Development and validation of a random forest model for diabetic nephropathy with cardiac autonomic neuropathy

doi:10.11958/20252563

Abstract

Abstract:

Objective To develop and validate the random forest model based on glycolipid metabolism and cardiometabolic index (CMI) and to predict the risk of diabetic kidney disease (DKD) combined with cardiac autonomic neuropathy (CAN) in patients with type 2 diabetes mellitus (T2DM). Methods A retrospective single-center study design was adopted. A total of 109 patients with DKD and CAN admitted between February 2023 and February 2025 were consecutively enrolled as the comorbidity group. Based on the baseline data of the case group, 109 T2DM patients without DKD or CAN during the same period were selected in a 1∶1 matching ratio as the non-comorbidity group. The baseline characteristics of the two groups were compared. Binary Logistic regression was used to analyze the influencing factors for the occurrence of DKD combined with CAN in T2DM patients. A random forest model was constructed using the R software package (version 4.1.0), and a receiver operating characteristic (ROC) curve was plotted to analyze the predictive value of the model. Results Binary Logistic regression revealed that the urinary albumin-to-creatinine ratio (UACR)>30 mg/g, elevated fasting plasma glucose (FPG), CMI, glycated hemoglobin (HbA1c) and triglyceride (TG) levels were risk factors for the development of DKD combined with CAN in T2DM patients, while estimated glomerular filtration rate (eGFR) and high-density lipoprotein cholesterol (HDL-C) were protective factors. Using the variable importance measure (%IncMSE) for scoring and feature importance ranking, the top three factors were CMI, HDL-C and FPG, with %IncMSE values of 26.700%, 16.300% and 13.400%, respectively. Based on these influencing factors, the random forest model established to predict the occurrence of DKD combined with CAN in T2DM patients achieved an AUC of 0.849, a sensitivity of 0.862, a specificity of 0.730 and a Youden index of 0.592, with a 95% confidence interval of 0.738-0.933, demonstrating good predictive performance. Conclusion UACR >30 mg/g，elevated levels of FPG, CMI, HbA1c and TG, along with decreased levels of eGFR and HDL-C are risk factors for the occurrence of DKD combined with CAN in patients with T2DM, among which CMI is the key driver factor.

Key words: diabetes mellitus, type 2, diabetic nephropathies, diabetic neuropathies, cardiac autonomic neuropathy, glucose-lipid metabolism, cardiac metabolic index

CLC Number:

R587.2

Figures/Tables 6

References 17

[1]	中华医学会糖尿病学分会. 中国2型糖尿病防治指南(2020年版)[J]. 中华内分泌代谢杂志, 2021, 37(4):311-398.
	Chinese Diabetes Society. Guideline for the prevention and treatment of type 2 diabetes mellitus in China (2020 edition)[J]. Chinese Journal of Diabetes, 2021, 37(4):311-398. doi:10.3760/cma.j.cn115791-20210221-00095
[2]	THIPSAWAT S. Early detection of diabetic nephropathy in patient with type 2 diabetes mellitus:a review of the literature[J]. Diab Vasc Dis Res, 2021, 18(6):14791641211058856. doi:10.1177/14791641211058856.
[3]	WILLIAMS S, RAHEIM S A, KHAN M I, et al. Cardiac autonomic neuropathy in type 1 and 2 diabetes:epidemiology,pathophysiology,and management[J]. Clin Ther, 2022, 44(10):1394-1416. doi:10.1016/j.clinthera.2022.09.002.
[4]	CHOWDHURY M, NEVITT S, ELEFTHERIADOU A, et al. Cardiac autonomic neuropathy and risk of cardiovascular disease and mortality in type 1 and type 2 diabetes:a meta-analysis[J]. BMJ Open Diabetes Res Care, 2021, 9(2):e002480. doi:10.1136/bmjdrc-2021-002480.
[5]	MICHOU V, TSAMOS G, VASDEKI D, et al. Diabetic kidney disease and cardiac autonomic neuropathy:insights on exercise rehabilitation[J]. J Nephrol, 2025, 38(2):457-471. doi:10.1007/s40620-025-02216-6.
[6]	李艳敏, 张维, 胡彦彦. 2型糖尿病肾病患者血清糖脂代谢指标、胱抑素C、同型半胱氨酸水平的变化及其临床意义[J]. 海南医学, 2025, 36(3):386-390.
	LI Y M, ZHANG W, HU Y Y. Changes of serum glucose and lipid metabolism indexes,cystatin C,and homocysteine levels in patients with type 2 diabetic nephropathy and their clinical significance[J]. Hainan Medical Journal, 2025, 36(3):386-390. doi:10.3969/j.issn.1003-6350.2025.03.016.
[7]	WAKABAYASHI I, DAIMON T. The "cardiometabolic index" as a new marker determined by adiposity and blood lipids for discrimination of diabetes mellitus[J]. Clin Chim Acta, 2015, 438:274-278. doi:10.1016/j.cca.2014.08.042.
[8]	田丽娜, 牛奔, 朱恩仙, 等. 2型糖尿病心脏自主神经病变合并视网膜病变和肾病的临床研究[J]. 重庆医科大学学报, 2024, 49(1):18-23.
	TIAN L N, NIU B, ZHU E X, et al. A clinical study of cardiac autonomic neuropathy with retinopathy and nephropathy in patients with type 2 diabetes mellitus[J]. Journal of Chongqing Medical University, 2024, 49(1):18-23. doi:10.13406/j.cnki.cyxb.003416.
[9]	SHILO S, KESHET A, ROSSMAN H, et al. Continuous glucose monitoring and intrapersonal variability in fasting glucose[J]. Nat Med, 2024, 30(5):1424-1431. doi:10.1038/s41591-024-02908-9.
[10]	DEMIRELLI B, BOZTEPE B, ŞENOL E G, et al. Non-diabetic nephropathy in diabetic patients:incidence,HbA1c variability and other predictive factors,and implications[J]. Int Urol Nephrol, 2024, 56(9):3091-3100. doi:10.1007/s11255-024-04066-w.
[11]	TAM A A, ALTAY F P, DEMIR P, et al. The association between diabetic nephropathy and triglyceride/glucose index and triglyceride/high-density lipoprotein cholesterol ratio in patients with type 2 diabetes mellitus[J]. J Clin Med, 2024, 13(22):6954. doi:10.3390/jcm13226954.
[12]	ZHANG L, FAN D, ZHU T, et al. The ratio of non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol is associated with diabetic kidney disease:a cross-sectional study[J]. PLoS One, 2024, 19(11):e0311620. doi:10.1371/journal.pone.0311620.
[13]	TANG W Z, CAI Q Y, LIU T H, et al. The association between the cardiac metabolic index and rapid kidney function decline and CKD in individuals with different glucose metabolism statuses:results from the China health and retirement longitudinal study[J]. Lipids Health Dis, 2025, 24(1):153. doi:10.1186/s12944-025-02572-z.
[14]	吴焱, 张秋荣, 余晓波, 等. 心脏代谢指数及血尿酸水平与2型糖尿病患者心脏自主神经病变的相关性研究[J]. 临床内科杂志, 2024, 41(8):566-568.
	WU Y, ZHANG Q R, YU X B, et al. Association between the cardiometabolic index,serum uric acid levels and cardiac autonomic neuropathy in patients with type 2 diabetes mellitus[J]. Journal of Clinical Internal Medicine, 2024, 41(8):566-568. doi:10.3969/j.issn.1001-9057.2024.08.017.
[15]	ZHANG L, LIANG C, YAN Z, et al. Association between cardiac metabolic index and diabetic kidney disease:a cross-sectional study of NHANES 1999-2018[J]. J Health Popul Nutr, 2025, 44(1):105. doi:10.1186/s41043-025-00826-1.
[16]	LI B, WANG W, GAO Z, et al. A novel index,Chinese visceral adiposity index is closely associated with urinary albumin-creatinine ratio in Chinese community adults,especially in hypertensive or hyperglycemic population:results from the REACTION study[J]. J Diabetes, 2022, 14(12):792-805. doi:10.1111/1753-0407.13336.
[17]	LI T, LIU T C, LIU N, et al. Urinary exosome proteins PAK6 and EGFR as noninvasive diagnostic biomarkers of diabetic nephropathy[J]. BMC Nephrol, 2023, 24(1):291. doi:10.1186/s12882-023-03343-7.

组别	BMI≥ 24 kg/m²		男性		高血压		吸烟史		饮酒史	UACR> 30 mg/g		FPG/ （mmol/L）		CMI		WHtR	eGFR/[mL/（min·1.73 m²）]			HbA1c/ %
A组	63（57.80）		61（55.96）		18（16.51）		35（32.11）		39（35.78）	34（31.19）		6.89±0.55		0.64±0.25		0.47±0.16	90.21±5.70			6.13±1.05
B组	61（55.96）		58（53.21）		20（18.35）		37（33.94）		42（38.53）	62（56.88）		7.31±0.53		0.95±0.38		0.49±0.15	88.31±6.60			6.85±1.03
χ²或t	0.075		0.167		0.127		0.083		0.177	14.593^＊＊		5.742^＊＊		7.162^＊＊		0.930	2.272^＊			5.105^＊＊
组别		TC/ （mmol/L）		TG/ （mmol/L）		HDL-C/ （mmol/L）		LDL-C/ （mmol/L）			年龄/岁		糖尿病病程/年		运动次数/（次/周）			SDNN/ ms	SDANN/ ms
A组		4.59±0.63		1.81±0.51		1.39±0.24		2.91±0.34			55.49±5.89		8.42±0.66		3.05±0.53			88.73±7.12	76.22±5.93
B组		4.73±0.61		2.19±0.54		1.20±0.19		2.98±0.37			55.56±5.97		8.56±0.62		3.01±0.51			86.91±6.98	74.63±6.14
t		1.784		5.391^＊＊		6.610^＊＊		1.474			0.091		1.597		0.524			1.913	1.941

组别	BMI≥ 24 kg/m²		男性		高血压		吸烟史		饮酒史	UACR> 30 mg/g		FPG/ （mmol/L）		CMI		WHtR	eGFR/[mL/（min·1.73 m²）]			HbA1c/ %
A组	63（57.80）		61（55.96）		18（16.51）		35（32.11）		39（35.78）	34（31.19）		6.89±0.55		0.64±0.25		0.47±0.16	90.21±5.70			6.13±1.05
B组	61（55.96）		58（53.21）		20（18.35）		37（33.94）		42（38.53）	62（56.88）		7.31±0.53		0.95±0.38		0.49±0.15	88.31±6.60			6.85±1.03
χ²或t	0.075		0.167		0.127		0.083		0.177	14.593^＊＊		5.742^＊＊		7.162^＊＊		0.930	2.272^＊			5.105^＊＊
组别		TC/ （mmol/L）		TG/ （mmol/L）		HDL-C/ （mmol/L）		LDL-C/ （mmol/L）			年龄/岁		糖尿病病程/年		运动次数/（次/周）			SDNN/ ms	SDANN/ ms
A组		4.59±0.63		1.81±0.51		1.39±0.24		2.91±0.34			55.49±5.89		8.42±0.66		3.05±0.53			88.73±7.12	76.22±5.93
B组		4.73±0.61		2.19±0.54		1.20±0.19		2.98±0.37			55.56±5.97		8.56±0.62		3.01±0.51			86.91±6.98	74.63±6.14
t		1.784		5.391^＊＊		6.610^＊＊		1.474			0.091		1.597		0.524			1.913	1.941

变量	β	SE	Wald χ²	P	OR	OR 95%CI	方差膨胀因子
UACR	0.844	0.407	4.312	0.038	2.326	1.048~5.160	1.073
FPG	1.453	0.394	13.618	<0.001	4.278	1.977~9.257	1.066
CMI	2.737	0.680	16.184	<0.001	15.466	4.070~58.615	1.161
eGFR	-0.104	0.034	9.470	0.002	0.901	0.843~0.963	1.027
HbA1c	0.954	0.213	20.063	<0.001	2.595	1.710~3.938	1.027
TG	1.490	0.434	11.766	0.001	4.436	1.894~10.393	1.073
HDL-C	-4.478	1.219	13.485	<0.001	0.011	0.001~0.124	1.210
常数项	-6.994	4.557	2.356	0.125	0.001

变量	β	SE	Wald χ²	P	OR	OR 95%CI	方差膨胀因子
UACR	0.844	0.407	4.312	0.038	2.326	1.048~5.160	1.073
FPG	1.453	0.394	13.618	<0.001	4.278	1.977~9.257	1.066
CMI	2.737	0.680	16.184	<0.001	15.466	4.070~58.615	1.161
eGFR	-0.104	0.034	9.470	0.002	0.901	0.843~0.963	1.027
HbA1c	0.954	0.213	20.063	<0.001	2.595	1.710~3.938	1.027
TG	1.490	0.434	11.766	0.001	4.436	1.894~10.393	1.073
HDL-C	-4.478	1.219	13.485	<0.001	0.011	0.001~0.124	1.210
常数项	-6.994	4.557	2.356	0.125	0.001

混淆矩阵		预测值		准确率/%
混淆矩阵		发生	未发生	准确率/%
真实值	发生	30	7	85.00
真实值	未发生	3	26	85.00