Preoperative serum lipids as prognostic predictors in esophageal squamous cell carcinoma patients with esophagectomy

This study was to evaluate the prognostic significance of serum lipids in esophageal squamous cell carcinoma patients who underwent esophagectomy. Preoperative serum lipids were collected from 214 patients who were diagnosed with esophageal squamous cell carcinoma. All of the patients received esophagectomy in Qilu Hospital of Shandong University from January 2007 to December 2008. The records and data were analyzed retrospectively. We found that low total cholesterol (for T stage, p = 0.006; for TNM stage, p = 0.039) and low-density lipoprotein cholesterol (for T stage, p = 0.031; for TNM stage, p = 0.035) were associated with advanced T stage and TNM stage. Kaplan-Meier survival analysis indicated that low total cholesterol and low-density lipoprotein cholesterol were associated with shorter disease-free survival(for total cholesterol, p = 0.045; for low-density lipoprotein cholesterol, p < 0.001) and overall survival (for total cholesterol, p = 0.043; for low-density lipoprotein cholesterol, p < 0.001). Lower low-density lipoprotein cholesterol/high-density lipoprotein cholesterol ratio (LHR) indicated poorer disease-free survival and overall survival (both p < 0.001). In the multivariate analysis, low-density lipoprotein cholesterol and LHR were independent prognostic factors for disease-free survival and overall survival. In conclusion, our study indicated that preoperative serum total cholesterol and low-density lipoprotein cholesterol are prognostic factors for esophageal squamous cell carcinoma patients who underwent esophagectomy. LHR can serve as a promising serum lipids-based prognostic indicator.


INTRODUCTION
In 2012, there were 455 800 new esophageal cancer (EC) cases and 400 200 deaths worldwide, especially in developing countries [1]. In high-risk region, including some Asian countries, esophageal squamous cell carcinoma (ESCC) were responsible for 90% of all cases with a significant increasing incidence rate [2]. According to the National Comprehensive Cancer Network (NCCN) guideline of ESCC (version 2, 2016), esophageal resection should be considered for all physiologically fit patients with resectable esophageal cancer [3]. Surgical treatment is still the recommended curative option in ESCC patients considering multiple anti-tumor modalities, unless contraindications exist. Nevertheless, the 5-year overall survival (OS) remains only 26.2% to 49.4% after surgery [4]. Therefore, there is an urgent need for verifying effective biomarkers to predict prognosis of ESCC patients after surgery.
Various studies had discussed the close relationship between serum lipids and cancer, including gastrointestinal cancer, breast cancer and other cancers [5][6][7][8]. Yan-Yan Liu et al found that pretreatment high-density lipoprotein cholesterol (HDL-C) was a poor prognostic factor in nasopharyngeal carcinoma patients by enhancing proliferation and migration of tumor cells [9]. It is broadly accepted that low total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-C) are highly relevant with  [8,10,11]. Saito et al found low LDL-C was associated with higher mortality, being a prognostic biomarker in liver patients [12]. As reported, serum TC and LDL-C significantly decreased in EC patients, indicating an altered lipid metabolism associated with EC [13]. Considering disequilibrium of lipids metabolism commonly occurred in cancer patients, the indicator containing several lipids may reflect the change of serum lipids more accurately and comprehensively. One research reported that LDL-C and high-density lipoprotein cholesterol (HDL-C) ratio (LHR) was a prognostic factor in colorectal patients, providing further prognositic information than single LDL-C or HDL-C [5]. Although serum lipids have been discussed in various cancers, studies about lipids and ESCC are limited. Relationships of serum lipids with survival prognosis in ESCC patients are also unclear. To our best knowledge, no research has been published previously regarding the LHR in ESCC patients. Consequently, we conducted this retrospective study to verify the correlations of serum lipids with clinicopathological characteristics and assess their prognostic values on disease-free survival (DFS) and OS in ESCC patients.

Associations of TG, TC, LDL-C and HDL-C with clinicopathological characteristics
Data about associations of TG, TC, LDL-C and HDL-C with clinicopathological characteristics were shown in Table 2 & 3. In ESCC patients, TC (for T stage, p = 0.006; for TNM stage, p = 0.039) and LDL-C (for T stage, p = 0.031; for TNM stage, p = 0.035) decreased with both advanced T stage and TNM stage while TG and HDL-C showed no such trends. All the lipids exhibited negative association with differential grade and N stage.
By adopting ROC analysis, we obtained the optimal cutoff values of 4.79 mmol/L for TC and 3.23 mmol/L for LDL-C (TG and HDL-C showed negative results, so the data was not given). Relied on the cutoff values, 131 cases (61.2%) were assigned into low TC group and 83 (38.8%) cases were assigned into high TC group. 134 cases (62.6%) were included in low LDL-C group while 80 cases (37.4%) were included in high LDL-C group. In Table 3, low TC was only significantly associated with gender male (p = 0.006). Differences were discernable in regard to T stage (p = 0.044) and TNM stage (p = 0.023) between two groups of LDL-C.

Survival and evaluation of the prognostic value of TC, LDL-C and LHR
In our research, the median DFS was 29.8 months and 5-year DFS was 32.7%. The median OS was 38.8 months and 5-year OS was 36.4%. Kaplan-Meier curves of DFS and OS based on TC and LDL-C were given in Figure 1 (TG and HDL-C had no relationship with survival in our analysis, so the data was not shown ). The curves implied that patients with high TC and LDL-C obviously got a trend towards superior DFS (for TC, p = 0.045; for LDL-C, p < 0.001) and OS (for TC, p = 0.043; for LDL-C, p < 0.001). Considering serum lipids were related to nutritional condition, we analyzed DFS and OS    (Table  4). Patients with lower TC and LDL-C showed inferior DFS (for TC, p = 0.034; for LDL-C, p < 0.001) and OS (for TC, p = 0.024; for LDL-C, p < 0.001). LHR was analyzed in this study and its optimal cutoff value was determined as 2.12. Kaplan-Meier curves of DFS and OS based on LHR was given in Figure  1. Patients with low LHR showed poor DFS and OS (both p < 0.001). To evaluate whether the relationship between LHR and survival was merely affected by elevated LDL-C, patients were stratified according to their LDL-C levels ( Table 5). In patients with no elevations in LDL-C, those who had lower LHR tended to show inferior DFS (p = 0.036) and OS (p = 0.044). In patients with elevated LDL-C, LHR showed no association with DFS (p = 0.071) and OS (p = 0.065).

Subgroup analysis by postoperative adjuvant treatment
For ESCC patients without postoperative adjuvant treatment, Kaplan-Meier analysis implied that LDL-C and LHR were both prognostic predictive factors for DFS (for LDL-C, p = 0.002; for LHR, p < 0.001) and OS (for LDL-C, p = 0.003; for LHR, p < 0.001) (Figure 2). LDL-C and LHR were not associated with DFS and OS when patients received postoperative adjuvant radiotherapy or chemotherapy (all p > 0.05) (the figures were not shown). However, DFS (for LDL-C, p = 0.001; for LHR, p = 0.002) and OS (for LDL-C, p = 0.001; for LHR, p = 0.005) differed significantly in patients who adopted postoperative adjuvant chemoradiotherapy according to LDL-C or LHR (Figure 3).

Subgroup analysis by pathological stage
In the analysis of DFS (Figure 4), patients with high LDL-C had longer DFS in stage III (p = 0.023) group. No relationship was observed between LDL-C and DFS in stage I (p = 0.174) and II (p = 0.053) groups. High LHR had significant relationship with better DFS in patients with stage I (p = 0.005) and II (p = 0.003), but no positive outcome was found in stage III (p = 0.573) group. OS of stage III (p = 0.020) patients with high LDL-C was superior. In patients with stage I (p = 0.157) and II (p = 0.055) tumors, LDL-C had no relationship with OS ( Figure 5). OS of patients with low LHR was significantly

The predictive effects of LDL-C and LHR for 1-year and 3-year OS
In this analysis, 1-year and 3-year OS were analyzed respectively ( Figure 6). Patients with low LDL-C had inferior 1-year (p = 0.002) and 3-year (p < 0.001) OS. Statistical relationship between low LHR and poor OS occurred in 3-year OS (p = 0.002) analysis, but not 1-year OS (p = 0.359).

DISCUSSION
In the present research, we evaluated the prognostic predictive values of preoperative serum lipids and the LDL-C/HDL-C-based LHR in ESCC patients who adopted esophageal resection. We found LDL-C and LHR were independent indicators for DFS and OS. Lower TC, LDL-C and LHR level could predict both poorer DFS and OS, especially the LDL-C and LHR. LDL-C and LHR owned different values for patients with different TNM stage and different survival time. These indicated that LDL-C and LHR could provide effective and important prognostic information in ESCC patients who underwent esophagectomy.
Cholesterol, as one of the structural component of cell membrane, is involved in some signaling pathways that are essential to malignant transformation [14]. LDL-C and HDL-C are key lipoproteins in the transportation of cholesterol. Studies had explored the relationship of cholesterol and LDL-C with cancer. It is reported that people with lower LDL-C level were more likely to generate tumor and most cancer-bearers got low LDL-C and TC levels [11,15]. Saito et al suggested that lower LDL-C level occurred in liver cancer patients, and mortality was the highest in the lowest TC group [12]. Other researches also implied that low LDL-C and cholesterol were accompanied by obviously elevated cancer risk and mortality [16][17][18][19]. In the present study, we found the median DFS and OS were 18.3 and 25.4 months in low LDL-C group while 50.0 and 59.8 months in high LDL-C group (p < 0.001). The median DFS and OS were 18.8 and 31.4 months in low TC group, compared with 38.2 and 53.5 months in high TC group (for DFS, p = 0.031; for OS, p = 0.016). The mortality of patients with low LDL-C was 70.1%, which was 1.44 times higher than that in high LDL-C group (p = 0.002). Studies [8,20] found that TC and LDL-C levels were obviously lower in patients with esophageal, gastric or breast cancer, even in early stage cancer-bearers. This suggested that the lipid changes might not be the products of reduced food intake or deteriorative nutritional status. We analyzed TC and LDL-C in patients with normal serum albumin and found the positive results. Previous researches also supported our results. Firstly, tumor-associated hypocholesterolemia had been hypothesized as elevated LDL-C uptake in tumor cells, and TC and LDL-C would reduce because of the tumor consumption [8,10]. As Y. Tomiki reported, we also found negative outcomes for HDL-C and TG implying low TC might stem from low LDL-C [8]. This might explain why TC exhibited slight statistical relationship with DFS and OS in our analyses. High pretreatment TC and LDL-C levels might represent low LDL-C uptake of tumor cells meaning low tumor metabolic activity. Secondly, increased LDL-C could enhance human immunological response and promote clearance of pathogen as reported [21]. Thirdly, other reports confirmed that oxidized low-density lipoprotein (oxLDL-C) had cytotoxicity on different cell lines and could activate apoptosis of esophageal cancer cells [22][23][24]. Fourthly, recent studies suggested that LDL-C could inhibit angiogenesis by disrupting hypoxia inducible factor (HIF) pathway and hindering endothelial cell migration [25,26]. Angiogenesis could promote tumor development and showed association with tumor necrosis, which was a hallmark of poor prognosis [27][28][29]. Besides, a cohort suggested that higher LDL-C was associated with a lower mortality of all causes [30].
In previous studies [5,31], LHR acted as a prognostic factor in colorectal cancer patients, providing different information from LDL-C. Our results indicated that ESCC patients with low LHR had poor survival outcomes. The median DFS in low and high LHR groups were 18.4 and 47.4 months (p < 0.001). The median OS of these two groups were 25.6 and 59.7 months (p < 0.001).
To confirm whether LHR was merely affected by LDL-C or not, we stratified patients according to elevation ( ≥3.4mmol/L) of LDL-C. In patients with elevated LDL-C, the LHR had no relationship with DFS and OS. But DFS and OS showed statistical difference between two LHR groups in patients with no elevations in LDL-C. The recurrence rate was 54.6% in low LHR group, which was 1.38 times higher than that in high LHR group (p = 0.030). The recurrence rate had no statistical difference according to LDL-C. These suggested LHR could provide much prognostic information, even when preoperative LDL-C was normal. The mechanisms might be involved in the relation of dysfunctional lipid metabolism and cancer development [6,8,9].
Considerding the effects of adjuvant treatment on prognosis, patients were stratified according to different adjuvant treatment modalities. We observed that LDL-C and LHR were correlated to DFS and OS when patients had no adjuvant treatment or received adjuvant chemoradiotherapy. The findings confirmed the prognostic predictive values of LDL-C and LHR both in patients with or without adjuvant treatment after esophageal resection.
Our subgroup analysis discussed the predictive functions of LDL-C and LHR in patients with different TNM stages and survival time. In Kaplan-Meier analysis, LDL-C level had statistical relation with DFS and OS in patients with stage III tumors. DFS and OS showed difference between two LHR groups in patients with stage I and II tumors. These suggested that LDL-C was more valuable for stage III patients, while LHR was better in stage I and II patients. LDL-C had close relationship with 1-year and 3-year OS, but no difference was found in two LHR groups regarding 1-year OS. The area under the ROC curve (AUC) was 0.636 (95% CI, 0.561-0.710; p = 0.001) for LDL-C and 0.622 (95% CI, 0.547-0.697; p = 0.002) for LHR in 3-year OS. As to 5-year OS, the AUC was 0.606 (95% CI, 0.528-0.685; p = 0.009) for LDL-C and 0.618 (95% CI, 0.541-0.695; p = 0.004) for LHR. These results implied that LHR had no prognostic effect on short-term survival, and LDL-C and LHR were associated with long-term survival.
There were some limitations in our research. First, it was a retrospective cohort with a small sample. Second, we did not figure out the causal relationship of low serum lipids and cancer development. It is hard to say whether low TC and LDL-C can promote cancer development or is the result of advanced cancer. Third, it was uncertain whether other cutoff values could act as better predictors. Fundamental knowledge is needed to elucidate the association of serum lipids and survival time in ESCC patients, and further basic and clinical researches are required.
In conclusion, low serum TC and LDL-C levels are predictive factors for poor prognosis in ESCC patients who underwent esophagectomy. LHR, as an independent factor, can serve as a novel and promising serum lipidsbased prognostic indicator, especially for patients with earlier pathological stage. Measurements of serum lipids are inexpensive, convenient and routinely tested in clinic, especially before surgery. This will provide potential prognostic information for clinicians to facilitate the personalized treatment.

Patients
During January 2007 to December 2008, 214 patients with esophageal squamous cell carcinoma who received esophagetomy in the Department of Thoracic Surgery, Qilu Hospital of Shandong University, were admitted into our retrospective study. A patient was excluded from the study: 1) if the patient had more than one primary tumors; 2) if the patient received neoadjuvant therapy; 3) if the patient had serious metabolic diseases before; 4) if the patient was lost during our follow-up.