2型糖尿病女性患者绝经后骨质疏松预测模型的构建与验证

Establishment and validation of a prediction model for postmenopausal osteoporosis in women with type 2 diabetes mellitus

摘要:

目的 探讨2型糖尿病（T2DM）女性患者绝经后骨质疏松现状，构建风险预测模型，并验证该模型的预测效果。方法 选取2019年1月—2021年12月海南医学第二附属医院收治的98例T2DM女性患者，根据是否发生骨质疏松将其分为骨质疏松组（53例）与非骨质疏松组（45例），搜集并比较两组患者的临床资料，对差异具有统计学意义的指标进行多因素Logistic回归分析，以构建预测模型。采用Hosmer-Lemeshow（H-L）检验模型的拟合优度，另选取50例患者通过受试者工作特征（ROC）曲线验证模型的预测效果。结果 两组在年龄、T2DM病程、体质量指数（BMI）、绝经年龄、绝经年限、生育次数、受教育程度及血清糖化血红蛋白、25（OH）D₃、碱性磷酸酶（ALP）水平方面比较，差异有统计学意义（P <0.05）。非条件逐步多因素Logistic回归分析结果：年龄≥ 65岁［O^R=9.625（95% CI：1.378，67.246）］、绝经年限≥15年［O^R=10.125（95% CI：1.466，69.934）］、生育次数≥ 3次［O^R=7.200（95% CI：1.081，47.962）］、血清ALP≥ 68.17 u/L ［O^R=12.500（95% CI：2.105，24.483）］是T2DM女性患者绝经后发生骨质疏松的危险因素（P <0.05）；BMI> 24.0 kg/m² ［O^R=0.068（95% CI：0.009，0.508）］、受教育程度初中及以上［O^R=0.069（95% CI：0.007，0.736）］、血清25（OH）D₃≥ 36.22 nmol/mL ［O^R=0.167（95% CI：0.028，0.983）］是T2DM女性患者绝经后发生骨质疏松的保护因素（P <0.05）。最终构建T2DM女性患者绝经后骨质疏松风险预测模型为：Logistic（P）=（-0.167）+2.264×年龄+（-2.686）×BMI+2.315×绝经年限+1.974×生育次数+（-2.667）×受教育程度+（-1.792）×25（OH）D₃+3.114×ALP。经H-L检验P=0.170，提示该模型拟合较好。该研究模型的ROC曲线下面积为0.861（95% CI：0.820，0.893），约登指数最大值为0.580时，选取最佳临界值为0.471，此时敏感性为0.887，特异性为0.822。在临床模型验证中，当截断值≥0.471时ROC曲线下面积为0.832，敏感性为0.850、特异性为0.775。结论 T2DM女性患者绝经后骨质疏松的发生受多因素影响，该研究构建的预测模型一致性与效能较好，可为临床早期识别T2DM女性患者绝经后骨质疏松高危人群提供参考，并制订相应干预措施，以提高患者骨密度，降低骨质疏松发生风险。

Abstract:

Objective To investigate the occurrence of postmenopausal osteoporosis in female patients with type 2 diabetes mellitus (T2DM), to establish a risk prediction model, and to verify the predictive efficacy of the model.Methods Ninety-eight female T2DM patients treated in the Second Affiliated Hospital of Hainan Medical College from January 2019 to December 2021 were selected. According to the presence of osteoporosis, they were divided into osteoporosis group (53 cases) and non-osteoporosis group (45 cases). The clinical data of these patients collected and compared, and the indicators that were different between the two groups were further analyzed via multivariable Logistic regression to establish the prediction model. Hosmer-Lemeshow (H-L) test was used to assess the goodness of fit of the model, and another 50 patients were selected to verify the predictive efficacy of the model through the receiver operating characteristic (ROC) curve analysis.Results There were significant differences between the two groups in age, disease duration, body mass index (BMI), age of menopause, years since menopause, number of pregnancies, level of education and serum levels of glycosylated hemoglobin (HbA1c), 25-hydroxyvitamin D₃ [25(OH)D₃] and alkaline phosphatase (ALP) (P < 0.05). The unconditional multivariable Logistic stepwise regression analysis revealed that age ≥ 65 years [O^R = 9.625 (95% CI: 1.378, 67.246) ], years since menopause ≥ 15 [O^R =10.125 (95% CI: 1.466, 69.934) ], number of pregnancies ≥ 3 [O^R = 7.200 (95% CI: 1.081, 47.962) ], and serum ALP ≥ 68.17 u/L [O^R = 12.500 (95% CI: 2.105, 24.483) ] were risk factors for the occurrence of postmenopausal osteoporosis in female T2DM patients, and that BMI > 24.0 kg/m² [O^R = 0.068 (95% CI: 0.009, 0.508) ], level of education no less than junior middle school [O^R = 0.069 (95% CI: 0.007, 0.736) ], and serum 25(OH)D₃ ≥ 36.22 nmol/mL [O^R= 0.167 (95% CI: 0.028, 0.983) ] were the protective factors (P < 0.05). The final risk prediction model for postmenopausal osteoporosis in female T2DM patients was established as Logistic (P) = (-0.167) + 2.264 × age + (-2.686) × BMI + 2.315 × years since menopause + 1.974 × number of pregnancies + (-2.667)× level of education + (-1.792) × serum level of 25(OH)D₃ + 3.114 × serum level of ALP. The H-L test demonstrated a P value of 0.170, suggesting that the model fitted well. The area under the ROC curve (AUC) of the model was 0.861 (95% CI: 0.820, 0.893) and the maximum value of Youden index was 0.580. Besides, the optimal cutoff value of the model was 0.471, with a sensitivity of 0.887 and a specificity of 0.822. In the validation of the clinical model, the AUC was 0.832 when the cutoff value was set as 0.471, with a sensitivity of 0.850 and a specificity of 0.775.Conclusions The occurrence of postmenopausal osteoporosis in female T2DM patients is affected by multiple factors. The prediction model established in this study exhibits great efficacy and consistency of performance, which facilitates the early identification of the high-risk group of postmenopausal osteoporosis in female T2DM patients and therefore the early interventions to improve the bone mineral density of the patients and to reduce the risk of osteoporosis.