基于血液学指标的普通人群非创伤性股骨头坏死风险预测模型构建及验证

Development and Validation of a Hematological Indicators-Based Risk Prediction Model for Non-Traumatic Osteonecrosis of the Femoral Head in the General Population

摘要: 背景：非创伤性股骨头坏死(NONFH)是一种以股骨头血运障碍为特征的进展性骨病，可导致疼痛、功能障碍及医疗负担增加。其早期诊断因症状隐匿和影像学检查局限性面临挑战。本研究旨在基于常规临床指标构建适用于普通人群的NONFH风险预测模型。方法：采用单中心回顾性病例对照设计，纳入602例研究对象(病例组202例，对照组400例)，排除激素或酒精相关NONFH病例。通过多因素逻辑回归及逐步回归筛选独立预测因子，构建列线图模型，并采用ROC曲线、校准曲线和决策曲线分析(DCA)评估模型性能。结果：研究确定5项独立危险因素：红细胞计数升高(OR = 1.81, P = 0.03)、血红蛋白升高(OR = 1.04, P < 0.001)、血清钠(OR = 0.90, P = 0.0018)、白蛋白(OR = 0.83, P < 0.001)及球蛋白降低(OR = 0.94, P = 0.0081)。列线图模型表现出良好区分度(AUC = 0.779)，校准误差仅为0.009，最佳截断值0.387时敏感度76.2%、特异度70.3%。DCA显示模型在10%~70%风险阈值范围内具有显著临床净获益。结论：本研究整合血液流变学及代谢指标，首次构建了针对普通人群(排除激素/酒精暴露)的NONFH风险预测工具。该列线图模型操作简便，为早期识别高风险个体、实施精准预防提供了依据，有望弥补传统筛查手段的不足。未来需多中心前瞻性研究进一步验证其普适性。

Abstract: Background: Non-traumatic osteonecrosis of the femoral head (NONFH) is a progressive bone disorder characterized by impaired blood supply, leading to pain, functional disability, and increased healthcare burdens. Early diagnosis remains challenging due to nonspecific symptoms and limited accessibility to advanced imaging. This study aimed to develop a practical risk prediction model for NONFH in the general population using routinely available clinical indicators. Methods: A retrospective case-control study analyzed 602 participants (202 NONFH cases, 400 controls) from a single center. Exclusion criteria included alcohol abuse or glucocorticoid-induced NONFH. Multivariable logistic regression and stepwise selection identified independent predictors. A nomogram was constructed and validated through ROC analysis, calibration curves, and decision curve analysis (DCA). Results: Five independent risk factors were identified: elevated red blood cell count (OR = 1.81, P = 0.03) and hemoglobin (OR = 1.04, P < 0.001), and decreased serum sodium (OR = 0.90, P = 0.0018), albumin (OR = 0.83, P < 0.001), and globulin (OR = 0.94, P = 0.0081). The nomogram demonstrated good discrimination (AUC = 0.779), calibration (mean absolute error = 0.009), and clinical utility, with sensitivity of 76.2% and specificity of 70.3% at the optimal cutoff (0.387). DCA confirmed significant net benefit across risk thresholds of 10%~70%. Conclusion: This model integrates hematological and biochemical markers to provide a clinically accessible tool for early NONFH risk stratification. By excluding glucocorticoid and alcohol-related cases, it highlights novel predictors linked to blood rheology and metabolic balance. The nomogram facilitates individualized prevention strategies, potentially reducing delayed diagnosis and disease progression. Further multicenter validation is warranted to enhance generalizability.