Measurement of the expression pattern of serum lipids and apolipoproteins in liver disease
We first compared the serum lipid and apolipoprotein levels in the four different groups. In patients with HCC, the serum level of ApoA1 was statistically lower than that in patients with benign liver disease but was higher than that in the liver cirrhosis group. No significant difference was observed between HCC and the hepatitis group. In addition, the expression pattern of HDL was similar to that of ApoA1. The serum ApoE level was significantly higher in the HCC group compared with the chronic liver hepatitis group, whereas the level did not differ significantly between the HCC group and the other two groups. Compared with the liver cirrhosis group, the TG and LDL levels were significantly elevated in the HCC, liver hepatitis and benign liver disease groups, whereas no significant differences were shown in the expression levels among these three groups. Moreover, we did not find any significant differences in the serum ApoB and TC levels among these four groups (Figure 1 and Figure S1).
Correlations between the lipid and apolipoprotein levels and clinical characteristics
As shown in Table 1, no significant differences were found in the relationships between lipid and apolipoprotein levels and clinical background factors such as age, liver cirrhosis, alanine aminotransferase (ALT), albumin (ALB), alpha-fetoprotein (AFP), and serum total bilirubin (TB). However, ApoA1, ApoE and HDL were closely associated with gender. With regard to metabolic profiles, TG and HDL had positive correlations with BMI, whereas LDL and ApoB were associated with glycometabolic indexes such as blood glucose and hemoglobin A1C (HbA1C).TG had a close relationship with HbA1C and the insulin resistance index HOMA2-IR. However, most lipids and apolipoproteins were not closely related to metabolic parameters in HCC patients. Finally, with respect to tumor-related covariates, we revealed that ApoA1 was significantly related to vascular invasion and that LDL was associated with tumor size. However, no significant relationships were observed between most lipids and tumor-related factors.
Prediction of Tumor Recurrence and Patient Survival in the training cohort
We next explored the prognostic significance of lipids and apolipoproteins in patients with HCC. The baseline characteristics of the patients were not significantly different between the training and validation cohorts (Table S1). Using X-tile 3.6.1 software, the optimal cutoff values of each lipid and apolipoprotein were established to separate patients in the training cohort into low and high groups[18]. Divisions of the validation cohort were also based on these numerical standards. A univariate analysis indicated that a lower ApoA1 level was significantly correlated with poor prognosis of HCC patients. The cumulative OS and DFS rates of patients in the low ApoA1 group were significantly lower than those of patients in the high ApoA1 group (5-year OS rate: 65 vs. 78.8%, P=0.02; 5-year DFS rate: 45.5 vs. 59.6%, P=0.01) (Figure 2). In addition, an elevated baseline level of LDL was significantly associated with reduced OS and DFS, while other lipids were not significant predictive factors in the univariate analysis (Figure S2 and S3). A multivariate survival analysis performed with the Cox regression model confirmed ApoA1 as an independent predictor of both OS and DFS (OS: Hazard ratio (HR): 0.587; 95% confidence interval (CI): 0.352-0.979; P=0.04; DFS: Hazard ratio (HR): 0.607; 95% confidence interval (CI): 0.413-0.893; P=0.01). Then, a multivariate Cox regression model was used to further assess the independent prognostic ability of TNM stage. The analysis showed that low ApoA1 remained anindependent unfavorable factor that influenced the OS and DFS (OS: HR: 0.517; 95% CI: 0.309-0.866; P=0.01; DFS: HR: 0.537; 95% CI: 0.363-0.793; P=0.002)(Table 2 and Table S2).
Prognostic value of lipid and apolipoprotein levels in the validation cohort
Lipids and apolipoproteins were further assessed for their prognostic power in an independent cohort of 229 patients. We found that HCC patients with lower serum ApoA1 levels had a significantly worse prognosis than those with higher serum ApoA1 levels. In the univariate analysis of OS, ApoA1 and ApoE were univariate predictors of OS (P=0.01; P=0.02, for ApoA1 and ApoE, respectively)(Figure 2 and Figure S4). A Cox regression analysis demonstrated that a lower serum ApoA1 level was an independent indicator of OS (HR: 0.418; 95% CI: 0.234-0.747; P=0.003). In addition, a higher serum ApoE level was also an independent risk factor for OS (HR:1.897; 95% CI: 1.123-3.204; P=0.02). In the univariate analysis of DFS, a lower ApoA1 level was a significant predictor of poor prognosis (P=0.007), and a Cox regression analysis showed that ApoA1 remained an independent prognostic factor for DFS (HR:0.609; 95% CI: 0.409-0.908; P=0.02)(Table S3 and Table S4). However, other lipids and apolipoproteins including TC, TG, HDL, LDL and ApoB did not show any statistically significant associations with OS and DFS(Figure S4 and S5).
Serum ApoA-1 levels are correlated with early recurrence in the training and validation cohorts
Furthermore, the prognostic significance of the serum ApoA-1 level in patients with AFP levels ≤400 ng/mL and/or in patients with early recurrence was investigated. Among patients with an AFP concentration ≤400 ng/mL, the OS and DFS rates in low ApoA1 patients were significantly lower than those in high ApoA1 patients in the training cohort (OS: 68.3% vs. 83.1%, P=0.006; DFS: 49% vs. 66.7%, P=0.01, respectively, for low ApoA1 and high ApoA1 patients). In accordance with the prognostic value of ApoA1 in the training cohort, ApoA1 remained associated with the OS and DFS rates in the lower AFP subgroup in the validation cohort (OS: 72.2% vs. 86.4%, P=0.025; DFS: 47.4% vs. 62.1%, P=0.04) (Figure S6).
In addition, it is worth noting that, for early recurrence in the training cohort, the serum ApoA-1 level was significantly associated with recurrence in the univariate analysis. In the multivariate analysis, the serum ApoA-1 level was an independent indicator of early recurrence (HR: 0.51; 95% CI: 0.297-0.875; P=0.02). Similar results were confirmed in the validation cohort: lower ApoA-1 was an independent predictor of early recurrence (HR, 0.478; 95% CI: 0.278-0.881; P=0.008)(Figure 3, Table 3 and Table S5). Previous studies have shown that predictive factors for early recurrence are mainly tumor-related clinicopathologic factors (i.e., tumor size, tumor number, tumor markers, microscopic vascular invasion)[19,20]. The results of the multivariate analysis in the present study also show that early recurrence was associated with larger tumor size, vascular invasion, and advanced TNM stage. Further stratification analyses were conducted in the subgroups based on AFP, tumor size, number, vascular invasion, and TNM stage. We found that low ApoA1 was remarkably related to early recurrence in patients with an AFP concentration≤400 μg/L, a tumor size ranging from 2cm to 5cm, TNM stage I and solitary tumor in both the training and validation cohorts. By contrast, no significant associations were observed between the ApoA1 level and the early recurrence rate in the vascular invasion-negative subgroup (Figure S7 and Figure S8).