Skin autofluorescence and the complexity of complications in patients with type 2 diabetes mellitus: a cross-sectional study

A total of 825 patients with T2DM aged 18 to 80 years old who were treated at the Hospital of Chengdu University of Traditional Chinese Medicine from January 2015 to May 2019 were enrolled in this study. The diagnosis of T2DM was based on the diagnosis and classification of diabetes mellitus by the American Diabetes Association [10]. The exclusion criteria were as follows: 1) other special types of diabetes, such as type I diabetes, mitochondrial gene mutation diabetes, gestational diabetes, and adult-type autoimmune diseases in young people; 2) malignant digestive tract tumours; 3) various acute and chronic infections; 4) mental and neurological diseases; and 5) liver and kidney failure and severe heart disease. The study was approved by the ethics committees at the Hospital of Chengdu University of Traditional Chinese Medicine (No. 2017KL-033).

All of the patients had full medical history records in the electronic system, which included all test results. The general indicators for these patients included age, sex, body mass index (BMI), fasting blood glucose (FBG), 2-h postprandial blood glucose (2-h PBG), HbA1c, fasting C-peptide and 2-h standardized postprandial C-peptide (2-h C-peptide). In addition, total cholesterol (Tch), triglycerides (TGs), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamyltransferase (GGT), blood creatinine (BCr), blood uric acid (BUA), blood urea nitrogen (BUN), cystatin-C (Cys-C), and homocysteine (HCY) were detected to estimate blood lipids, liver function, and kidney function. The Cockcroft-Gault equation was used to determine the estimated glomerular filtration rate (eGFR) [11]. Smoking/drinking behaviour was defined as smoking/drinking or never. Hypertension was defined as either 1) using anti-hypertensive medication or 2) systolic blood pressure (SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg. The diagnosis of fatty liver was primarily confirmed via ultrasonography by an experienced physician.

Upon excitation at 370 nm, AGEs have a characteristic fluorescence spectrum at 440 nm. Skin long wavelength fluorescence ~ excitation (ex)/emission (em) 370/440 nm was introduced as a surrogate marker for AGE formation and has since been widely used in both clinical and experimental studies of diabetes [12, 13]. In this study, skin AGEs were assessed by an autofluorescence reader (Hefei Institutes of Physical Science, Chinese Academy of Sciences) following a 2–3-min measurement of the left volar forearm [14‐16]. In short, an excitation light source with a peak wavelength of 370 nm was used to excite the AGEs in the skin that have fluorescence properties in a frequency range of 420–600 nm, and a broadband light source in a frequency range of 420–600 nm illuminated the skin to measure tissue absorption and scattering. Emission light, including reflected light (diffuse reflectance light) and fluorescence from the skin, was measured with a spectrometer in the 300–600-nm range through a 3 × 1 fibre bundle. AGE-related autofluorescence was determined from the fluorescence and reflected light. The final values were relative (the ratio of fluorescence intensity and diffuse reflectance light intensity), so there was no uniform range, and the range of AGE values we used was 0–150.

The diabetic complication risk score was also given by an autofluorescence reader with a built-in algorithm (Hefei Institutes of Physical Science, Chinese Academy of Sciences). The device has been certified by the CFDA (China Food and Drug Administration) and CE. The risk score is a prediction model that utilizes an algorithm to combine various biological markers of diabetic complications. In detail, many studies have shown that AGEs can cause neuropathy, nephropathy, retinopathy, and cardiovascular complications through both receptor and non-receptor pathways [17]. The higher the AGEs of the individual, the higher is the risk of complications. Second, the four questions in the embedded software regarding the clinical symptoms are as follows: numbness of the hands and feet corresponding to neuropathy [18], blurred vision corresponding to retinopathy [19], hypertension and hyperlipidaemia corresponding to cardiovascular disease [20], and body oedema corresponding to kidney disease [21]. Third, some common tissue spectral features were used to evaluate complications, for example, the ratio of reflectance at 390 and 360 nm, deviation of UV reflectance from a straight line, ratio of reflectance at 470 and 500 nm, ratio of emission at 470 and 500 nm, and ratio of emission at 470 and 570 nm. Multiple logistic regression analysis was used to determine the β coefficient of the AGE value, questionnaire and characteristic spectrum. Finally, the total score range was converted to 0–100.

All measurements were performed at room temperature in a semi-dark environment by trained nurses. For every patient, the fluorescence of the skin was measured 3 times on the volar side of the arm on normal skin without visible vessels, scars, lichenification, or other skin abnormalities.

Major T2DM complications in this study included diabetic retinopathy (R), diabetic kidney disease (K), cardiovascular disease (C) and diabetic peripheral neuropathy (N). Diabetic retinopathy was diagnosed with five grades by professional ophthalmologists in accordance with the Global Diabetic Retinopathy Project G 2002 [22], including 0 (no), 1 (mild), 2 (moderate), 3 (severe), and 4 (proliferative). In this study, as a result of the sample size, we divided the classification into three groups for better comparison: 0, ≤ 2, and ≥ 3. Diabetic kidney disease was diagnosed when UACR ≥30 mg/g/Cr and/or eGFR ≤30 ml/min/1.73 m² according to the Classification of Diabetic Nephropathy 2014 [23]. Cardiovascular disease was defined as a history of ischaemic heart disease and/or a history of percutaneous coronary intervention or coronary bypass surgery according to International Classification of Diseases (ICD) codes I20–25. Diabetic peripheral neuropathy includes 1) neuropathy at or after the diagnosis of diabetes; 2) clinical signs and symptoms consistent with the performance of DPN; 3) clinical symptoms (pain, numbness, etc.); or 4) one abnormality of the five examinations (ankle reflexes, discrimination, vibration perception, pinprick sensation, or 10-g monofilament) with symptoms or two without symptoms.

A total of 825 patients with T2DM were included in this study. All of the participants were classified into five groups based on their number of complications (74 patients in group 0, 133 in group 1, 201 in group 2, 275 in group 3, and 142 in group 4). From groups 0 to 4, the median age increased from 44 to 68 (P <  0.001). Similarly, T2DM duration, categorized by years as 0–5, 10–15, 15–20, and > 20, was significantly different among the five groups (P <  0.05). However, no significant difference was apparent for years 5–10 (P = 0.255). Additionally, there was no difference in the history of smoking or drinking among the five groups. For the laboratory tests, Tch (P = 0.013), TGs (P <  0.001), ALT (P = 0.028), BCr (P <  0.001), BUA (P <  0.001), Cys-C (P = 0.025), and eGFR (P <  0.001) were significantly different among the groups, but there was no significant difference in FBG, 2-h PBG, fasting C-peptide, 2-h C-peptide, HbA1c, HDL-C, LDL-C, AST, GGT, BUN, or HCY. Expressly, the SAF levels were significantly different among the five groups, and the median SAF value increased from group 0 to group 4: 76.35 (95% CI: 71.45–80.15) in group 0; 79.6 (72.45–89.9) in group 1; 84.7 (77.05–94.7) in group 2; 86.6 (76.7–96.2) in group 3; and 91.85 (81.63–102.5) in group 4 (P <  0.001). There were also significant differences among the four types of risk scores associated with the complications (Table 1).

Univariate linear regression showed an association between SAF and age (R = 0.479, P <  0.001), sex (R = − 0.070, P = 0.045), smoking (R = 0.140, P <  0.001), drinking (R = 0.151, P <  0.001), T2DM duration (R = 0.260, P <  0.001), hypertension (R = 0.134, P <  0.001), and fatty liver (R = 0.103, P = 0.021) as well as the levels of 2-h PBG (R = 0.141, P = 0.007), 2-h C-peptide (R = − 0.140, P = 0.007), TGs (R = − 0.136, P <  0.001), ALT (R = − 0.143, P <  0.001), BCr (R = 0.387, P <  0.001), BUA (R = 0,178, P <  0.001), Cys-C (R = 0.243, P <  0.001), and eGFR (R = − 0.301, P <  0.001). However, BMI, FBG, fasting C-peptide, HbA1c, Tch, HDL-C, LDL-C, AST, GGT, BUN, and HCY showed no relationship with SAF levels. Moreover, multivariate analysis indicated that age (β = 0.389, P <  0.001), sex (β = − 2.221, P = 0.004), 2-h C-peptide (β = − 0.182, P = 0.017), ALT (β = − 0.158, P = 0.041), BCr (β = 0.206, P = 0.009), and fatty liver (β = 0.161, P = 0.026) were independent determinants of SAF (Table 2).

ANOVA tests were performed to compare the SAF values of enrolled patients with different numbers of complications, and the results showed that the SAF values of patients with 1, 2, 3, and 4 complications were significantly increased compared with those with 0 complications. Moreover, there were significant differences between the patients with 1 vs. 2 and 3 vs. 4 complications (P <  0.05), while there was no significant difference between 2 and 3 complications, as shown in Fig. 1a. In addition, we compared the SAF values of no complications and each individual complication (0 vs. R, diabetic retinopathy; 0 vs. K, diabetic kidney diseases; 0 vs. N, diabetic peripheral neuropathy; 0 vs. C, cardiovascular disease), and the P-values among the four groups were <  0.001, as shown in Fig. 1b. Logistic regression in the unadjusted model indicated that the SAF value had a significant association with the number of complications in reference to no complications (P <  0.001). After adjusting for age, sex, and BMI in model 2, the P-value for the trend was 0.007. Similarly, we calculated the variables, including FBG, HbA1c, C-peptide, history of smoking, drinking, and hypertension, in model 3, and this showed that the P-value for the trend was 0.023; however, after adding TGs, BUA, BCr, and eGFR to model 4, there was no significant association (P = 0.129), as shown in Table 3. In addition, the correlation between SAF and certain types of T2DM complications was analysed after adjusting for the relative risk factors, as presented in Table 4. In the unadjusted model, SAF had a significant relationship with only diabetic kidney disease, diabetic cardiovascular disease, and diabetic neuropathy (P <  0.001), but SAF demonstrated no relation with diabetic retinopathy. Nevertheless, after adjusting for the associated variables in model 2, there was no significant association between SAF and diabetic cardiovascular disease. Ultimately, diabetic kidney disease (P = 0.007) and diabetic neuropathy (P <  0.001) showed a statistically significant connection with SAF.

The SAF test reported four risk scores for corresponding complications, and the evaluation effect between each complication and no complications was calculated via Mann-Whitney and ANOVA tests (Fig. 2). For diabetic retinopathy, there were significant differences between non-diabetic retinopathy and retinopathy grade ≥ 3 (P = 0.003) as well as between grade ≤ 2 and ≥ 3 (P <  0.001), but there was no difference between non-diabetic retinopathy and grade ≤ 2 (P = 0.073). Additionally, there was a difference between non-diabetic kidney disease and diabetic kidney disease (P <  0.001), and the same result was observed in diabetic cardiovascular disease (P = 0.037). It should be noted that there were significant differences between grade 1 vs. grade 2 (P <  0.001) and grade 2 vs. grade 3 (P = 0.01) with regard to diabetic peripheral neuropathy, while only grade 3 diabetic peripheral neuropathy showed a significant difference compared with non-diabetic peripheral neuropathy (P <  0.001).