Development and validation of a new prognostic immune-inflammatory-nutritional score for predicting outcomes after curative resection for intrahepatic cholangiocarcinoma: A multicenter study

Abstract

Background: Immune function, nutrition status, and inflammation influence tumor initiation and progression. This was a retrospective multicenter cohort study that investigated the prognostic value and clinical relevance of immune-, inflammatory-, and nutritional-related biomarkers to develop a novel prognostic immune-inflammatory-nutritional score (PIIN score) for patients with intrahepatic cholangiocarcinoma (ICC).

Methods: The clinical data of 571 patients (406 in the training set and 165 in the validation set) were collected from four large hepato-pancreatico-biliary centers of patients with ICC who underwent surgical resection between January 2011 and September 2017. Twelve blood biomarkers were collected to develop the PIIN score using the LASSO Cox regression model. The predictive value was further assessed using validation datasets. Afterward, nomograms combining the PIIN score and other clinicopathological parameters were developed and validated based on the calibration curve, time-dependent AUC curves, and decision curve analysis (DCA). The primary outcomes evaluated were overall survival (OS) and recurrence-free survival (RFS) from the day of primary resection of ICC.

Results: Based on the albumin-bilirubin (ALBI) grade, neutrophil- to- lymphocyte ratio (NLR), prognostic nutritional index (PNI), and systemic immune- inflammation index (SII) biomarkers, the PIIN score that classified patients into high-risk and low-risk groups could be calculated. Patients with high-risk scores had shorter OS (training set, p < 0.001; validation set, p = 0.003) and RFS (training set, p < 0.001; validation set, p = 0.002) than patients with low-risk scores. The high PIIN score was also associated with larger tumors (≥5 cm), lymph node metastasis (N1 stage), multiple tumors, and high tumor grade or TNM (tumor (T), nodes (N), and metastases (M)) stage. Furthermore, the high PIIN score was a significant independent prognostic factor of OS and RFS in both the training (p < 0.001) and validation (p = 0.003) cohorts, respectively. A PIIN-nomogram for individualized prognostic prediction was constructed by integrating the PIIN score with the clinicopathological variables that yielded better predictive performance than the TNM stage.

Conclusion: The PIIN score, a novel immune-inflammatory-nutritional-related prognostic biomarker, predicts the prognosis in patients with resected ICC and can be a reliable tool for ICC prognosis prediction after surgery. Our study findings provide novel insights into the role of cancer-related immune disorders, inflammation, and malnutrition.

Keywords: immunity; inflammation; intrahepatic cholangiocarcinoma; nomogram; nutrition; prognosis.

Figure 1

Kaplan–Meier curves for overall survival (OS), stratified by (A) ALT, (B) AST, (C) AAPR, (D) AGR, (E) FIB, and (F) ALBI grades and (G) GAR, (H) NLR, (I) PLR, (J) PNI, (K) SII, and (L) CONUT scores in patients with ICC. ICC, intrahepatic cholangiocarcinoma; ALT, alanine aminotransferase; AST, aspartate aminotransferase; AAPR, albumin–alkaline phosphatase ratio; AGR, albumin–globulin ratio; ALBI, albumin–bilirubin grade; FIB, fibrinogen; GAR, GGT–albumin ratio; NLR, neutrophil- to- lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; PNI, prognostic nutritional index; SII, systemic immune inflammation index; CONUT, controlling nutritional status.

Figure 2

Kaplan–Meier curves for recurrence-free survival (RFS), stratified by (A) ALT, (B) AST, (C) AAPR, (D) AGR, (E) FIB, and (F) ALBI grades and (G) GAR, (H) NLR, (I) PLR, (J) PNI, (K) SII, and (L) PLR in patients with ICC.

Figure 3

Construction of the PIIN score using the LASSO Cox regression model. (A) Heatmap of the correlations of the immune–inflammatory–nutritional- related biomarkers. (B) Forest plot of the univariate Cox regression analysis for OS. (C) Partial likelihood deviance for LASSO coefficient profiles. The red dots represent the partial likelihood values, the grey lines represent the standard error (SE), and the vertical dotted line shows the optimal values by 1-s.e. (D) Least absolute shrinkage and selection operator (LASSO) coefficient profiles of 12 immune–inflammatory–nutritional- related biomarkers. (E-L) Differential analysis of the distribution of the PIIN scores in different clinicopathologic features. A comparison between the two groups was performed using the Wilcoxon test. Three group comparisons were performed using the Kruskal–Wallis test. *P< 0.05; **P< 0.01; ***P< 0.001; ns not significant. The ROC curves for predicting OS at 1-, 3-, and 5 years in the training set (M) and the validation set (N). ROC, receiver operating characteristic.

Figure 4

Prognostic implications of the PIIN score. Kaplan–Meier curves of OS (A) and RFS (B) for patients in the low- and high-risk groups according to the PIIN score in the training set. Kaplan–Meier curves of OS (C) and RFS (D) for patients in the low- and high-risk groups according to the PIIN score in the validation set. Forest plot of multivariable Cox regression analysis of OS (E) and RFS (F) in the training set. Forest plot of multivariable Cox regression analysis of OS (G) and RFS (H) in the validation set.

Figure 5

Construction and validation of the nomograms. Nomograms incorporating the PIIN score and other clinicopathological parameters for OS (A) and RFS (B) prediction in the training cohort. (C) Time-dependent AUC curves of the PIIN-nomogram and the AJCC-TNM staging system for the prediction of OS in the training and validation sets. (D) Time-dependent AUC curves of the PIIN-nomogram and the AJCC-TNM staging system for predicting RFS in the training and validation sets. AUC, area under the curve; AJCC-TNM, American Joint Committee on Cancer tumor–node–metastasis.

Figure 6

Calibration curves. The calibration curves of the nomograms between predicted and observed 1-, 3-, and 5-year OS of patients in the training set (A–C) and the validation set (G–I). The calibration curves of the nomograms between predicted and observed 1-, 3-, and 5-year RFS in the training set (D–F) and the validation set (J–L). The dashed line of 45° represents the perfect prediction of the nomogram.

Figure 7

DCA of OS and RFS prediction by the nomograms. The DCA of the nomogram and AJCC-TNM stage for 1-year OS (A), 3-year OS (B), and 5-year OS (C) and for 1-year RFS (D), 3-year RFS (E), and 5-year RFS (F) in the training set. DCA of the nomogram and AJCC-TNM stage for 1-year OS (G), 3-year OS (H), and 5-year OS (I) and for 1-year RFS (J), 3-year RFS (K), and 5-year RFS (L) in the validation set. DCA, decision curve analysis; AJCC-TNM, American Joint Committee on Cancer tumor–node–metastasis.