Albumin-to-alkaline phosphatase ratio serves as a prognostic indicator in unresectable pancreatic ductal adenocarcinoma: a propensity score matching analysis

Background Recent evidence suggests that albumin-to-Alkaline Phosphatase Ratio (AAPR) functions as a novel prognostic marker in several malignancies. However, whether it can predict the prognosis of unresectable pancreatic ductal adenocarcinoma (PDAC) remains unclear. Herein, we seek to investigate this possibility by a propensity score matching (PSM) analysis. Methods This was a retrospective cohort study in which 419 patients diagnosed with unresectable PDAC and receiving chemotherapy were recruited. Patients were stratified based on the cutoff value of AAPR. The PSM analysis was performed to identify 156 well-balanced patients in each group for overall survival (OS) comparison and subgroup analysis. Univariate and multivariate analyses were carried out to examine the potential of AAPR to indicate the prognosis of unresectable PDAC. The prediction performance of conventional model and combined model including AAPR was compared using the Akaike Information Criterion (AIC) and concordance index (C-index). Results We identified an AAPR of 0.4 to be the optimal cutoff for OS prediction. Patients with AAPR≤0.4 had significantly shorter OS compared with patients with AAPR> 0.4 (6.4 versus 9.3 months; P < 0.001). Based on the PSM cohort and entire cohort, multivariate Cox analysis revealed that high pretreatment for AAPR was an independent marker predicting favorable survival in unresectable PDAC (hazard ratio, 0.556; 95% confidence interval, 0.408 to 0.757; P < 0.001). Significant differences in OS were observed in all subgroups except for the group of patients age ≤ 60. Combined prognostic model including AAPR had lower AIC and higher C-index than conventional prognostic model. Conclusions Pretreatment AAPR servers as an independent prognostic indicator for patients with unresectable PDAC. Inclusion of AAPR improved the prediction performance of conventional prognostic model, potentially helping clinicians to identify patients at high risk and guide individualized treatment.