Su Yongcheng, Xu Beibei, Ma Miaomiao, Zhang Wenqing, Ouyang Zhong, Hu Tianhui
Shenzhen Research Institute of Xiamen University, Shenzhen, China.
Xiamen Key Laboratory for Tumor Metastasis, Cancer Research Center, School of Medicine, Xiamen University, Xiamen, China.
Int J Surg. 2025 Aug 1;111(8):4972-4985. doi: 10.1097/JS9.0000000000002543. Epub 2025 Jun 5.
Breast cancer (BC) remains one of the most prevalent cancers affecting women globally, imposing significant health and economic burdens on both patients and society. This study aims to investigate the relationship between the neutrophil percentage-to-albumin ratio (NPAR) and BC risk and mortality.
Clinical data from 13 540 participants in the NHANES database were analyzed, including 331 individuals with a documented history of BC. Survival analysis and advanced machine learning (ML) techniques were applied to assess the data.
Higher NPAR levels were significantly associated with increased BC risk in the unadjusted model, with quartile comparisons revealing an odds ratio (OR) of 1.51 (95% CI: 0.99-2.29, P = 0.057). After adjustment, the OR increased to 1.70 (95% CI: 1.12-2.57, P < 0.05), indicating the robustness of this association. Elevated NPAR levels were also linked to higher all-cause mortality (ACM). Multivariate Cox regression models showed that a one-unit increase in NPAR was associated with adjusted hazard ratios of 1.09 (95% CI: 1.07-1.12) for overall mortality and 1.17 (95% CI: 1.13-1.22) for cardiovascular disease mortality, both with P values <0.001. Restricted cubic splines analysis revealed a linear correlation between NPAR and BC risk ( P for nonlinearity = 0.15), while a nonlinear relationship was observed for ACM ( P for nonlinearity < 0.01). Among nine ML models evaluated, the LightGBM model exhibited the best diagnostic performance, achieving an area under the receiver operating characteristic curve of 0.995, outperforming models such as CATBoost, Naive Bayes, logistic regression, random forest, K-nearest neighbors, support vector machine, decision tree, and XGBoost. After model selection, an online calculator was built for use in the clinic, and the web-service is available at https://fast.statsape.com/tool/detail?id=11 .
NPAR emerged as a crucial biomarker in BC risk assessment. This study suggests that NPAR may serve as a dual-purpose biomarker for both BC risk evaluation and prognostic assessment, potentially aiding in early screening and personalized treatment strategies.
乳腺癌（BC）仍然是全球影响女性的最常见癌症之一，给患者和社会带来了巨大的健康和经济负担。本研究旨在探讨中性粒细胞百分比与白蛋白比值（NPAR）与BC风险及死亡率之间的关系。
分析了美国国家健康与营养检查调查（NHANES）数据库中13540名参与者的临床数据，其中包括331名有BC病史记录的个体。应用生存分析和先进的机器学习（ML）技术对数据进行评估。
在未调整模型中，较高的NPAR水平与BC风险增加显著相关，四分位数比较显示优势比（OR）为1.51（95%可信区间：0.99 - 2.29，P = 0.057）。调整后，OR增加到1.70（95%可信区间：1.12 - 2.57，P < 0.05），表明这种关联的稳健性。NPAR水平升高还与全因死亡率（ACM）升高有关。多变量Cox回归模型显示，NPAR每增加一个单位，总体死亡率的调整后风险比为1.09（95%可信区间：1.07 - 1.12），心血管疾病死亡率的调整后风险比为1.17（95%可信区间：1.13 - 1.22），P值均<0.001。受限立方样条分析显示NPAR与BC风险之间存在线性相关性（非线性P值 = 0.15），而在ACM方面观察到非线性关系（非线性P值 < 0.01）。在评估的9种ML模型中，LightGBM模型表现出最佳诊断性能，受试者工作特征曲线下面积为0.995，优于CATBoost、朴素贝叶斯、逻辑回归、随机森林、K近邻、支持向量机、决策树和XGBoost等模型。模型选择后，构建了一个在线计算器供临床使用，网络服务可在https://fast.statsape.com/tool/detail?id=11获取。
NPAR成为BC风险评估中的关键生物标志物。本研究表明，NPAR可能作为一种双重用途的生物标志物，用于BC风险评估和预后评估，可能有助于早期筛查和个性化治疗策略。
