Using Metabolic and Biochemical Indicators to Predict Diabetic Retinopathy by Back-Propagation Artificial Neural Network

Figures & data

Table 1 Characteristics of Participants (datadryad website data)

Variables	Diabetes without DR	Diabetes with DR	P-value
Gender (male, %)	162 (62.5)	78 (46.7)	0.001
History of hypertension (yes, %)	145 (56)	94 (56.3)	0.013
Age (years)	57.75 ± 14.16	60.41 ± 12.90	0.05
Duration of diabetes (year)	8.02 ± 7.62	12.38 ± 7.27	<0.001
SBP (mmHg)	135.21 ± 19.01	143.28 ± 23.31	0.82
DBP (mmHg)	80.22 ± 11.60	80.15 ± 12.47	0.95
HbA1c (%)	9.54 ± 2.72	9.71 ± 2.28	0.51
TG (mmol/L)	2.92 ± 1.92	2.22 ± 2.42	0.52
UA (μmol/L)	359.83 ± 110.34	378.11 ± 127.80	0.12
Cr (μmol/L)	77.66 ± 29.43	111.12 ± 92.42	<0.001
BUN (mmol/L)	5.88 ± 2.41	9.86 ± 2.72	0.02
ALT (U/L)	27.68 ± 15.82	19.58 ± 13.53	0.005
AST (U/L)	25.03 ± 17.75	19.87 ± 9.25	0.021
ALB (g/L)	37.96 ± 4.57	35.70 ± 5.69	<0.001
HDL-c (mmol/L)	1.00 ± 0.34	1.07 ± 0.32	0.04
LDL-c (mmol/L)	3.10 ± 0.93	3.37 ± 1.11	0.008
AchE (U/L)	8463.67 ± 2011.18	8231.12 ± 2228.64	0.27
CHOL (mmol/L)	4.83 ± 1.45	5.27 ± 1.85	0.007

Note: Data from Association of diabetic retinopathy and diabetic macular edema with renal function in southern Chinese patients with type 2 diabetes mellitus: a single-center observational study, Dryad, Dataset, Available from: https://datadryad.org/stash/dataset/doi:10.5061/dryad.6kg1sd7.¹⁹

Abbreviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting plasma glucose; HbA1c, glycosylated hemoglobin; TG, triglyceride; UA, Uric acid; Cr, creatinine; BUN, Urea nitrogen; ALT, alanine aminotransferase; AST, Ceramic oxalacetic transaminase; ALB, albumin; ACHE, cholinesterase.

Table 2 Characteristics of Participants (database of our hospital)

Variables	Diabetes without DR	Diabetes with DR	P-value
Gender (male, %)	51 (56)	49 (53.3)	0.7
History of hypertension (yes, %)	47 (51.6)	68 (73.9)	0.002
Age (years)	60.41 ± 13.92	58.64 ± 13.33	0.38
Duration of diabetes (year)	11.27 ± 8.38	13.72 ± 8.62	0.05
SBP (mmHg)	136.23 ± 17.65	140.01 ± 18.87	0.16
DBP (mmHg)	79.29 ± 11.41	81.28 ± 10.32	0.22
HbA1c (%)	9.53 ± 2.25	9.74 ± 2.23	0.52
TG (mmol/L)	1.90 ± 1.02	1.94 ± 1.42	0.91
UA (μmol/L)	311.59 ± 93.22	322.90 ± 97.54	0.42
Cr (μmol/L)	62.99 ± 20.38	81.36 ± 88.23	0.05
BUN (mmol/L)	5.72 ± 1.94	6.70 ± 3.37	0.02
ALT (U/L)	21.39 ± 17.02	21.79 ± 16.25	0.87
AST (U/L)	18.74 ± 9.02	20.89 ± 12.27	0.18
ALB (g/L)	39.64 ± 3.94	39.21 ± 5.07	0.52
HDL-c (mmol/L)	1.21 ± 0.31	1.17 ± 0.29	0.42
LDL-c (mmol/L)	2.90 ± 1.10	2.91 ± 0.98	0.94
AchE (U/L)	8284.01 ± 1747.39	8800.62 ± 2253.91	0.09
CHOL (mmol/L)	4.77 ± 1.48	4.80 ± 1.30	0.88

Table 3 Factors Associated with the Development of Diabetic Retinopathy (datadryad website data)

Variables	Regression Coefficients	OR (95% CI)	P-value
Gender	−0.89	0.41 (0.25–0.69)	<0.001
History of hypertension	0.12	1.12 (0.55–2.28)	0.75
Age (years)	−0.027	0.97 (0.95–1)	0.02
Duration of diabetes (year)	0.076	1.08 (1.04–1.12)	<0.001
SBP (mmHg)	0.014	1.01 (1–1.03)	0.13
DBP (mmHg)	−0.001	1.00 (0.97–1.03)	0.95
HbA1c (%)	0.05	1.05 (0.95–1.16)	0.32
TG (mmol/L)	−0.13	0.88 (0.73–1.07)	0.19
UA (μmol/L)	0.001	1.00 (1–1.02)	0.61
Cr (μmol/L)	0.009	1.01 (1.00–1.02)	0.06
BUN (mmol/L)	0.033	1.03 (0.91–1.18)	0.62
ALT (U/L)	−0.004	1.00 (0.97–1.02)	0.77
AST (U/L)	−0.024	0.98 (0.94–1.01)	0.21
ALB (g/L)	−0.05	0.95 (0.90–1.00)	0.05
HDL-c (mmol/L)	−0.157	0.86 (0.31–2.36)	0.76
LDL-c (mmol/L)	−0.181	0.83 (0.42–1.67)	0.61
AchE (U/L)	0	1 (1–1)	0.31
CHOL (mmol/L)	0.37	1.45 (0.81–2.58)	0.21

Note: Data from Association of diabetic retinopathy and diabetic macular edema with renal function in southern Chinese patients with type 2 diabetes mellitus: a single-center observational study, Dryad, Dataset, Available from: https://datadryad.org/stash/dataset/doi:10.5061/dryad.6kg1sd7 ¹⁹

Abbreviations: OR, odds ratio; CI, confidence interval.

Table 4 Factors Associated with the Development of Diabetic Retinopathy (database of our hospital)

Variables	Regression Coefficients	OR (95% CI)	P-value
Gender	−0.22	0.80 (0.38–1.69)	0.56
History of hypertension	−1.24	0.29 (0.13–0.67)	0.003
Age (years)	−0.05	0.95 (0.92–0.99)	0.008
Duration of diabetes (year)	0.05	1.05 (1.00–1.10)	0.04
SBP (mmHg)	−0.009	0.99 (0.97–1.02)	0.47
DBP (mmHg)	0.02	1.02 (0.97–1.06)	0.22
HbA1c (%)	0.12	1.13 (0.94–1.35)	0.19
TG (mmol/L)	−0.1	0.90 (0.68–1.2)	0.483
UA (μmol/L)	−0.002	1.00 (0.99–1.00)	0.32
Cr (μmol/L)	0.004	0.99 (0.97–1.02)	0.58
BUN (mmol/L)	0.15	1.16 (0.95–1.41)	0.14
ALT (U/L)	−0.03	0.97 (0.94–1.01)	0.12
AST (U/L)	0.07	1.07 (1–1.14)	0.04
ALB (g/L)	−0.03	0.97 (0.88–1.07)	0.51
HDL-c (mmol/L)	−0.12	0.89 (0.21–3.79)	0.87
LDL-c (mmol/L)	−0.22	0.80 (0.31–2.03)	0.64
AchE (U/L)	0	1 (1–1)	0.34
CHOL (mmol/L)	0.08	1.08 (0.48–2.42)	0.85

Table 5 Multivariable Logistic Regression Analysis Results of DR Related Factors (datadryad website data)

Variables	Regression Coefficients	Wald Chi-Square	OR (95% CI)	P-value
Cr (μmol/L)	−0.011	12.41	0.99 (0.98–1.00)	<0.001
Gender	0.9	14.95	2.46 (1.56–3.88)	<0.001
Age (years)	0.02	4.53	1.02 (1.01–1.04)	0.03
Duration of diabetes (year)	−0.074	18.69	0.93 (0.90–0.96)	<0.001
ALB (g/L)	0.069	8.40	1.07 (1.02–1.12)	0.004

Note: Data from Association of diabetic retinopathy and diabetic macular edema with renal function in southern Chinese patients with type 2 diabetes mellitus: a single-center observational study, Dryad, Dataset, Available from: https://datadryad.org/stash/dataset/doi:10.5061/dryad.6kg1sd7 ¹⁹

Figure 1 ROC curve of DR predicted by logistics regression model (Datadryad database).

Table 6 Multivariable Logistic Regression Analysis Results of DR Related Factors (database of our hospital)

Variables	Regression Coefficients	Wald Chi-Square	OR (95% CI)	P-value
AST (U/L)	−0.016	1.1	0.99 (0.96–1.01)	0.29
History of hypertension	1.173	10.84	3.23 (1.61–6.50)	0.001
Age (years)	0.043	8.48	1.04 (1.01–1.07)	0.004
Duration of diabetes (year)	−0.055	6.26	0.95 (0.91–0.99)	0.012

Figure 2 ROC curve of DR predicted by logistics regression model (database of our hospital).

Figure 3 The BP-ANN training process.

Figure 4 (A)The covariate-adjusted ROC curve by BP-ANN model to predict DR; (B) fitting performance of the BP-ANN (7 nodes hidden layer); (C) changes in parameters of the BP-ANN (7 nodes hidden layer); (D) regression performance of BP-ANN (7 nodes hidden layer).

Figure 5 ROC curve of DR predicted by BP-ANN model (data of our hospital).

Association of diabetic retinopathy and diabetic macular edema with renal function in southern Chinese patients with type 2 diabetes mellitus: a single-center observational study, Dryad, Dataset, Available from: https://datadryad.org/stash/dataset/doi:10.5061/dryad.6kg1sd7.

 Google Scholar