Good prediction of treatment responses to neoadjuvant chemoradiotherapy for esophageal cancer based on preoperative inflammatory status and tumor glucose metabolism

To develop a tool for predicting pathologic complete response (pCR) after neoadjuvant chemoradiotherapy (neoCRT) in patients with esophageal cancer by combining inflammatory status and tumor glucose metabolic activity. This study included 127 patients with locally advanced esophageal cancer who had received neoCRT followed by esophagectomy from 2007 to 2016. We collected their neutrophil–lymphocyte ratio (NLR) and standardized uptake value (SUV) obtained from fluorodeoxyglucose positron emission tomography (PET/CT) before and after neoCRT. Univariate and multivariate logistic regression analyses were performed to identify potential predictive factors for pCR. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of predictors were calculated. Between pCR and non-pCR groups, there were no statistically significant differences in patient characteristics, such as sex, age, site, and clinical T/N stage. Multivariate analyses identified four independent predictors for pCR, including pre-OP NLR < 5.4 [OR 11.179; 95% CI 8.385–13.495; p = 0.003], NLR change (ΔNLR) < 3 [OR 4.891; 95% CI 2.274–9.180; p = 0.005], changes in SUV (ΔSUV) > 7.2 [OR 3.033; 95% CI 1.354–6.791; p = 0.007], and SUV changes ratio (ΔSUV ratio) > 58% [OR 3.585; 95% CI 1.576–8.152; p = 0.002]. ΔNLR had the highest accuracy and NPV (84.3% and 90.3%, respectively). Combined factors of ΔNLR < 3 and ΔSUV ratio > 58% had the best PPV for pCR (84.8%). Inflammatory status (ΔNLR) and tumor glucose metabolic activity (ΔSUV ratio), when considered together, constitute a promising low-invasive tool with high efficacy for prediction of treatment response before surgery.

Apart from inflammatory factors, functional images, such as Fluorodeoxyglucose Positron Emission Tomography (FDG-PET/CT), are also indicators of responses to neoCRT. FDG-PET/CT uses the radiolabeled glucose analog 18F-fluorodeoxyglucose (FDG) to detect the glucose metabolic activity of cancer cells. Changes in FDG uptake after neoCRT reflect changes in glucose metabolic activity by tumor cells which can be used to predict treatment responses 14,15 . A systemic review demonstrated the ratio of change in maximum standardized uptake value at one hour (ΔSUV ratio) after neoCRT is associated with treatment response. However, the cut-off point of the ΔSUV ratio varies, and the sensitivity and specificity of the ΔSUV ratio are heterogeneous 14 . It is risky to predict treatment response after neoCRT based solely on tumor glucose metabolic activity.
Here we aimed to determine the predictive values of combined inflammatory status and tumor glucose metabolism in patients with esophageal cancer treated with neoCRT followed by surgery.

Materials and methods
Study population. We first reviewed patients with histologically proven squamous cell carcinoma (SCC) of the esophagus at clinical stage T2-4N0-3M0 based on the 8th edition of the American Joint Committee on Cancer staging manual. Patients who received neoCRT followed by surgery were then included for analysis. The exclusion criteria were: (a) histology other than SCC; (b) distant metastasis at initial diagnosis; (c) synchronous malignancy; (d) previous history of malignancies within five years, except non-melanoma skin cancer or adequately treated carcinoma in situ of the cervix; (e) concurrent chemotherapy other than cisplatin and 5-fluorouracil (5-FU) and (f) incomplete treatment.
Our study was approved by the Institutional Review Board of Taichung Veterans General Hospital. Written informed consent was obtained from all participating patients. All patients were treated according to the National Comprehensive Cancer Network (NCCN) guideline for esophageal cancer.
Radiotherapy was given with a total dose from 50 to 50.4 Gy in 1.8 to 2.0 Gy per fraction, five daily fractions per week.
During the treatment period of radiotherapy, concurrent chemotherapy was given intravenously with cisplatin at 20 mg/m 2 drip for 2 h and fluorouracil 800 mg/m 2 for 24 h on Days 1-4 (cycle 1), and Days 29-32 (cycle 2).
A second PET-CT scan was performed within 3-4 weeks after the completion of neoCRT to evaluate the treatment response before surgery. Patients received surgery within 4-6 weeks after completion of neoCRT. The surgery procedure included esophagectomy, lymph node dissection, and esophagus reconstruction with the gastric tube. Pathologic reports included differentiation, margin status, tumor regression grade (TRG), and nodal metastasis. Pathological complete response (pCR) was defined as the absence of tumor cells in the primary tumor and all dissected lymph nodes.
Inflammatory status and tumor glucose metabolic activity. Data on inflammatory status were extracted from medical records. The neutrophil-to-lymphocyte ratio (NLR) was defined as the absolute neutrophil count divided by the absolute lymphocyte count. All differential counts were taken within 1 week before neoCRT (Pre-CRT NLR) and 1 week before surgery (Pre-OP NLR). The change of NLR (ΔNLR) was defined as pre-OP NLR substrate pre-CRT NLR. Tumor glucose metabolic activity was defined as the maximum standardized uptake value at one hour of primary esophageal tumor which was obtained from PET/CT before neoCRT (pre-CRT SUV) and after neoCRT (pre-OP SUV). The change of SUV (ΔSUV) after neoCRT was defined as pre-CRT SUV minus pre-OP SUV. The ratio of change in SUV (ΔSUV ratio) was defined as ΔSUV divided by pre-CRT SUV.

Statistical analyses.
Patients' characteristics' included sex, age, histological differentiation, site, clinical T/N classification, pre-treatment tumor length, time interval between neoCRT to surgery, and numbers of dissected lymph nodes. Between pCR and non-pCR groups, the chi-square test or Fisher's exact test was used to compare categorical variables, and the 2-sample t-test was used to compare continuous variables. The associations between clinical parameters and pCR were evaluated by univariate and multivariate analyses using the logistic regression model. Variates with p < 0.05 and confounders, including clinical T stage and clinical N stage in the univariate analysis, were further explored using multivariate analysis. Receiver operating characteristic (ROC) curves for pCR prediction were used to verify the best cut-off point for pre-CRT NLR, pre-OP NLR, ΔNLR, pre-CRT SUV, pre-OP SUV, ΔSUV, and SUV ratio. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated for both tumor glucose metabolic and inflammatory factors.
Survival outcomes were analyzed using the Kaplan-Meier method and log-rank tests. OS was measured from the date of biopsy until death from any cause or time of the most recent follow-up. Progression-free survival (PFS) started from the date of biopsy until death or disease progression. Data were analyzed statistically using SPSS software (version 25.0, IBM, NY).

Results
We retrospectively analyzed 127 patients who had received neoCRT between October 2007 and July 2016. All of them completed neoCRT and then received surgery. Patients' characteristics are shown in ROC curves for pre-OP NLR, ΔNLR, ΔSUV, and ΔSUV ratio in predicting pCR were plotted in Supplementary Information 1. The optimized cut-off values for pre-OP NLR and ΔNLR were 5.4 and 3, respectively. The best cut-off points for ΔSUV and ΔSUV ratio were 7.2 and 58%, respectively. The prediction efficacy of the inflammatory (pre-OP NLR and ΔNLR) and the tumor glucose metabolic factors (ΔSUV and ΔSUV ratio) are shown in Table 3 and 4. The distribution of ΔNLR and ΔSUV ratio in patients with or without pCR is shown in Fig. 1. ΔNLR showed the highest accuracy and NPV (84.3% and 90.3%, respectively). In tumor glucose metabolic markers, ΔSUV ratio showed great sensitivity and NPV (77.8% and 72.7%, respectively). Combined factors of inflammatory status (ΔNLR < 3) and tumor glucose metabolic marker (ΔSUV ratio > 58%)  Table 3. Predictive efficacy of inflammatory factors and tumor glucose metabolism for pCR. PPV positive predict value, NPV negative predict value, pre-OP NLR neutrophil-lymphocyte ratio before operation, CCRT concurrent chemoradiotherapy, OP operation, pre-CCRT NLR neutrophil-lymphocyte ratio before concurrent chemoradiotherapy, pre-OP NLR neutrophil-lymphocyte ratio before operation, ΔNLR the change of neutrophil-lymphocyte ratio before and after concurrent chemoradiotherapy, ΔSUV the change of maximum standardized uptake value of the esophageal tumor before and after concurrent chemoradiotherapy.

Discussion
We reviewed the data of patients with locally advanced esophageal SCC to determine the predictive value of combining the conditions of inflammatory status and tumor glucose metabolism. We found that patients with both low ΔNLR (< 3) and high ΔSUV ratio (> 58%) were associated with a higher probability of pCR [OR 21.543, (95% CI 8.047-57.670)]. The PPV was 84.8% based on the combined condition of a low ΔNLR and a high ΔSUV ratio. Patients with high ΔNLR and low ΔSUV showed a higher probability of having residual tumors. To the best of our knowledge, this is the first study to assess the accuracy of combining both inflammatory status and tumor glucose metabolic activity for predicting pCR in patients with esophageal SCC after neoCRT. NLR was calculated with absolute neutrophil and lymphocyte counts derived from differential counts, which is a low-cost and low-invasive method. SUV obtained from PET-CT is also a minimally invasive procedure. After neoCRT, ΔNLR combined with ΔSUV can help physicians predict the treatment response before surgery.
In previous studies, high NLR was found to be associated with poor survival outcomes in esophageal cancer patients 11 . However, the time points of NLR vary among different reports. Sharaiha et al. analyzed esophageal cancer patients who received surgery with or without neoCRT. They found that high pre-OP NLR (> 5) was associated with poor OS and PFS 16 . Feng et al. also confirmed this result with a different cut-off point of NLR (> 3.5) 17 . In a few studies, high pre-CRT NLR was reported to have a negative impact on survival in patients treated with neoCRT followed by surgery 18 . Yoo et al. reported that a high pre-CRT NLR was also a negative predictor in the definitive CCRT setting 19 .
Some recent studies have used NLR to predict pCR after neoadjuvant treatments. McLaren et al. analyzed 60 esophageal cancer patients and found that a high pre-CRT NLR was a negative predictor of pCR 20 . Hyder et al. reported the smaller the increase in NLR (< 4.6), the higher the probability of achieving pCR 21 . In these studies, an association was found between NLR and pCR. However, these studies involved only a few cases and their predictive efficacy is unclear. In our analysis, we reviewed 127 patients and found that lower NLR increase (ΔNLR < 3) had a 4-to fivefold probability of achieving pCR [OR 1.891; 95% CI 2.274-9.180; p = 0.005].
In the published reports on esophageal cancer patients, changes in pre-OP PET-CT parameters were found to reflect the tumor response after neoCRT and have an impact on survival outcomes 15,22 . Patients with ΔSUV > 60% after neoadjuvant CCRT showed better OS and disease-free survival 15 . However, a systematic review of the predictive value of PET-CT for tumor response, based on 20 reports, showed variable ranges in sensitivity (33-100%), specificity (30-100%), and differences in PPV and NPV 14 . Hence, prediction of treatment response based on PET-CT parameters alone may be risky as the level of accuracy reported in the literature varies considerably.
To improve the predictive value of treatment response, we herein combined ΔNLR and ΔSUV ratio. Inflammation stimulates angiogenesis, influences immune surveillance on tumor cells, and affects treatment response 10 . As a marker for systemic inflammation, low NLR reflects a low neutrophil count and a high lymphocyte count. Neutrophils are associated with the production of various inflammatory mediators, which are related to systemic inflammation. Lymphocytes are crucial in cell-mediated immunity to initiate tumor cell death. Low NLR represents a low level of systemic inflammation and better cell-mediated immunity toward tumor cells and consequently better response to treatment.
Generally, the intensity of tumor proliferation is correlated with its glucose metabolic activity. Tumor proliferation affects the clinical behavior, which determines prognosis and treatment response. In clinical practice, PET/ CT with a radiolabeled glucose analog can detect glucose metabolic activity, which helps to predict prognosis in   23 . Overexpression of glucose transporter 1 (GLUT-1) in SCC was associated with a higher FDG-uptake and a negative survival impact 24 . In esophageal SCC, FDG-PET/CT was a useful tool for predicting survival and an indicator of residual tumor burden before surgery 15 . Greater drops in FDG-uptake, measured as SUV after treatment, reflect a decline in tumor metabolism of glucose, or a better response to treatment. Both NLR and SUV are therefore predictors of anticancer treatment response. Esophagectomy is a major surgical operation associated with morbidities such as leakage (10%), mortality (3 to 5%), and impairment of life quality 9 . Elderly patients with multiple comorbidities may have higher risks of post-esophagectomy complications 25 . A more accurate preoperative assessment of treatment response, such as the ΔNLR and ΔSUV ratio, would allow surgeons to make an individualized treatment plan for their patients. Those who are at high risk for postoperative complications, but have a high probability of achieving a pCR, could potentially be spared the morbidity and mortality of surgery. The organ-preserving approach with active surveillance is an alternative treatment for these patients [26][27][28] . Van der Wilk et al. reported that patients with esophageal cancer who achieved a clinical complete response after neoCRT and underwent active surveillance had a survival outcome comparable to immediate surgery after neoCRT. The 3-year OS was 77% vs. 55% (p = 0.104) and the 3-year PFS was 60% v. 54% (p = 0.871) in the active surveillance and immediate surgery groups, respectively 27 . Taketa et al. described similar findings. The median OS was not statistically different between the patients who declined surgery (DS) and those who received surgery (trimodality therapy, TMT) after neoCRT (57.9 months v. 50.8 months, p = 0.28). There was a favorable trend toward TMT in relapse-free survival (RFS), but the difference in 3-year RFS was not statistically significant (23.8% v. 45.2%, p = 0.45) 28 .
This study had some limitations. The investigation was retrospective in nature, was conducted in a single institution, and the sample size was relatively small in 9 years. One advantage of this study was the highly homogenous treatment protocol that was applied in all reviewed patients. Another advantage was the high compliance of patients who received neoCRT. All of our patients completed radiotherapy and two cycles of concurrent chemotherapy.

Conclusion
Change of both NLR and SUV ratio after neoCRT were good predictors of pCR. This finding may help decisionmaking with respect to the organ-preserving approach in elderly patients with comorbidities.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.