Verification of inflammation‐based prognostic marker as a prognostic indicator in hepatocellular carcinoma

Abstract Aim Although inflammation‐based markers in cancer have been used for prognostic prediction, the most useful marker for hepatocellular carcinoma (HCC) has not been established. We investigated the usefulness of various inflammation‐based markers in HCC patients after hepatectomy. Methods A total of 478 patients who underwent initial hepatectomy for HCC from 2009 to 2015 and were diagnosed with pathological HCC were included in this retrospective study. Inflammation‐based markers consisted of the C‐reactive protein to albumin ratio (CAR), Glasgow prognostic score (GPS), neutrophil to lymphocyte ratio, lymphocyte to monocyte ratio, platelet to lymphocyte ratio and prognostic index. Univariate and multivariate analyses for overall survival (OS) and disease‐free survival (DFS) using the Cox proportional hazard model were carried out. Kaplan‐Meier analysis and log‐rank test were used for comparison of OS and DFS. To reduce influences of selection bias and confounders for stratifying CAR, clinicopathological characteristics of patients were balanced by propensity score matching. Results Multivariate analysis identified only high CAR (>0.027) as an indicator of poor OS, and high CAR and high GPS (1‐2) as indicators of poor DFS among inflammation‐based markers. After propensity score matching, 124 patients each with low CAR and high CAR were matched. High CAR was correlated with both poor OS and DFS. Conclusion C‐reactive protein to albumin ratio was the most valuable prognostic indicator after hepatectomy for HCC among inflammation‐based markers.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC) is the fifth most common cancer and the second leading cause of cancer-related death globally. 1,2 Liver transplantation is an effective treatment, but the limited number of donor livers prevents widespread use of this oncological therapy. 3 Surgical resection is hence still the most effective method for the treatment of HCC. However, the high recurrence rate after curative resection indicates that the prognosis of HCC is still insufficient despite recent progress in treatment. 4 Thus, identification of patients with probable poor prognosis using reliable biomarkers is essential for improving survival outcomes of HCC patients.
Previous studies have shown that various staging systems and serum biomarkers have predicted prognosis using tumor-related factors. [5][6][7] There is increasing evidence that inflammation is crucial in the progression of cancer development and the presence of systemic inflammatory response has been shown to be associated with clinical outcomes in several malignancies. [8][9][10] Recent studies have also proposed markers based on a number of inflammation-based prognostic indicators of HCC. Inflammation-based markers include the C-reactive protein (CRP) to albumin ratio (CAR), 11 Glasgow prognostic score (GPS), 12 neutrophil to lymphocyte ratio (NLR), 13 lymphocyte to monocyte ratio (LMR), 14 platelet to lymphocyte ratio (PLR) 15 and prognostic index (PI). 16 Nevertheless, it remains unclear which inflammation-based marker more accurately predicts the prognosis after hepatectomy for HCC. In the present study, the value of these inflammation-based markers as predictive indicators of prognosis was explored in patients with HCC after hepatectomy.

| Patients
This study was based on a retrospective analysis of 478 HCC patients who underwent hepatectomy at our institution between January 2009 and December 2015. Inclusion criteria were follows: (i) tumor was histologically diagnosed as HCC; (ii) no distant metastasis was detected in the preoperative image; (iii) resection margin was negative; (iv) first hepatectomy for HCC; and (v) no other malignancies.
Baseline clinicopathological findings were retrieved and reviewed from the hospital database. Characteristics of patient cohorts are shown in Table 1

| Treatment and patient follow up
Hepatectomy procedure was determined after evaluating tumor size, number of tumors, tumor location, liver function and patient status. Hepatectomy and liver function were classified according to

| Definition of inflammation-based prognostic systems
Prior to the operation, blood samples were collected. CAR was cal- were assigned a PNI of 1.

| Statistical analysis
Continuous variables were expressed as median and range and com-   Table 3). Among all the inflammation-based markers, high CAR was the only factor negatively influencing both OS and DFS.

| RE SULTS
Kaplan-Meier analyses showing OS and DFS using inflammationbased markers are shown in Figures 2 and 3. As shown in Figure 2, Kaplan-Meier analyses indicated that CAR, GPS, NLR and PI were correlated with OS, whereas LMR and PLR were not. In addition, Figure 3 shows that CAR, GPS, LMR, and PI were correlated with DFS, but NLR and PLR were not.
All baseline clinical characteristics were compared between the low and high CAR groups. Before propensity score matching, baseline  (Table 4).
After propensity score matching, all baseline clinical characteristics between the low and high CAR groups were well balanced (Table 4).
In total, 124 of the 301 patients with low CAR and an equal number of the 177 patients with high CAR were matched. As shown in Figure 4, Kaplan-Meier analyses indicated that high CAR was correlated with poor OS and DFS. Usefulness of CAR in predicting the prognosis of HCC has been reported previously. 11,[18][19][20] The present study included only cases in which hepatectomy was carried out for HCC and the diagnosis of HCC was confirmed pathologically. Hence, this study probably had the largest cohort among the papers already published. 11,20 In addition, the results from the propensity score matching can enhance the usefulness of CAR. Similarity of the cut-off values between our study and those reported by Shimizu et al, 11 who have also examined cases of hepatectomy in confirmed HCC, could explain the F I G U R E 2 Relationship between the two groups of overall survival of different inflammation-based markers. CAR, C-reactive protein to albumin ratio; GPS, Glasgow prognostic score; LMR, lymphocyte to monocyte ratio; NLR, neutrophil to lymphocyte ratio; PI, prognostic index; PLR, platelet to lymphocyte ratio F I G U R E 3 Relationship between the two groups of disease-free survival of different inflammation-based markers. CAR, C-reactive protein to albumin ratio; GPS, Glasgow prognostic score; LMR, lymphocyte to monocyte ratio; NLR, neutrophil to lymphocyte ratio; PI, prognostic index; PLR, platelet to lymphocyte ratio TA B L E 4 Clinicopathological characteristics of the patients before and after propensity score matching Therefore, CRP level plays a key role in the progression of HCC. 11 With a high CAR, serum albumin level is found to be decreased.

| D ISCUSS I ON
Hypoalbuminemia not only reflects liver dysfunction as a result of the underlying chronic liver disease, but is also associated with a sustained systemic inflammatory response, either from the tumor itself or as a host reaction. Circulating catabolic factors, such as tumor necrosis factor-α and interleukins, mediate the hypoalbuminemia process. 25 As albumin decreases along with the disease status, weight and muscle mass decrease, leading to a decrease in performance status and an increase in mortality. 26 Among the inflammation-based markers, lymphocytes reflect the host immune system's ability to recognize and eliminate tumors. 27 Neutrophils can prompt the secretion of vascular endothelial growth factor (VEGF) and induce tumor growth. 28  There are a few limitations in the present study. First, it was a retrospective study, although propensity score matching was used. As such, selection bias and potential confounders could not be avoided completely. Second, it was a single-center study. Prospective cohort studies in multiple institutions should be carried out to confirm these results.
In conclusion, our study showed that CAR was the most useful prognostic indicator among inflammation-based markers for patients who had undergone hepatectomy for HCC.

ACK N OWLED G M ENT
We would like to thank Editage (www.edita ge.jp) for English language editing.

D I SCLOS U R E
Conflicts of Interest: Authors declare no conflicts of interest for this article.