Neutrophil-albumin ratio and multi-phase computed tomography for lymph node metastasis in pancreatic cancer

Abstract

BACKGROUND

Reliable preoperative detection of lymph node metastasis (LNM) in pancreatic cancer remains elusive: Conventional computed tomography (CT) underestimates micrometastases, and carbohydrate antigen 19-9 is hampered by low specificity. The neutrophil-albumin ratio (NAR) simultaneously reflects systemic inflammation and nutritional depletion, but its contribution to LNM prediction in pancreatic cancer is unexplored. We hypothesised that integrating NAR with multi-phase CT findings would significantly improve the accuracy of preoperative LNM assessment in patients undergoing curative-intent resection.

AIM

To determine whether preoperative NAR plus multi-phase CT reliably predicts nodal metastasis in pancreatic cancer.

METHODS

In this single-centre retrospective cohort study (February 2022 to February 2025, Ordos Central Hospital, China), 129 consecutive patients undergoing curative pancreatic resection were histologically classified as LNM⁺ (n = 61) and LNM^- (n = 68). Preoperative NAR and platelet-albumin ratio (PAR) were calculated; optimal cut-offs were determined with X-tile. Multi-phase CT images were re-reviewed by two blinded radiologists. Independent predictors of nodal metastasis were identified by multivariate logistic regression, and model performance was evaluated with receiver operating characteristic (ROC) analysis.

RESULTS

Between the two cohorts, univariate comparison revealed significant divergence in age, tumour diameter, concomitant hemangioma thrombosis, PAR, NAR, and CT-detected nodal status (P < 0.05). Subsequent multivariate modelling identified hemangioma thrombosis, PAR above 6.35, NAR exceeding 0.13, and radiologically positive lymph nodes as independent predictors of nodal metastasis (P < 0.05). ROC evaluation indicated that the NAR-plus-CT-nodes model (model 1) reached an area under the curve (AUC) of 0.758, whereas the four-variable composite (model 3) achieved the best performance with an AUC of 0.830 (95%CI: 0.753-0.890), sensitivity 83.61%, and specificity 67.65%.

CONCLUSION

The model 3 (NAR > 0.13, PAR > 6.35, CT nodal positivity, hemangioma thrombosis) provides robust, clinically actionable preoperative identification of pancreatic cancer patients at high risk of LNM.

Keywords: Neutrophil-albumin ratio; Multiphase enhanced computed tomography image; Pancreatic cancer; Lymph node metastasis; Predictive model

Core Tip: For the first time, we integrated systemic inflammation-nutrition indices [neutrophil-albumin ratio (NAR), platelet-albumin ratio (PAR)] with multi-phase computed tomography (CT) radiomics to predict lymph node metastasis of pancreatic cancer. The resulting four-marker model (NAR > 0.13, PAR > 6.35, hemangioma thrombosis, CT-positive nodes) achieved an area under the curve of 0.830, surpassing conventional imaging and offering a cost-free, pre-operative tool for personalized surgical planning and neoadjuvant selection.