥ 1 (P = 0.001), and CRP/albumin ratio ≥ 9.5 (P = 0.002) were still significant predictors of worse OS, similar to the results in the cervical ESCC group. Worse OS (9.0 months versus 14.4 months, P = 0.038, Fig. 3A) was also found in the 35 patients with ≥ 2.5 than the other 28 patients with NLR <2.5; and patients with PLR ≥ 103 had worse OS compared to others with PLR <103 (9.0 months versus 10.9 months, P = 0.022, Fig. 3B). The total of 35 patients who had mGPS ≥ 1 had worse OS compared to the other 28 patients with mGPS of 0 (8.9 months versus 20.0 months, P = 0.001, Fig. 3C); the 28 patients who had CRP/albumin ratio ≥ 9.5 were found to have worse OS in comparison with the 35 patients who had CRP/albumin ratio <9.5 (9.0 months versus 15.9 months, P = 0.002, Fig. 3D). Patients with NLR ≥ 2.5 (P = 0.012, HR: 2.21, 95% CI [1.19–4.12]) and mGPS ≥ 1 (P < 0.001, HR: 3.13, 95% CI [1.66–5.88]) had worse OS than others with NLR <2.5 and mGPS of 0 in the multivariable analysis. These univariate and multivariable survival analyses are shown in Table 3.