Subjects
In this study, participants aged 18 to 80 were recruited continuously at Beijing Hospital of Traditional Chinese Medicine affiliated to Capital Medical University from January 2010 to January 2021, to conduct the retrospective study.
Initially, a total of 16,589 patients with type 2 diabetes mellitus were enrolled in this study. After screening, 13,757 patients were excluded for the following reasons: (i) receiving renal replacement therapy; (ii) with definite diagnosis of non-diabetic kidney disease; (iii) with diabetic acute complications, such as diabetic ketoacidosis, and hyperosmolar hyperglycemic coma; (iv) with chronic diseases that may affect metabolic function including hypothalamic disease, adrenal disease, any thyroid medication (levothyroxine or antithyroid medication), and a history of thyroid diseases prior to diabetes mellitus; (v) with severe respiratory, digestive, or hematological diseases, or current acute or severe infections, autoimmune diseases or malignancies; (vi) lack of necessary information, including serum creatinine, urinary albumin to creatinine ratio (ACR). Ultimately, a total of 2,832 eligible patients were enrolled in the study, including 1,710 males and 1,122 females.
Diagnosis of DKD
According to the 2012 KDIGO clinical practice guidelines, DKD was defined as a urinary albumin-creatinine ratio (uACR) ≥30 mg/g, and/or an estimated glomerular filtration rate(eGFR) <60 mL/min/1.73 m2 , in the absence of signs or symptoms of other primary causes of kidney damage[6]. eGFR was estimated according to the 2009 CKD Epidemiology Collaboration (CKD-EPI) equation [14].
Stages of DKD
According to the KDIGO risk categories, the DKD patients were divided into four stages, including low risk, moderate risk, high risk and very high risk[5] .
Clinical and biochemical indexes
All the data of this study came from the scientific research sharing platform(Yidu Cloud Research Collaboration Platform)of Beijing Hospital of Traditional Chinese Medicine affiliated to Capital Medical University.
All subjects were given standardized questionnaires for demographics and medical histories.
diabetes mellitus was defined as fasting plasma glucose ≥7.0 mmol/L, or 2-h plasma glucose ≥11.1mmol/l during an OGTT, or taking hypoglycemic drugs or receiving parenteral insulin therapy[15].
Hypertension was defined as systolic blood pressure(SBP) ≥140 mmHg, a diastolic blood pressure(DBP) ≥90 mmHg, or treatment with antihypertensive medication[16].
BMI was calculated as weight in kilograms, which was divided by height in meters squared.
The results of laboratory tests were performed for serum samples obtained by venipuncture after fasting for 8 hours in patients prior to clinical treatment. Serum thyroid hormones concentration was measured by chemiluminescence, including triiodothyronine (T3), thyroxine (T4), free triiodothyronine (FT3), free thyroxine (FT4), and thyroid-stimulating hormone (TSH). The reference ranges of TSH, T3, T4, FT3 and FT4 were 0.51-6.27 uIU /mL, 0.60-1.81ng/ml, 4.50-10.90 ug/dL, 2.30-4.20 pg/mL and 0.89-1.76 ng/dL, respectively.
The creatinine concentration was determined by enzymatic method and the urine microalbumin was determined by immunoturbidimetric method.
Serum triglycerides (TG), total cholesterol (TC), high density lipoprotein cholesterol (HDL-C),and low density lipoprotein cholesterol (LDL-C) were measured with an automated biochemical analyzer. hemoglobin A1c(HbA1c) was determined by high performance liquid chromatography. All tests were performed by trained staff in the laboratory of Beijing Hospital of Traditional Chinese Medicine, Capital Medical University.
Statistical Analysis
Statistical analysis was performed with Stata version 16 (StataCorp, College Station, TX, USA). Continuous variables with normal distribution were expressed as mean±SD, continuous variables with non-normal distribution were expressed as median (quaternion interval), and classification variables were shown as percentages. Differences between these groups were assessed, by categorical variables with χ2 test, and continuous data were analyzed by Wilcoxon rank sum test. The association between thyroid hormones and risk of DKD before and after adjustment for confounders was studied through logistic regression analysis, with the results expressed as odds ratios (OR) and 95% confidence intervals (95% CI). And then, bias in group-based equivalents was reduced with propensity score matching (PSM), calibration ability was assessed with Akaike Information Standard (AIC), Bayesian Information Standard (BIC) and -2 log likelihood ratio test, and discrimination of significant thyroid hormones was judged by the net reclassification improvement (NRI) and integrated differential improvement (IDI). And the association between thyroid hormones, ACR and eGFR was evaluated through Spearman correlation analysis. The R programming environment (version 3.5.2) was used for nod diagram and accuracy evaluation of the "RMS" package. P value of less than 0.05 was considered statistically significant.
Characteristics of Subjects
Table 1 shows the baseline characteristics of subjects, 39.62% of which were females, with the median age of 61 years old (IQR 53–69 years old) and the median known duration of diabetes mellitus of 10 years old (IQR 6–20 years old). By the KDIGO risk categories, 2,832 people with type 2 diabetes mellitus were staged into low risk, moderate risk, high risk and very high risk. Low-risk patients were used as a control group. For thyroid hormones, only FT3 were significantly different at all stages compared to the control group (P<0.001), with a continuous downward trend, T3, T4 and TSH did not change significantly in the moderate-risk group compared with the control group (P>0.05).
Table 1 Baseline characteristics of patients stratified by the KDIGO risk categories
Characteristics
|
Patients with diabetes mellitus
|
|
total
|
Low risk
|
Moderate risk
|
High risk
|
very high risk
|
|
Samplesize
|
2832
|
736
|
501
|
578
|
1017
|
|
Age(years)
|
61(53-69)
|
58(51-66)
|
62(55-70)***
|
61(53-69)***
|
62(54-70)***
|
|
Female(n,%)
|
1122,39.62
|
340,46.20
|
208,41.50
|
200,34.60***
|
374,36.80***
|
|
Duration of diabetes(years)
|
10(6-20)
|
7(3-12)
|
10(5-16)***
|
10(5-18)***
|
14.5(9-20)***
|
|
Diabetic retinopathy(n,%)
|
990,34.96
|
122,16.60
|
125,25***
|
217,37.5***
|
526,51.7***
|
|
Hypertension(n,%)
|
2045,72.20
|
256,64.60
|
350,71.3*
|
483,83.60***
|
956,94.00***
|
|
SBP(mmHg)
|
140(129-155)
|
130(120-140)
|
135(125-148)***
|
140(130-155)***
|
150(139-165)***
|
|
DBP(mmHg)
|
80(74-90)
|
80(71-85)
|
80(72-88)
|
80(75-90)***
|
80(75-90)***
|
|
BMI(Kg/m2)
|
26(23.4-28.4)
|
26.2(24.2-29.2)
|
26.05(23.90-29.00)
|
25.3(22.75-28.20)
|
25.70(23.1-28.1)
|
|
|
HbA1c(%)
|
7.10(6.20-8.60)
|
7.50(6.50-9.10)
|
7.60(6.60-9.30)
|
7.30(6.40-8.95)
|
6.60(5.90-7.70)***
|
|
|
|
TG(mmol/L)
|
1.67(1.19-2.42)
|
1.54(1.11-2.18)
|
1.56(1.14-2.28)
|
1.77(1.25-2.56)***
|
1.72(1.24-2.52)***
|
|
TC(mmol/L)
|
4.77(3.97-5.73)
|
4.64(3.93-5.38)
|
4.57(3.77-5.34)
|
4.90(4.10-6.15)***
|
4.88(3.99-6.12)***
|
|
LDL-C(mmol/L)
|
2.75(2.15- 3.46)
|
2.66(2.12-3.26)
|
2.61(2.03-3.29)
|
2.87(2.22-3.71)***
|
2.85(2.19-3.69)***
|
|
HDL-C(mmol/L)
|
1.17(0.99-1.38)
|
1.19(1.01-1.37)
|
1.12(0.97-1.32)***
|
1.19(0.99-1.42)
|
1.16(0.97-1.39)
|
|
eGFR(ml/min/1.73m²)
|
73.62(36.92-99.39)
|
100.21(91.43-108.85)
|
91.82(73.58-103.57)***
|
82.57(64.38-99.98)***
|
25.10(12.21-39.74)***
|
|
ACR(mg/g)
|
309.06(22.96-2182.22)
|
9.24(4.96-15.91)
|
75.83(39.50-148.20)***
|
957.54(399.50-2444.80)***
|
2251.14(869.76-4418.82)***
|
|
ALB(mg/L)
|
37.90(33.30-41.80)
|
41.2(38.2-44.2)
|
40.20(37.50-43.50)***
|
36.00(30.40-40.40)***
|
34.30(28.80-38.50)***
|
|
T3(ng/mL)
|
0.90(0.77-1.04)
|
0.98(0.85-1.09)
|
0.94(0.81-1.08)
|
0.91(0.79-1.07)***
|
0.83(0.70-0.96)***
|
|
T4(μg/dL)
|
8.00(6.70-9.20)
|
8.27(7.10-9.20)
|
8.20(7.00-9.30)
|
7.79(6.60-8.98)***
|
7.90(6.50-9.20)*
|
|
FT3(pg/mL)
|
2.73(2.43-3.02)
|
2.95(2.71-3.20)
|
2.86(2.62-3.07)***
|
2.76(2.50-3.05)***
|
2.45(2.19-2.70)***
|
|
FT4(ng/dL)
|
1.15(1.02-1.28)
|
1.18(1.06-1.31)
|
1.22(1.07-1.34)*
|
1.17(1.03-1.30)
|
1.10(0.98-1.23)***
|
|
TSH(μIU/mL)
|
1.91(1.25-3.32)
|
1.70(1.10-2.65)
|
1.61(1.09-2.53)
|
2.01(1.24-3.24)***
|
2.24(1.46-4.11)***
|
|
P values are calculated by Wilcoxon rank sum tests for continuous variables expressed as median (interquartile range) andχ2 tests for categorical variables expressed as count and percent.
SBP, Systolic Blood Pressure; DBP, diastolic blood pressure;BMI, Body Mass Index;HbA1c,hemoglobin A1c;TG, serum triglyceride;TC, total cholesterol;LDL-C, low-density lipoprotein cholesterol;HDL-C, high-density lipoprotein cholesterol;eGFR, estimated glomerular filtration rate;ACR, albumin-to-creatinine ratio;ALB, albumin;T3, triiodothyronine; T4, total thyroxine; FT3, free triiodothyronine; FT4, free thyroxine; TSH, thyroid stimulating hormone. *P<0.05; **P< 0.01;***P<0.001
Correlation analyses
The correlation plot of 5 thyroid hormones under study is presented in Figure 1. There was a correlation between FT3 and T3 (Spearman correlation coefficient r: 0.611, P<0.001), FT4 and T4 (Spearman correlation coefficient r: 0.226, P<0.001). Since free thyroid hormones are the physiologically active form of thyroid hormones, only FT3, FT4 and TSH were retained in the following analysis.
Table 2 shows the correlations among serum FT3, FT4, TSH and ACR and eGFR categories. ACR and eGFR categories were important parts of the KDIGO risk categories, so we further explored the correlation between three thyroid hormones and ACR and eGFR categories.
Table2 correlation between serum FT3, FT4, TSH and ACR and eGFR catogories.
|
FT3
|
FT4
|
TSH
|
ACR catagories
|
-0.290
|
-0.112
|
0.210
|
|
P<0.001
|
P<0.001
|
P<0.001
|
eGFR catagories
|
-0.490
|
-0.178
|
0.197
|
|
P<0.001
|
P<0.001
|
P<0.001
|
Correlation between thyroid hormones, ACR and eGFR was evaluated through Spearman correlation analysis.
FT3, free triiodothyronine; FT4, free thyroxine; TSH, thyroid stimulating hormone.
Risk prediction for DKD
Table 3 shows the risk prediction of serum FT3, FT4, and TSH in different stages of DKD before and after propensity score matching(PSM). According to Bonferroni correction, P less than 0.05/9 were considered significant differences. After balancing age, gender, Hypertension, HbA1c, TC and duration of diabetes, PSM analysis showed that, the risk of DKD was significantly reduced by 13-21% for every 0.2 pg/mL increase in serum FT3 in the moderate-risk stage (OR: 0.87, 95% CI: 0.81-0.93, P<0.001) and high-risk stage (OR: 0.79, 95% CI: 0.74-0.85, P<0.001) , and reduced by 44% in the very high-risk stage(OR:0.56, 95% Cl: 0.52-0.61, P<0.001). While after PSM, Both of serum FT4 and TSH at all stages had no statistical significance.
Table 3 the risk prediction of serum FT3, FT4, and TSH in different stages of DKD
Significant risk factors
|
Low risk
|
Moderate risk
|
High risk
|
very high risk
|
FT3 (+0.2 pg/mL)
|
-
|
0.87, 0.81 to 0.93, <0.001
|
0.79, 0.74 to 0.85, <0.001
|
0.56, 0.52 to 0.61, <0.001
|
FT4 (+0.2 ng/dL)
|
-
|
1.20, 1.06 to 1.38, 0.005
|
1.04, 0.92 to 1.18, 0.491
|
0.82, 0.73 to 0.92, 0.001
|
TSH (+0.5 μIU/mL)
|
-
|
1.00, 0.99 to 1.00, 0.394
|
1.00, 0.99 to 1.00, 0.378
|
1.00, 1.00 to 1.01, 0.344
|
After balancing age, gender, hypertension, HbA1c, TC and duration of diabetes
|
FT3 (+0.2 pg/mL)
|
-
|
0.86, 0.77 to 0.95, 0.003
|
0.83, 0.76 to 0.93, 0.001
|
0.56, 0.49 to 0.66, <0.001
|
FT4 (+0.2 ng/dL)
|
-
|
1.20, 1.00 to 1.43, 0.045
|
1.08, 0.91 to 1.28, 0.354
|
0.90, 0.75 to 1.09,
0.311
|
TSH (+0.5 μIU/mL)
|
-
|
1.00, 1.00 to 1.01, 0.352
|
1.00, 0.98 to 1.01, 0.263
|
1.00, 0.99 to 1.01, 0.517
|
FT3, free triiodothyronine; FT4, free thyroxine; TSH , thyroid-stimulating hormone. Data are expressed as odds ratio, 95% confidence interval, P value.
Prediction accuracy assessment
Table 4 shows an evaluation of the accuracy of the predictions respectively by the degree to which thyroid hormones additions were understood to the basic model ( including age, gender, hypertension, HbA1c, TC and duration of diabetes). The predicted probabilities of thyroid hormones additions reflected the actual observed risk and overall adaptation to the modified risk model. In terms of calibration, the reduction in AIC and BIC statistics was greater than 10 after the addition of three thyroid hormones to the basic model at each stage. Moreover, the likelihood ratio test showed that the difference was statistically significant in FT3 in all stages .
Table 4 Prediction accuracy gained by adding thyroid hormones to basic model for DKD at different stages.
|
Statistics
|
moderate risk
|
High risk
|
very high risk
|
|
Basic model
|
Basic model plus FT3
|
Basic model
|
Basic model plus FT3
|
Basic model
|
Basic model plus FT3
|
Calibration
|
AIC
|
766.27
|
670.59
|
761.25
|
658.17
|
836.01
|
612.24
|
BIC
|
796.71
|
704.40
|
792.46
|
692.98
|
870.34
|
650.59
|
LR test (χ2 )
|
Ref.
|
12.23
|
Ref.
|
30.43
|
Ref.
|
128.88
|
LR test P value
|
Ref.
|
<0.001
|
Ref.
|
<0.001
|
Ref.
|
<0.001
|
Discrimination
|
NRI (P value)
|
Ref.
|
<0.001
|
Ref.
|
0.007
|
Ref.
|
<0.001
|
IDI (P value)
|
Ref.
|
<0.001
|
Ref.
|
<0.001
|
Ref.
|
<0.001
|
AIC, Akaike information criterion; BIC, Bayesian information criterion; LR, likelihood ratio; NRI net reclassification improvement; IDI, integrated discrimination improvement; FT3, free triiodothyronine.
Nomogram prediction model
In order to further analyze the common contribution of FT3, a nomogram prediction model was established for moderate-risk stage, high-risk stage and very high-risk stage, respectively, as shown in Figure 2. The predictive accuracy and discriminative capability were assessed by C-index ( C-index 0.673, P<0.001 for moderate-risk stage; 0.810, P<0.001 for high-risk stage; 0.907, P<0.001 for very high-risk stage), indicating significant improvement in model performance, especially in the high-risk stage and the very high-risk stage.
The nomogram prediction models were established for moderate-risk stage (panel A), high-risk stage (panel B), very high-risk stage (panel C).
SBP, Systolic Blood Pressure; HbA1c,hemoglobin A1c; FT3, free triiodothyronine; TC, total cholesterol.
From the nomogram of the predictive model, the single scores of FT3 were all greater than 80, indicating that FT3 had great weight in the model. Taking very high-risk stage as an example, assuming a patient with diabetic retinopathy (10 points), a 20-year history of diabetes (8.75 points), SBP15mmHg (25 points), TC 4.5 mmol/L (6.25 points), FT3 2.5pg/mL (63.75 points) , the probability of having very high-risk stage DKD was estimated to be over 95%.