Baseline
As shown in Table 1, gender, presence of hypertension, smoking, drinking, diabetes duration, FPG, WBC, Hb, BUN, Scr and TG are significantly different between DR and NDR group (P<0.05). Specifically, compared with the NDR group, DR patients have longer diabetes duration, higher levels of FPG, WBC, BUN, Scr and TG, a higher proportion of female and hypertension presence. However, levels of Hb were lower in DR patients. In addition, no difference was observed in age, BMI, HBA1c, TC, LDL-C, HDL-C levels as well as the treatment of diabetes. According to figure 1, levels of NLR (2.36±1.16 in DR group versus 1.97±1.06 in NDR group, p<0.001) and PLR(11.62±4.55 in DR group versus10.56±4.45 in NDR group, p=0.012)were significantly higher in DR group.
Table 1 Baseline characteristics of the DR and NDR groups
Variable
|
DR group
N=264
|
NDR group
N=206
|
P
|
Gender(male/female)
|
131/133
|
146/60
|
0.000**
|
Hypertension(no/yes)
|
95/169
|
108/98
|
0.000**
|
Smoking(no/yes)
|
168/96
|
107/99
|
0.007*
|
Drinking(no/yes)
|
180/84
|
119/87
|
0.013*
|
Sulfonylureas (no/yes)
|
147/117
|
110/96
|
0.344
|
Biguanides (no/yes)
|
129/135
|
95/110
|
0.327
|
Insulin (no/yes)
|
127/137
|
108/98
|
0.201
|
Age(years)
|
56.48±9.86
|
55.44±11.27
|
0.295
|
NLR
|
2.36±1.16
|
1.97±1.06
|
0.000**
|
MLR(*10)
|
2.28±1.03
|
2.34±1.28
|
0.628
|
PLR(*0.1)
|
11.62±4.55
|
10.56±4.45
|
0.012*
|
Neutrophils(*10^9/L)
|
4.25±1.31
|
3.72±1.22
|
0.000**
|
Lymphocytes(*10^9/L)
|
1.98±0.64
|
2.11±0.67
|
0.031*
|
Monocytes(*10^9/L)
|
0.42±0.15
|
0.44±0.15
|
0.097
|
Platelets(*10^9/L)
|
213.84±69.19
|
203.90±54.66
|
0.092
|
Diabetes course(month)
|
122.86±87.38
|
66.16±66.25
|
0.000**
|
BMI(kg/m2)
|
24.12±3.25
|
24.62±3.14
|
0.090
|
FPG(mmol/L)
|
9.60±3.79
|
8.80±3.27
|
0.019 *
|
HbA1c(%)
|
9.96±2.45
|
9.68±2.66
|
0.246
|
WBC(*10^9/L)
|
6.83±1.65
|
6.45±1.50
|
0.010*
|
Hb(g/L)
|
136.25±19.33
|
153.20±15.70
|
0.000**
|
BUN(mmol/L)
|
7.07±7.82
|
5.32±1.64
|
0.000**
|
Scr(mmol/L)
|
84.21±46.63
|
69.14±17.87
|
0.000**
|
TG(mg/dL)
|
3.02±2.37
|
2.29±1.99
|
0.000**
|
TC(mg/dL)
|
4.73±2.09
|
4.79±1.13
|
0.694
|
HDL-C(mg/dL)
|
1.16±0.33
|
1.12±0.57
|
0.434
|
LDL-C(mg/dL)
|
3.11±0.98
|
2.97±0.88
|
0.110
|
DR diabetic retinopathy, NDR non-diabetic retinopathy,NLR neutrophil-to-lymphocyte ratio, MLR monocyte-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio, BMI body mass index, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c, WBC white blood cell, Hb hemoglobin, BUN blood urea nitrogen,Scr serum creatinine,TG triglycerides, TC total cholesterol, HDL-C high density lipoprotein cholesterol, LDL-C low density lipoprotein cholesterol.
*P<0.05, **P<0.01.
Correlation analysis of NLR, PLR and Hb and major clinical factors.
As shown in Table 2, according to correlation analysis, NLR positively correlated with MLR, PLR, diabetes duration, BUN and Scr, and negatively correlated with Hb, but displayed insignificant correlation with age, gender, hypertension, BMI, FPG, and HbA1c. Meanwhile, PLR correlated positively with NLR, MLR and diabetes duration, and negatively with BMI and Hb. In addition, Hb correlated positively with BMI and negatively correlated with gender, hypertension, age, NLR, PLR, diabetes course, and BUN.
Table 2 Correlation analysis of NLR, PLR and Hb and major clinical factors
Variable
|
NLR
|
PLR (*0.1)
|
Hb(g/L)
|
R P
|
R P
|
R P
|
Gender(male/female)
|
-0.061
|
0.190
|
0.080
|
0.082
|
-0.465
|
0.000**
|
Hypertension(no/yes)
|
0.053
|
0.247
|
-0.008
|
0.855
|
-0.130
|
0.005**
|
Age(years)
|
0.055
|
0.238
|
0.029
|
0.534
|
-0.150
|
0.001**
|
NLR
|
1
|
1
|
0.520
|
0.000**
|
-0.115
|
0.012*
|
MLR(*10)
|
0.609
|
0.000**
|
0.446
|
0.000**
|
0.004
|
0.934
|
PLR(*0.1)
|
0.520
|
0.000**
|
1
|
1
|
-0.289
|
0.000**
|
Diabetes course(month)
|
0.125
|
0.007**
|
0.093
|
0.044*
|
-0.266
|
0.000**
|
BMI(kg/m2)
|
-0.021
|
0.651
|
-0.117
|
0.011*
|
0.194
|
0.000**
|
FPG(mmol/L)
|
0.015
|
0.755
|
-0.063
|
0.184
|
0.080
|
0.091
|
HbA1c(%)
|
-0.001
|
0.980
|
-0.072
|
0.129
|
0.076
|
0.110
|
Hb(g/L)
|
-0.115
|
0.012*
|
-0.289
|
0.000**
|
1
|
1
|
BUN(mmol/L)
|
0.100
|
0.033*
|
-0.042
|
0.371
|
-0.220
|
0.000**
|
Scr(mmol/L)
|
0.173
|
0.000**
|
0.079
|
0.092
|
-0.065
|
0.168
|
Sulfonylureas(no/yes)
|
-0.078
|
0.090
|
0.049
|
0.291
|
-0.016
|
0.725
|
Biguanides(no/yes)
|
0.033
|
0.476
|
-0.058
|
0.209
|
0.072
|
0.120
|
Insulin(no/yes)
|
0.081
|
0.081
|
-0.032
|
0.491
|
0.009
|
0.839
|
NLR neutrophil-to-lymphocyte ratio, MLR monocyte-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio, BMI body mass index, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c, Hb hemoglobin, BUN blood urea nitrogen,Scr serum creatinine.
*P<0.05, **P<0.01.
Univariate analysis
As shown in Table 3, our univariate regression analysis revealed that gender of female, presence of hypertension, long course of diabetes,higher levels of FPG, WBC, NLR, PLR BUN, Scr and TG adds risk to presence of DR, while higher counts of Hb was related to lower risk of DR. However, no association was found between age, BMI, HbA1c, TC, HDL-C, LDL-C, MLR and presence of DR.
Table 3 Univariate analysis
Variable
|
OR
|
95%CI.low
|
95%CI.upp
|
P
|
Gender(Male/Female)
|
2.47
|
1.68
|
3.63
|
0.000**
|
Hypertension(no/yes)
|
1.96
|
1.35
|
2.84
|
0.000**
|
Age(years)
|
1.01
|
0.99
|
1.03
|
0.295
|
NLR
|
1.46
|
1.19
|
1.79
|
0.000**
|
MLR(*10)
|
0.96
|
0.82
|
1.13
|
0.619
|
PLR(*0.1)
|
1.06
|
1.01
|
1.10
|
0.013*
|
Diabetes course(month)
|
1.01
|
1.01
|
1.01
|
0.000**
|
BMI(Kg/m2)
|
0.95
|
0.90
|
1.01
|
0.090
|
FPG(mmol/L)
|
1.07
|
1.01
|
1.13
|
0.020*
|
HbA1c(%)
|
1.05
|
0.97
|
1.13
|
0.239
|
WBC(*10^9/L)
|
1.17
|
1.04
|
1.31
|
0.010*
|
Hb(g/L)
|
0.95
|
0.93
|
0.96
|
0.000**
|
BUN(mmol/L)
|
1.42
|
1.26
|
1.59
|
0.000**
|
Scr(mmol/L)
|
1.02
|
1.01
|
1.02
|
0.000**
|
TG(mg/dL)
|
1.18
|
1.07
|
1.31
|
0.001**
|
TC(mg/dL)
|
0.98
|
0.88
|
1.09
|
0.693
|
HDL-C(mg/dL)
|
1.21
|
0.76
|
1.93
|
0.412
|
LDL-C(mg/dL)
|
1.17
|
0.12
|
1.43
|
0.110
|
NLR neutrophil-to-lymphocyte ratio, MLR monocyte-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio, BMI body mass index, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c, WBC white blood cell, Hb Hemoglobin, BUN blood urea nitrogen,Scr serum creatinine,TG triglycerides, TC total cholesterol, HDL-C high density lipoprotein cholesterol, LDL-C low density lipoprotein cholesterol.
*P<0.05, **P<0.01.
Associations between NLR, PLR, and DR
NLR and DR
By conducting multivariate analysis, we found that NLR was associated with DR independent of other known factors. With a unit increase of NLR, the risk for DR would raise 37%. Furthermore, when treated as a category variate divided according to its quantile, the association of NLR and DR still exists. As table 4 demonstrated, from the crude model to simple or complex model, there was a 2.8 fold increased risk for DR in the highest quantile of NLR (OR, 95%CI: 2.80 (1.32, 5.95) p=0.007) (Table4).
Table 4 Independent association between NLR and DR
|
Exposure
|
MODE
|
MODE I
|
MODE II
|
NLR group
|
OR 95%CI P
|
OR 95%CI P
|
OR 95%CI P
|
ContinuousNLR
|
1.46(1.19,1.78) 0.000**
|
1.46(1.18,1.81) 0.000**
|
1.37(1.06,1.78) 0.018*
|
0.46-1.49
|
1
|
1
|
1
|
1.50-1.95
|
1.89(1.12,3.17) 0.016*
|
2.11(1.19,3.73) 0.011*
|
1.23(0.62, 2.44) 0.556
|
1.95-2.54
|
2.25(1.33,3.79) 0.002**
|
2.51(1.42,4.46) 0.002**
|
1.77(0.88, 3.55) 0.110
|
2.54-11.05
|
3.44(2.01,5.89) 0.000**
|
3.61(1.99,6.55) 0.000**
|
2.80(1.32, 5.95) 0.007**
|
P for trend
|
0.000**
|
0.000**
|
0.004**
|
|
|
|
|
|
MODE: crude model; MODE I: adjusting for age, gender, diabetes duration; MODE II: adjusting for age, gender, diabetes duration, hypertension, BMI, Scr, WBC, Hb, BUN, TG, FPG and HbA1c.
NLR neutrophil-to-lymphocyte ratio,OR odds ratios, CI confidence intervals, BMI body mass index, Scr serum creatinine, WBC white blood cell, Hb hemoglobin, BUN blood urea nitrogen, TG triglycerides, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c.
* P <0.05, ** P <0.01.v
PLR and DR
We also included PLR in multivariate logistics regression by adjusting other co-founders and observed that PLR was not an independent risk factor for DR as a continuous variable, but the ranked level of the index assists the risk stratification. Specifically, after full adjustment, the highest quantile of PLR held add 2.2 times of risk to the presence of DR compared with the first quantile(OR(95%CI)2.20 (1.05, 4.59), P=0.037) (Table5).
Table 5 Independent association between PLR and DR
Exposure
|
MODE
|
MODE I
|
MODE II
|
PLRgroup
|
OR 95%CI P
|
OR 95%CI P
|
OR 95%CI P
|
Continuous PLR
|
1.06(1.01,1.10) 0.013*
|
1.04(0.99,1.09) 0.109
|
1.05(0.99, 1.11) 0.135
|
2.75-7.98
|
1
|
1
|
1
|
7.99-10.34
|
1.65(0.98, 2.76) 0.058
|
1.86(1.06, 3.27) 0.031*
|
1.37(0.70, 2.71) 0.360
|
10.36-13.40
|
1.17(0.70, 1.94) 0.558
|
1.34(0.77, 2.34) 0.307
|
0.95(0.47, 1.92) 0.891
|
13.49-35.26
|
2.51(1.47, 4.27) 0.000**
|
2.15(1.20, 3.85) 0.001**
|
2.20(1.05, 4.59) 0.037*
|
P for trend
|
0.003**
|
0.033*
|
0.073
|
MODE: crude model; MODE I: adjusting for age, gender, diabetes duration; MODE II: adjusting for age, gender, diabetes duration, hypertension, BMI, Scr, WBC, Hb, BUN, TG, FPG and HbA1c.
PLR platelet-to-lymphocyte ratio,OR odds ratios,CI confidence intervals,BMI body mass index, Scr serum creatinine, WBC white blood cell, Hb hemoglobin, BUN blood urea nitrogen, TG triglycerides, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c.
* P <0.05, ** P <0.01.
NLR and PLR for reclassification of DR
Hb and DR
Particularly, we also assess the performance of the previous factor-Hb in our diabetic patients' group. By applying crude and different multivariate models, we confirmed that no matter treated as a continuous variable or categorical variable, Hb was stably related to lower risk of DR with the higher quartile displaying lower risk probability (Table6).
Table 6 Independent association between Hb and DR
Exposure
|
MODE
|
MODE I
|
MODE II
|
Hb group
|
OR 95%CI P
|
OR 95%CI P
|
OR 95%CI P
|
Continuous Hb
|
0.95(0.93, 0.96) 0.000**
|
0.95(0.94,0.97) 0.000**
|
0.96(0.94, 0.97) 0.000**
|
70-132(g/L)
|
1
|
1
|
1
|
133-143(g/L)
|
0.32(0.17, 0.61) 0.000**
|
0.38(0.19, 0.73) 0.004**
|
0.88(0.44, 1.74) 0.707
|
144-157(g/L)
|
1.61(0.95, 2.71) 0.076
|
2.02(1.12, 3.62) 0.019*
|
0.27(0.12, 0.60) 0.001**
|
158-195(g/L)
|
0.07(0.04, 0.13) 0.000**
|
0.09(0.05,0.20) 0.000**
|
0.95(0.93, 0.97) 0.000**
|
P for trend
|
0.000**
|
0.000**
|
0.000**
|
MODE: Univariate model; MODE I: adjusting for age, gender, and diabetes course; MODE II: adjusting for age, gender, diabetes course, hypertension, BMI, Scr, WBC, NLR, PLR, BUN, TG, FPG, HbA1C.
Hb hemoglobin, OR odds ratios,CI confidence intervals,BMI body mass index, Scr serum creatinine, WBC white blood cell, NLR neutrophil-to-lymphocyte ratio, PLR platelet-to-lymphocyte ratio,BUN blood urea nitrogen, TG triglycerides, FPG fasting plasma glucose, HbA1c glycated hemoglobin A1c.
* P <0.05, ** P <0.01.
A combination of PLR, NLR, and Hb for predicting DR
In order to evaluate the prognostic value of NLR and PLR for improving risk stratification of DR, we performed a receiver-operating characteristic (ROC) analyses to calculate the area under the curve (AUC) of each factor and assess the performance of a combination of these factors (Table7). Though NLR and PLR alone didn't perform better than Hb, a combination of NLR, PLR, and Hb indeed result in a model with increased predictive performance (area under the ROC curve 0.78 (95%CI:0.74-0.82) versus. 0.76 (95%CI:0.72-0.81))(figure2). Furthermore, the addition of NLR, PLR significantly improved the risk reclassification over using Hb alone, with a considerable category-free net reclassification improvement(NRI) and a meaningful integrated discrimination improvement (IDI) for DR among diabetic patients without a family history.( NRI(95%CI) 0.53 (0.36-0.71) p<0.001; IDI(95%CI) 0.03 (0.01-0.04);p<0.001.)
Table 7 A combination of PLR, NLR and Hb for predicting DR
Marker
|
AUC
|
95%CI
up
|
95%CI down
|
Cut off
|
specificity
|
sensitivity
|
NLR
|
0.64
|
0.58
|
0.69
|
1.84
|
0.56
|
0.64
|
PLR
|
0.58
|
0.53
|
0.63
|
128.11
|
0.79
|
0.35
|
Hb
|
0.76
|
0.72
|
0.81
|
146.50
|
0.68
|
0.72
|
NLR+PLR+Hb
|
0.78
|
0.74
|
0.82
|
-
|
0.68
|
0.75
|
Based on model: logit (DR) = 8.74874 -0.06060*Hb +0.49400*NLR -0.00682*PLR.
AUC area under curve, CI confidence intervals, NLR neutrophil-to-lymphocyte ratio, PLR platelet-to- lymphocyte ratio, Hb hemoglobin, NRI net reclassification index, IDI integrated discrimination improvement.