We screened 2,490 patients diagnosed with HCC between January 2010 and June 2017 at Beijing Ditan Hospital, Capital Medical University. Patients aged between 18 and 75 years diagnosed with HBV-related HCC were recruited for this study. The exclusion criteria were as follows: (1) age <18 or >75 years, (2) presence of other types of tumors or liver transplantation, (3) other viral infections or human immunodeficiency virus infection, (4) incomplete clinical data, and (5) lost of follow-up within 1 year. As per these criteria, 1,639 patients were finally enrolled in the study. We also included 373 patients as an internal validation cohort between July 2017 and October 2019 (Figure 1). This study followed the ethical principles of the Declaration of Helsinki and approval was obtained from the Ethical Review Committee of the Beijing Ditan Hospital.

Chronic hepatitis B was defined as HBsAg positivity for >6 months (21). The diagnosis of cirrhosis was based on evidence from liver biopsy, endoscopy, ultrasound, or elastography, and/or signs of complications associated with portal hypertension (22). HCC diagnosis was as per the criteria of the Asia-Pacific clinical guidelines (23). Every 3 months, routine laboratory tests [including routine blood examination, liver, renal, coagulation function tests, HBV DNA, and alpha-fetoprotein (AFP)] and radiological examination, such as computed tomography (CT), magnetic resonance imaging (MRI), or ultrasound were performed. The outcome was the occurrence of mortality within 3 years or at the end of the 3 years follow-up period.

Baseline demographic characteristics and laboratory data, including age, sex, hypertension, diabetes, smoking history, alcohol consumption history, complications, liver function, renal function, coagulation function, serum lipid level, and AFP level, were collected from electronic medical records at enrollment. In addition, tumor characteristics, such as tumor number, size, vascular invasion, and tumor metastasis, were recorded based on the imaging data at baseline. NHR was calculated as the neutrophil count divided by the HDL-C value, while MHR as the monocyte count divided by the HDL-C value, and LHR as the lymphocyte count divided by the HDL-C value. The model for end-stage liver disease (MELD) was used to estimate the severity of the liver disease (24).

SPSS (version 25.0; SPSS, Inc., Chicago, IL, United States) and R (version 3.6.3; The R Foundation, Vienna, Austria) software were used for the statistical analyses. Continuous variables were reported as mean ± standard deviation or median with interquartile range (IQR), while categorical variables were reported as frequency (percentage). Continuous variables were compared using Student’s t-test or the Mann–Whitney test; the chi–squared test or Fisher’s exact test was used for two groups, as appropriate. Univariate and multivariate Cox regression analyses were used to assess the association between the HDL-C, NHR, MHR, LHR (continuous and tertile), and mortality. Results were considered statistically significant at p -value <0.05.

The predictive value of lipid-related inflammatory markers was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). The prognostic power of these indicators was compared with that of the MELD score at 1, 2, and 3 years using the Delong test (25). The association between the NHR and 3-year mortality were evaluated on a continuous scale using a restricted cubic spline (RCS) with four knots at the 5th, 35th, 65th, and 95th percentiles and to test for nonlinearity (26). The optimal cut-off values were determined for the NHR and MELD scores for mortality using the X-tile software (Yale University School of Medicine, New Haven, CT, United States). Differences in survival rates among the groups were analyzed using Kaplan–Meier curves and compared using log-rank tests.

A total of 1,639 and 373 patients in the training and validation cohorts, respectively, were included in the analysis. Baseline characteristics and laboratory data of the patients are shown in Table 1. The median age of the training cohort was 57.0 years (IQR, 50.0–63.0), with male predominance (n = 1,563, 77.6%). Of those patients, 181 (9.0%) patients underwent liver resection, 1,358 (67.5%) underwent minimally invasive therapy, whereas 473 (23.5%) received palliative therapy. Of the 1,639 patients in the training cohort, 1,563 patients (77.6%) were male and 1,306 patients (79.6%) were diagnosed with cirrhosis. During the 3-year follow-up period, 666 patients (40.6%) and 138 patients (37.0%) died in the training and validation cohorts, respectively. Overall, the patients in the two cohorts were similar when their baseline characteristics were considered.

Furthermore, we compared the survival and death characteristics of patients in the training cohort (Table 2). Patients who died were older, had a higher proportion of diabetes, tumor size ≥5 cm, AFP ≥ 400 ng/mL, higher total bilirubin (TBIL), γ-glutamyl transferase (GGT), creatinine (Cr), and international prothrombin ratio (INR), and lower albumin levels (all p  < 0.001) than those who survived. Regarding inflammation and lipid-related markers, dead patients had higher levels of neutrophils, monocytes, NHR, MHR, and LHR, and lower levels of lymphocytes, total cholesterol (TC), and HDL-C compared with the patients who survived.

Univariate analysis showed that low HDL-C, high NHR, MHR, and LHR levels significantly increased the risk of 3-year mortality as both continuous and categorical variables (all p < 0.001; Table 3). In addition, univariate analysis indicated that age, sex, diabetes, alcohol consumption, alanine aminotransferase, aspartate aminotransferase, platelet count, alpha-fetoprotein, total cholesterol, Child-Pugh class, MELD score, Barcelona Clinic Liver Cancer (BCLC) stage, tumor size, and type of treatment were significantly associated with the 3-year mortality (all p < 0.05). These significant factors were included in the multivariate Cox regression analysis. After adjustment for confounding variables, these significant associations were found with low HDL-C (aHR, 0.31; 95% CI: 0.23–0.41, p < 0.001) and high NHR levels (aHR, 1.02; 95% CI: 1.01–1.03, p < 0.001) as continuous variables. However, the association of 3-year mortality with LHR and MHR was attenuated.

Figures 2A–C show the 1-, 2-, and 3-year prognostic values of HDL-C, NHR, MHR, and LHR. Moreover, the performance of these markers was compared with that of the MELD score, a well-established prognosis score. In the training cohort, the AUCs of the NHR for mortality were 0.777 (95% CI 0.752–0.802), 0.760 (95% CI 0.722–0.784), and 0.740 (95% CI 0.715–0.767) at 1, 2, and 3 years, respectively. The NHR showed a similar predictive ability, compared to the MELD score (AUCs 1, 2, and 3 years: 0.786, 0.770, and 0.761, respectively). In addition, we compared the NHR with other lipid-related indicators, including HDL-C, MHR, and LHR. The results indicated that the AUC of the NHR was significantly higher than those of the HDL-C, MHR, and LHR at 1, 2, and 3 years (all p  < 0.05).

As shown in Figures 2D,E, the risk was relatively low in the low NHR range and then increased in patients with HCC. These results indicate that the NHR was associated with the 1-and 3-year mortality risk and the test for nonlinearity was statistically significant (p < 0.001).

The optimal cut-off points for the NHR and MELD scores were determined using the X-tile software. When NHR ≥ 3.5 and MELD ≥ 9, the difference was the most statistically significant. The sensitivity and specificity of the NHR ≥ 3.5 were 70.1 and 74.7%, respectively, and those of the MELD ≥ 9 were 76.3 and 67.4%, respectively. Scatterplots were used to visualize the relationship between NHR, MELD score, and mortality in patients with HCC (Figure 2F). The scatterplots revealed that patients with NHR ≥ 3.5 and MELD ≥ 9 had poor outcomes in patients with HCC.

The Kaplan–Meier survival curves based on the NHR and MELD optimal cut-off values are shown in Figure 3. We observed a statistical difference between cut-off values of these two biomarkers and survival probability. The 3-year survival probability was 73.7% in patients with NHR < 3.5, whereas those with NHR ≥ 3.5 were 37.0% ( p  < 0.0001; Figure 3A). Furthermore, patients with MELD score < 9 had a significantly higher survival probability than those with MELD score ≥ 9 (77.3 vs. 39.0%, p  < 0.0001; Figure 3B). Next, all patients were divided into three groups: low- (NHR < 3.5 and MELD <9, n = 628), medium- (NHR ≥ 3.5 or MELD ≥9, n = 615), and high-risk groups (NHR ≥ 3.5 and MELD ≥9, n = 396). The 3-year survival probabilities were 81.8, 62.1, and 19.4% in the low-, medium-, and high-risk groups, respectively ( p  < 0.0001; Figure 3C).

To examine the risk stratification for different ages, we divided patients into the following three subgroups: <40, 40–60, and >60 years. The Kaplan–Meier curve analysis showed that the 3-year survival probability in patients with a NHR < 3.5 and a MELD score < 9 was significantly higher than in patients with a NHR ≥ 3.5 and a MELD score ≥ 9, regardless of patient’s ages (all p  < 0.0001; Figures 3D–F).

For the diabetic subgroup, the NHR had the highest AUC at 3-year (0.735; 95% CI: 0.681–0.779). The AUCs of the HDL-C, MHR, LHR, and MELD scores were 0.631 (95% CI: 0.573–0.688), 0.617 (95% CI: 0.558–0.675), 0.538 (95% CI: 0.474–0.592), and 0.753 (95% CI: 0.704–0.805), respectively. The NHR showed better performance than HDL-C, MHR, and LHR (all p  < 0.05, Figure 4A). The 3-year survival probabilities of patients in the low-, medium-, and high-risk groups were 74.2, 54.8, and 18.4%, respectively ( p  < 0.0001; Figure 4B).

In the validation cohort, baseline NHR offered good prognostic value for mortality with the AUC at 1, 2, and 3 years: 0.751 (95% CI: 0.696–0.807), 0.739 (95% CI: 0.695–0.785), and 0.734 (95% CI: 0.684–0.777), respectively. In addition, the NHR showed a performance similar to that of the MELD score (0.756; 95% CI 0.706–0.807) and was noted to have the highest AUC (0.734), which was followed by AUCs of the HDL-C (0.606; 95% CI 0.517–0.640), MHR (0.605; 95% CI 0.542–0.667), and LHR (0.500; 95% CI 0.438–0.561) at 3 years (all p  < 0.05; Figures 5A–C).

Furthermore, patients with a NHR < 3.5 had a significantly higher survival probability than those with a NHR ≥ 3.5 (72.5 vs. 44.4%, p  < 0.0001; Figure 5D). Patients with a MELD score ≥ 9 were associated with an increased risk of death compared with those with a MELD score < 9 ( p  < 0.0001; Figure 5E). The 3-year survival probabilities in patients in the low- (NHR < 3.5 and MELD <9, n = 143), medium- (NHR ≥ 3.5 or MELD ≥9, n = 161), and high-risk groups (NHR ≥ 3.5 and MELD ≥9, n = 69) were 84.6, 59.0, and 27.5%, respectively (p < 0.001; Figure 5F).