Development of Predictive Scale for Diabetic Kidney Disease Progression Based on Decision Tree Classification Model
-
摘要:
目的 基于决策树分类模型建立糖尿病肾病(diabetic kidney disease,DKD)进展预测量表。 方法 回顾性收集六盘水市第二人民医院内分泌科2020年7月至2021年7月收治的308例糖尿病肾病患者作为研究对象,并将其分为微量蛋白尿组(n = 224)与显性蛋白尿组(n = 84)。对2组患者的人口学资料、既往基础病史等指标进行单因素和多因素Logistic回归分析,并运用决策树分类模型建立DKD进展预测量表。 结果 308例研究对象中84例(27.27%)为显性蛋白尿,224例(72.73%)为微量蛋白尿。多因素Logistic回归分析显示收缩压(OR = 1.022,P = 0.003)和血肌酐(OR = 1.012,P < 0.001)和总蛋白水平(OR = 0.953,P = 0.003)是引起显性蛋白尿的独立风险因素。运用决策树分类模型建立预测量表,量表总分为60分,诊断阈值为33分,决策树模型ROC曲线面积(0.781)大于多因素Logistic回归(0.769),灵敏度为95.2%,特异度为78.9%。 结论 DKD进展预测量表能够较准确的评估DKD进展,对于早期预测DKD进展具有较好的临床价值。 Abstract:Objective To establish a diabetic kidney disease (DKD) progression prediction scale based on the decision tree classification model. Methods A retrospective analysis was conducted on 308 patients with diabetic kidney disease admitted to Department of Endocrinology, the Second People's Hospital of Liupanshui from July 2020 to July 2021. The patients were divided into two groups: microalbuminuria group (n = 224) and macroalbuminuria group (n = 84). Univariate and multivariate Logistic regression analysis were performed on demographic data, past medical history and other indicators of the two groups of patients, and a DKD progression prediction scale was established using the decision tree classification model. Results Among the 308 subjects, 84 (27.27%) had macroalbuminuria and 224 (72.73%) had microalbuminuria. Multivariate Logistic regression analysis showed that systolic blood pressure (OR = 1.022, P = 0.003) and serum creatinine (OR = 1.012, P < 0.001) and total protein levels (OR = 0.953, P = 0.003) were risk factors for macroalbuminuria. The decision tree classification model was used to establish a prediction scale with a total score of 60 points and a diagnostic threshold of 33 points. The area under the ROC curve of the decision tree model (0.781) was greater than that of the multivariate logistic regression model (0.769). The sensitivity was 95.2% and the specificity was 78.9%. Conclusion DKD progression prediction scale can accurately assess the progression of DKD and has good clinical value for the early prediction of DKD progression. -
表 1 研究对象一般特征[n(%)/($\bar x \pm s $)/M(Q25,Q75)]
Table 1. Characteristics of participants [n(%)/($\bar x \pm s $)/M(Q25,Q75)]
一般特征 n=308 性别 男 169 (54.87) 女 139 (45.13%) 年龄(岁) 60.09±12.40 民族 汉族 303 (98.38) 少数民族 5 (1.62) BMI(kg/m2) 23.87±4.04 婚育史 有 303 (98.38) 无 5 (1.62) 文化程度 初中及以上 134 (43.51) 初中以下 174 (56.49) 吸烟 是 114 (37.01) 否 194 (62.99) 饮酒 是 78 (25.32) 否 230 (74.68) 糖尿病家族史 是 27 (8.77) 否 281 (91.23) 糖尿病病程(a) 6.00 (2.00,10.00) 高血压病史 是 137 (44.48) 否 171 (55.52) 高血压病程(a) 2.00 (0.00,4.00) 高血脂 是 40 (12.99) 否 268 (87.01) 高尿酸 是 23 (7.47) 否 285 (92.53) 冠心病 是 15 (4.87) 否 293 (95.13) 脑血管病史 是 48 (15.58) 否 260 (84.42) 表 2 2组间一般特征比较[n(%)/($\bar x \pm s $)/M(Q25,Q75)]
Table 2. Comparison of characteristics between the two groups [n(%)/($\bar x \pm s $)/M(Q25,Q75)]
一般特征 微量蛋白尿组(n=224) 显性蛋白尿组(n=84) t/χ2/Z P 性别 男 128 (57.14) 41 (48.81) 1.712 0.194 女 96 (42.86) 43 (51.19) 年龄(岁) 60.06±12.75 60.17±11.46 −0.071 0.945 汉族 汉族 221 (98.66) 82 (97.62) 0.422 0.523 少数民族 3 (1.34) 2 (2.38) BMI(kg/m2) 23.86±4.16 23.89±3.74 −0.056 0.945 婚育史 有 220 (98.21) 83 (98.81) 0.144 0.713 无 4 (1.79) 1 (1.19) 文化程度 初中及以上 101 (45.09) 33 (39.29) 0.842 0.366 初中以下 123 (54.91) 51 (60.71) 吸烟 是 84 (37.50) 30 (35.71) 0.084 0.777 否 140 (62.50) 54 (64.29) 饮酒 是 58 (25.89) 20 (23.81) 0.145 0.716 否 166 (74.11) 64 (76.19) 糖尿病家族史 是 21 (9.38) 6 (7.14) 0.385 0.544 否 203 (90.63) 78 (92.86) 糖尿病病程(a) 5.00 (1.00,10.00) 8.00 (3.00,14.50) −3.877 <0.001* 高血压病史 是 93 (41.52) 44 (52.38) 2.922 0.088 否 131 (58.48) 40 (47.62) 高血压病程(a) 0.00 (0.00,4.50) 0.03 (0.00,4.00) −0.863 0.392 高血脂 是 33 (14.73) 7 (8.33) 2.214 0.147 否 191 (85.27) 77 (91.67) 高尿酸 是 17 (7.59) 6 (7.14) 0.022 0.893 否 207 (92.41) 78 (92.86) 冠心病 是 9 (4.02) 6 (7.14) 1.294 0.262 否 215 (95.98) 78 (92.86) 脑血管病史 是 33 (14.73) 15 (17.86) 0.457 0.502 否 191 (85.27) 69 (82.14) *P < 0.05。 表 3 2组临床指标的差异性分析[$\bar x \pm s $/M(Q25,Q75)]
Table 3. Difference analysis of clinical indexes between the two groups [$\bar x \pm s $/M(Q25,Q75)]
临床指标 微量蛋白尿组(n=224) 显性蛋白尿组(n=84) Z/t P 收缩压(mmHg) 129.96±17.67 139.96±24.24 −3.984 <0.001* 舒张压(mmHg) 78.22±10.77 80.98±13.59 −1.852 0.065 总蛋白(g/L) 66.83±8.74 61.72±9.08 4.511 <0.001* 白蛋白(g/L) 42.05±25.87 37.07±11.10 1.713 0.089 空腹血糖(mmol/L) 11.65±4.54 10.45±3.97 2.142 0.033* 餐后血糖(mmol/L) 18.86±6.06 16.64±5.04 3.001 0.003* 空腹C肽(nmol/L) 1.96 (1.32,3.20) 2.03 (1.23,3.05) 0.863 0.394 餐后2hC肽(nmol/L) 3.28 (2.26,4.98) 3.76 (2.19,5.63) −0.633 0.532 甘油三酯(mmol/L) 1.89 (1.23,2.96) 1.82 (1.24,2.72) 0.334 0.742 胆固醇(mmol/L) 4.61±1.25 4.77±1.43 −0.943 0.351 高密度脂蛋白(mmol/L) 1.20 (1.03,1.39) 1.26 (1.05,1.47) −1.113 0.272 低密度脂蛋白(mmol/L) 3.08±1.05 3.20±1.19 −0.872 0.383 总胆红素(μmol/L) 13.55 (9.75,17.30) 8.90 (5.69,13.75) 5.842 <0.001* 直接胆红(μmol/L) 4.01 (2.72,5.49) 2.39 (1.87,3.63) 5.873 <0.001* 血红蛋白(g/L) 133.14±23.43 116.69±25.73 5.344 <0.001* 糖化血红蛋白(%) 11.30 (9.52,13.42) 10.16 (8.37,13.16) 2.832 0.005* 血尿酸(μmol/L) 345.37 (263.31,446.50) 391.50 (300.00,456.39) −1.734 0.084 血肌酐(μmol/L) 67.99 (55.48,89.13) 101.20 (67.90,195.74) −5.812 <0.001* 胱抑素(mg/L) 0.38 (0.30,0.60) 0.66 (0.38,1.17) −3.543 <0.001* 肾小球滤过率[mL/(min·1.73 m2)] 96.17 (70.44,119.82) 60.78 (23.29,92.33) 5.984 <0.001* 24h尿蛋白定量(g/24 h) 168.00 (100.00,300.00) 1188.00 (552.50,3877.50 )−9.547 <0.001* UACR(mg/g) 76.52 (47.95,110.90) 644.70 (473.30,972.95) −13.411 <0.001* *P < 0.05。 表 4 多因素Logistic回归分析
Table 4. Multivariate Logistic regression analysis
变量 B SE Wald df P OR 95%CI 收缩压 0.021 0.007 9.055 1 0.003* 1.022 1.007 1.036 总蛋白 −0.048 0.016 8.538 1 0.003* 0.953 0.923 0.984 血肌酐 0.012 0.003 20.835 1 <0.001* 1.012 1.007 1.017 常量 −1.99 1.470 1.833 1 0.176 0.137 *P < 0.05。 表 5 风险筛查评分量表
Table 5. Risk screening rating scale
自变量 OR(95%CI) 权重 赋值 收缩压 1.022 (1.007,1.036) 10 0(≤110);10(110~130);20(>130) 总蛋白 0.953 (0.923,0.984) 10 10(≤52.7);0(>52.7) 血肌酐 1.012 (1.007,1.017) 10 0(≤45.78);10(45.78~96.20);20(96.20~188.87);20(>188.87) 表 6 决策树模型和Logistic回归分析的比较
Table 6. Comparison of decision tree model and Logistic regression analysis
检验方法 AUC SE P 95%CI Logistic回归 0.769 0.032 <0.001* 0.705 0.832 决策树 0.781 0.030 <0.001* 0.721 0.840 *P < 0.05。 -
[1] Wang S,Yang Y,He X,et al. Cdk5-Mediated phosphorylation of sirt1 contributes to podocyte mitochondrial dysfunction in diabetic nephropathy[J]. Antioxidants & Redox Signaling,2021,34(3):171-190. [2] Luo Y,Lu Z,Waaga-Gasser A M,et al. Modulation of calcium homeostasis may be associated with susceptibility to renal cell carcinoma in diabetic nephropathy Rats[J]. Cancer Management and Research,2020,12:9679-9689. doi: 10.2147/CMAR.S268402 [3] 王晋文,芮章茹,任晓燕. 云南慢性肾衰竭患者病因与转归分析[J]. 昆明医学院学报,2010,31(10):102-104+117. [4] Mann J F E,Green D,Jamerson K,et al. Avosentan for overt diabetic nephropathy[J]. Journal of the American Society of Nephrology: JASN,2010,21(3):527-535. doi: 10.1681/ASN.2009060593 [5] Hamat I,Abderraman G M,Cisse M M,et al. Profile of diabetic nephropathy at the National Reference General Hospital of N’Djamena[J]. The Pan African Medical Journal,2016,24(1):193-196. [6] 张圣,胡振杰,叶璐,等. 决策树分析在急性心肌梗死事件预测中的应用[J]. 浙江大学学报(医学版),2019,48(6):594-602. doi: 10.3785/j.issn.1008-9292.2019.12.02 [7] 张向伟. 早期糖尿病肾脏病向临床期糖尿病肾脏病进展预测因子及模型研究[D]. 北京: 北京中医药大学,2017. [8] 李佳佳,黄皓,陶立坚,等. 糖尿病肾病主要发病机制的研究进展[J]. 生命科学,2023,35(3):396-404. [9] 孟继娴,刘蕾,甄紫伊,等. 2型糖尿病患者糖尿病肾病发生风险预测模型的研究进展[J]. 沈阳医学院学报,2023,25(5):525-528,534. [10] 文晓晨,马晓燕,宫成军. 糖尿病肾脏病进展风险因素和预测模型构建[J]. 辽宁中医药大学学报,2023,25(1):161-170. [11] 王富军,王文琦. 中国2型糖尿病防治指南(2020年版)解读[J]. 河北医科大学学报,2021,42(12):1365-1371. doi: 10.3969/j.issn.1007-3205.2021.12.001 [12] 孙玲莉,董世杰,杨贵军. 常用多重插补法的插补重数选择[J]. 统计与决策,2019,35(23):5-10. [13] 柳红芳,姜旻. 基于临床研究的糖尿病肾脏病预测量表的研制[J]. 北京中医药大学学报,2018,41(5):418-422. doi: 10.3969/j.issn.1006-2157.2018.05.011 [14] Tuttle K R,Agarwal R,Alpers C E,et al. Molecular mechanisms and therapeutic targets for diabetic kidney disease[J]. Kidney International,2022,102(2):248-260. doi: 10.1016/j.kint.2022.05.012 [15] 张曼,戴丽芬,田福璐,等. 糖尿病肾病发生发展的相关危险因素[J]. 中国老年保健医学,2020,18(2):80-82. [16] 王力宁. 糖尿病患者慢性肾脏病的早期筛查和防治[J]. 肾脏病与透析肾移植杂志,2007,(6):542-543. doi: 10.3969/j.issn.1006-298X.2007.06.009 [17] 李桂霞,翟晓丽,黄艺,等. 糖尿病肾脏病合并高血压患者昼夜血压节律变化与脂代谢、内皮功能的相关性研究[J]. 中国处方药,2023,21(6):1-4. doi: 10.3969/j.issn.1671-945X.2023.06.002 [18] 徐德凤,朱德琪. 糖尿病与高血压[J]. 山东医药,1988,(10):9. [19] Barutta F,Bellini S,Canepa S,et al. Novel biomarkers of diabetic kidney disease: Current status and potential clinical application[J]. Acta Diabetologica,2021,58(7):819-830. doi: 10.1007/s00592-020-01656-9 [20] 李明,罗绍珍,叶婧. 低分子肝素联合血管转换酶抑制剂对糖尿病肾病显性蛋白尿的治疗观察[J]. 内科,2006,(2):116-117. doi: 10.3969/j.issn.1673-7768.2006.02.009 [21] 林秋璇,林彩战. 糖肾方治疗糖尿病肾病显性蛋白尿期临床观察[J]. 光明中医,2022,37(13):2292-2294. [22] 王征,李艳芳. 糖肾方治疗糖尿病肾病显性蛋白尿期的临床疗效及其作用机制探讨[J]. 中国中西医结合肾病杂志,2020,21(4):328-330. [23] 梁瑛楠,刘玉宁,周静威,等. 以大量蛋白尿为表现的糖尿病肾脏病临床特点及中医证型分析[J]. 中国中西医结合肾病杂志,2023,24(11):974-977. [24] Naaman S C,Bakris G L. Slowing diabetic kidney disease progression: Where do we stand today?[J]. Compendia,2021,2021(1):28-32. doi: 10.2337/db20211-28 [25] van Raalte D H,Bjornstad P,Cherney D Z I,et al. Combination therapy for kidney disease in people with diabetes mellitus[J]. Nature Reviews Nephrology,2024,6(1):1-14. [26] An N,Wu B,Yang Y,et al. Re-understanding and focusing on normoalbuminuric diabetic kidney disease[J]. Frontiers in Endocrinology,2022,13:1077929. doi: 10.3389/fendo.2022.1077929