AGLR is a novel index for the prognosis of hepatocellular carcinoma patients: a retrospective study

Abstract

Background: Most hepatocellular carcinoma (HCC) patients' liver function indexes are abnormal. We aimed to investigate the relationship between (alkaline phosphatase + gamma-glutamyl transpeptidase)/lymphocyte ratio (AGLR) and the progression as well as the prognosis of HCC.

Methods: A total of 495 HCC patients undergoing radical hepatectomy were retrospectively analyzed. We randomly divided these patients into the training cohort (n = 248) and the validation cohort (n = 247). In the training cohort, receiver operating characteristic (ROC) curve was used to determine the optimal cut-off value of AGLR for predicting postoperative survival of HCC patients, and the predictive value of AGLR was evaluated by concordance index (C-index). Further analysis of clinical and biochemical data of patients and the correlation analysis between AGLR and other clinicopathological factors were finished. Univariate and multivariate analyses were performed to identify prognostic factors for HCC patients. Survival curves were analyzed using the Kaplan-Meier method.

Results: According to the ROC curve analysis, the optimal predictive cut-off value of AGLR was 90. The C-index of AGLR was 0.637 in the training cohort and 0.654 in the validation cohort, respectively. Based on this value, the HCC patients were divided into the low-AGLR group (AGLR ≤ 90) and the high-AGLR group (AGLR > 90). Preoperative AGLR level was positively correlated with alpha-fetoprotein (AFP), tumor size, tumor-node-metastasis (TNM) stage, and microvascular invasion (MVI) (all p < 0.05). In the training and validation cohorts, patients with AGLR > 90 had significantly shorter OS than patients with AGLR ≤ 90 (p < 0.001). Univariate and multivariate analyses of the training cohort (HR, 1.79; 95% CI 1.21-2.69; p < 0.001) and validation cohort (HR, 1.82; 95% CI 1.35-2.57; p < 0.001) had identified AGLR as an independent prognostic factor. A new prognostic scoring model was established based on the independent predictors determined in multivariate analysis.

Conclusions: The elevated preoperative AGLR level indicated poor prognosis for patients with HCC; the novel prognostic scoring model had favorable predictive capability for postoperative prognosis of HCC patients, which may bring convenience for clinical management.

Keywords: AGLR; Biomarker; Hepatocellular carcinoma; Prognosis.

Fig. 1

AGLR's predictive capability and its comparison with AFP, NLR and ALBI grade. ROC of AGLR in the training cohort (a) and the relationships between AGLR level and tumor size, TNM stage and MVI in the training cohort (b)

Fig. 2

ROC of AGLR in the validation cohort (a) and the relationships between AGLR level and tumor size, TNM stage and MVI in the validation cohort (b)

Fig. 3

Prognostic significance of AGLR in patients with HCC. Kaplan–Meier analysis of survival in the training cohort. High AGLR level was closely associated with a worse prognosis. The green line represents AGLR level > 90, whereas the blue line represents AGLR level ≤ 90. Kaplan–Meier curves depict OS (a) and DFS (b) in HCC patients with AGLR > 90 or ≤ 90

Fig. 4

Kaplan–Meier analysis of HCC patients’ survival in the validation cohort. High AGLR level was closely associated with a worse prognosis. Kaplan–Meier curves depict OS (a) and DFS (b) in HCC patients with AGLR > 90 or ≤ 90

Fig. 5

In the training cohort, comparison of prognostic effects of different scoring groups, there was no statistical significance of survival between patients with a score of 2 vs. 3, for both DFS (a) (p = 0.173) and OS (b) (p = 0.126). Kaplan–Meier analysis of survival for different risks groups. There were statistical significance between the low-, medium- and high-risk groups (all p < 0.001) for both DFS (c) and OS (d)

Fig. 6

In the validation cohort, DFS’s patients with a score of 2 vs. 3 (p = 0.195) (a), OS’s patients with a score of 2 vs. 3 (p = 0.111) and 4 vs. 5 (p = 0.068) (b) had no statistical significance. There were statistical significance between the low-, medium- and high-risk groups (all p < 0.001) for both DFS (c) and OS (d)