The Association of Skin Intrinsic Fluorescence With Type 1 Diabetes Complications in the DCCT/EDIC Study

Abstract

OBJECTIVE

To determine whether skin intrinsic fluorescence (SIF) is associated with long-term complications of type 1 diabetes (T1D) and, if so, whether it is independent of chronic glycemic exposure and previous intensive therapy.

RESEARCH DESIGN AND METHODS

We studied 1,185 (92%) of 1,289 active Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) participants from 2010 to 2011. SIF was determined using a fluorescence spectrometer and related cross-sectionally to recently determined measures of retinopathy (stereo fundus photography), cardiac autonomic neuropathy (CAN; R-R interval), confirmed clinical neuropathy, nephropathy (albumin excretion rate [AER]), and coronary artery calcification (CAC).

RESULTS

Overall, moderately strong associations were seen with all complications, before adjustment for mean HbA_1c over time, which rendered these associations nonsignificant with the exception of sustained AER >30 mg/24 h and CAC, which were largely unaffected by adjustment. However, when examined within the former DCCT treatment group, associations were generally weaker in the intensive group and nonsignificant after adjustment, while in the conventional group, associations remained significant for CAN, sustained AER >30 mg/24 h, and CAC even after mean HbA_1c adjustment.

CONCLUSIONS

SIF is associated with T1D complications in DCCT\EDIC. Much of this association appears to be related to historical glycemic exposure, particularly in the previously intensively treated participants, in whom adjustment for HbA_1c eliminates statistical significance.

The micro- and macrovascular complications of type 1 diabetes (T1D) are thought to result, primarily, from exposure of tissues to high glucose concentrations and to be reduced by intensive glycemic control (1,2). Although the pathogenesis is poorly understood, one mechanism of hyperglycemia-associated tissue damage appears to operate through enhanced formation and accumulation of advanced glycation end products (AGEs) (3,4). AGEs are the stable final products of complex reactions, many involving free radical oxidation, between reducing sugars and free amino groups in proteins. AGE formation is considered a stoichiometric, nonenzymatic process that is enhanced in diabetes not only by elevated glucose concentrations, but also by increased oxidative stress (3,5). AGEs include glucose- or carbohydrate-derived adducts to proteins, and some include cross-links within and between protein molecules.

Numerous studies have focused on AGE modification of skin collagen (6–16) as an accessible, long-lived protein that may reflect cumulative AGE-mediated tissue damage elsewhere. Skin AGE content accumulates on long-lived proteins with advancing chronological age and is accelerated in diabetes (17). In a study of skin biopsy samples collected near the end of the Diabetes Control and Complications Trial (DCCT) (from 1992 to 1993), levels of AGEs in collagen were lower in the intensive treatment group compared with the conventional group (8). Furthermore, in a more recent analysis of the same tissue samples, the initial glycation product (assessed as furosine content) and the AGE carboxymethyllysine were shown to predict the progression of both retinopathy and nephropathy independently of HbA_1c levels (13). As measures requiring skin biopsy are obviously not generally applicable in the clinical setting, a noninvasive, surrogate measure of AGE content reflecting skin collagen might be of great potential value.

Some AGE products fluoresce when excited with near-ultraviolet and blue light, and measures of fluorescence in tissue biopsies, including skin collagen, have been found to correlate with levels of specific AGE products (r = 0.55) (10,11). It is thus possible that skin fluorescence may act as a surrogate marker for AGE content and provide an opportunity for noninvasive measurement of these products (6,8,10,11). It should, however, be recognized that skin fluorescence has multiple determinants beyond AGE and thus may also provide additional information concerning risk for complications. The SCOUT DS instrument noninvasively measures AGE-related, and other, fluorescence in human skin, called skin intrinsic fluorescence (SIF), controlling for subject-specific light scattering and skin melanin and hemoglobin content (10,11).

Preliminary SIF data from the Pittsburgh Epidemiology of Diabetes Complication (EDC) study showed fairly strong associations with both neuropathy and coronary artery calcification (CAC) in a long-duration T1D cohort (18–20). While the cross-sectional association of SIF with the presence of these complications was superior to that based on historical HbA_1c measures (up to seven values over 18 years of follow-up), the documentation of glycemic exposure was insufficient to assess whether any association of SIF with complications was fully independent of prior glycemia.

In the present article, we thus assess the association of SIF with long-term T1D complications and, using the detailed glycemic history and larger sample size of the DCCT/Epidemiology of Diabetes Interventions and Complications (EDIC) cohort, determine whether the relationship is independent of chronic glycemia and renal function. We also determine whether there are any differences in these associations by DCCT cohort (primary prevention vs. secondary intervention) and treatment group (intensive therapy vs. conventional therapy).

RESEARCH DESIGN AND METHODS

Study sample

The inclusion and exclusion criteria for the DCCT and the treatment protocol have been described in detail (21). Briefly, 1,441 subjects with T1D between 13 and 39 years of age were recruited into the DCCT between 1983 and 1989; 53% were male. The primary prevention cohort consisted of 726 subjects who, at study baseline, had no retinopathy, a urinary albumin excretion rate (AER) <40 mg/24 h, and diabetes duration of 1–5 years. The secondary intervention cohort consisted of 715 subjects who had very mild to moderate nonproliferative retinopathy, urinary AER ≤200 mg/24 h, and diabetes duration of 1–15 years. As part of the screening for the DCCT, individuals were excluded if they had hypertension, a history of symptomatic ischemic heart disease, the presence of major electrocardiogram abnormalities, or severe hypercholesterolemia. Subjects were randomly assigned to either intensive (n = 711) or conventional (n = 730) treatment arms and assessed for complications at frequent follow-up visits. Intensive therapy included either multiple (at least three) daily insulin injections or continuous subcutaneous insulin infusion with external insulin pumps, with the goal of achieving glycemic control as close to the nondiabetic range as safely possible, specifically, HbA_1c <6.05% guided by frequent daily self–blood glucose monitoring. The intensive and conventional treatment groups maintained median HbA_1c levels of ∼7.0 and 9.0%, respectively, during the 6.5-year mean DCCT follow-up.

In 1994, after completion of the DCCT, 1,375 subjects (96% of the surviving cohort), 688 from the conventional arm and 687 from the intensive arm, agreed to participate in the EDIC follow-up study, which included annual examinations measuring diabetes complications (22).

For the current SIF complications analyses, all living EDIC subjects who participated in the annual exam during years 16 or 17 of EDIC (2010 to 2011) were eligible for inclusion. Ninety-two percent (1,185) of the 1,289 active EDIC participants had SIF measured.

Clinical measures

Demographic data and health history were self-reported, and a standardized physical examination was performed annually. BMI was measured every 3 months during the DCCT and yearly during the EDIC Study. All laboratory measurements were performed at the DCCT/EDIC Central Biochemistry Laboratory at the University of Minnesota as previously described (21). HbA_1c was measured every 3 months during the DCCT and yearly during the EDIC Study (21,22). AER was measured annually during the DCCT and on alternate years during the EDIC study using a timed 4-h urine collection and expressed per 24 h (21,22). Serum lipids were measured using conventional enzymatic methods from fasting samples obtained yearly during DCCT and on alternate years during EDIC. Serum creatinine was measured annually in DCCT/EDIC and estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation (23,24).

Total glycemic exposure (mean HbA_1c) was calculated as: (pre-DCCT: DCCT eligibility HbA_1c × duration of diabetes at study baseline) + (DCCT mean HbA_1c × years of follow-up in DCCT) + (EDIC mean HbA_1c × years of follow-up in EDIC). Glycemic exposure was also examined using each of these time periods separately.

Clinic latitude was determined as a surrogate for potential differences in vitamin D levels due to sun exposure and was incorporated into the data analysis as a categorical variable with EDIC clinics below and above 37° latitude designated as southern (n = 9) and northern clinics (n = 19), respectively (25–28). Smoking status was determined by subject self-report and categorized as never smoked (<100 cigarettes in a subject's lifetime), previous smoker (quit >1 year ago), or current smoker.

SIF measurement

Repeat measurements of SIF were obtained from the skin on the underside of the left forearm near the elbow using the SCOUT DS skin fluorescence spectrometer. SIF was excited with a light-emitting diodes centered at 375, 405, 416, 435, and 456 nm and was detected over the emission range of 435–655 nm. For these analyses, the 375-nm excited fluorescence was used, with k_x set to 0.6 and k_m set to 0.2. The resulting intrinsic fluorescence, f_xm, was integrated over the 435–655-nm spectral region and multiplied by 1,000 to give the SIF sum. These values of k_x and k_m were previously determined in the Pittsburgh EDC study to be relevant for the 375-nm excited fluorescence, which had the strongest association with diabetes-related complications in the Pittsburgh EDC cohort (19). The intrasubject, same-day variation in SIF was assessed using the method of the Hoorn Study (18). The intraday Hoorn coefficient of variation was 4.2%, and the between-measurement correlation was 0.963.

Definition of complications

The complications of diabetes reported in this article are retinopathy, nephropathy, neuropathy, and CAC. Each participant was categorized according to the presence or absence of each of these complications at his or her most recent clinical assessment.

Retinopathy.

Presence or absence of moderate nonproliferative diabetic retinopathy (NPDR) or worse indicated by microaneurysms plus immunoradiometric assay or moderate retinal hemorrhages (Early Therapy Diabetic Retinopathy Study [ETDRS] score of ≥6) between EDIC years 13–16. Subjects who received pan-retinal scatter photocoagulation (laser) therapy in either eye were counted as having the most severe level of retinopathy thereafter. Retinopathy was measured by standardized seven-field fundus photography biannually during DCCT. During EDIC, it was assessed with identical methods, but in approximately one-quarter of the cohort each year and in the entire cohort at EDIC years 4 and 10. All photographs were graded centrally using the final ETDRS grading scale (29,30) and DCCT methods with the graders masked to DCCT therapy assignment (21,30).

Nephropathy.

Presence or absence of sustained AER ≥30 mg/24 h (i.e., at two consecutive visits at EDIC years 13/14 and 15/16) (31,32).

Neuropathy.

Two measures of neuropathy were used. First, the presence or absence of cardiac autonomic neuropathy (CAN) was assessed by measuring sinus arrhythmia (33). Electrocardiographic R-R variation was computed as a dimensionless circular mean vector of R-R intervals (33). Abnormal R-R variation was defined as <15 in testing done in EDIC years 16/17 (33). Secondly, presence or absence of confirmed clinical neuropathy (CCN) was defined as in previous DCCT/EDIC publications (34,35). The data used for this assessment were obtained from nerve conduction tests done in EDIC years 13/14 (35).

CAC was detected with either multislice or electron beam computed tomography, as previously described (36). We defined categories of CAC (present/absent) using thresholds of CAC >0 and CAC >200 Agatston units from testing done in EDIC year 12.

Statistical analyses

Characteristics were compared using the Wilcoxon rank sum test to evaluate treatment group differences for ordinal and numeric variables. The contingency χ² test was used for categorical variables.

SIF was log transformed, and the first SIF measurement per subject was used for all analyses. The association between log SIF and the most recent complication status was modeled using logistic regression adjusting for variables that have been previously shown to be related to SIF and were statistically significant (37). These comprised age, any eGFR <60 mL/min/1.73 m² (categorical adjustment to control for retained AGE products as a consequence of renal function impairment), smoking status (never, ever, or current), skin tone, and clinic latitude. Hypertension and lipid concentrations were not related to SIF (37). Further adjustments were made for total glycemic exposure. Odds ratios (ORs) and 95% CIs are presented per 1 SD (0.21) change in log SIF. Interactions between treatment group and log SIF were also assessed.

RESULTS

At the time of SIF determination, the population had a mean age of 51.5 years and diabetes duration of 29.8 years (Table 1). As previously reported, the DCCT former intensive therapy group had a lower prevalence of most complications despite current and prior mean HbA_1c being similar between the two groups starting 5 years from the end of the DCCT in 1993, although no difference is seen for end-stage renal disease and CAC. There was no difference in SIF by former treatment group. Compared with the few (N = 104) active DCCT/EDIC participants without an SIF measure, the examined population showed no major differences in baseline age (P = 0.97) and duration of diabetes (P = 0.64). Nonparticipants had a higher mean DCCT/EDIC/pre-DCCT HbA_1c (8.4 ± 1.0 vs. 8.2 ± 0.9; P = 0.02).

Table 1.

Characteristics of the DCCT/EDIC SIF population at time of SIF assessment overall and by original DCCT treatment group

Table 2 presents the ORs for each complication per 1 SD change in log SIF after adjustment for age, any eGFR <60 mL/min/1.73 m², smoking status, skin tone, and clinic latitude with and without adjustment for total glycemic exposure. Data are presented overall and stratified by DCCT treatment group. Overall, a moderately strong positive risk for each complication was seen in the analyses unadjusted for total glycemic exposure, although magnitude varies (ORs range from 1.15 [CAC >200] to 1.87 [sustained AER ≥30]). These associations were a little higher if eGFR was not included as a covariate (e.g., for sustained AER, the OR unadjusted for mean HbA_1c was 2.19 [95% CI 1.79–2.67] compared with 1.87 [1.52–2.31]) (Table 2). All of the overall associations were significant except those for CAC >200, for which the prevalence was low (7%). After adjustment for total glycemic exposure, all ORs were markedly reduced, although those for sustained AER ≥30 and CAC >0 remained significant (P < 0.05). When analyzed as continuous outcomes, similar significant associations were seen; however, log (and square root) CAC were now significantly associated with SIF.

Table 2.

Log SIF association with most recent complication status overall and by original DCCT treatment group

Subtle differences emerged on stratification by treatment group, although generally, treatment group by log SIF interactions were significant. In the former intensive therapy group, the strength of the glycemic unadjusted associations with CAN, CCN, and sustained AER ≥30 was somewhat lower than seen for the former conventional group. For CAN, the interaction term for treatment group was significant (P < 0.05). In contrast, the associations with retinopathy and CAC were similar. When analyzed continuously, all associations between SIF and CAN, sustained AER, and both CAC transformations were significant for both treatment groups, except for CAN and square root CAC in the former intensive group.

The overall effect of adjustment for total glycemic exposure showed a reduction in the ORs to borderline or no significance for the retinopathy, neuropathy, and nephropathy markers except for sustained AER ≥30 (overall and conventional group) and CAN (conventional group), where significance was retained. The interaction term for treatment group remained significant for CAN (P < 0.05). The magnitude of the adjustment effect was fairly consistent across treatment groups. However, the relationship between SIF and CAC appeared to be little affected by the HbA_1c adjustment in either group. When examined continuously, similar patterns were seen by former treatment group, although in the intensive group, neither log or square root CAC was significantly associated with log SIF.

We also simultaneously examined the association between glycemia adjusted log SIF and complications by both DCCT primary prevention versus secondary intervention cohort and treatment group (Table 3). In the primary prevention cohort, log SIF associations with the microvascular complications were generally low (and nonsignificant) in the former intensive therapy group and stronger and significant for CAN and sustained AER ≥30 in the conventional group. The treatment group interaction term for CAN remained significant (P = 0.04). For CAC, little difference is seen between former treatment groups, and all associations were nonsignificant. In the secondary intervention cohort, for retinopathy, the relative direction of the associations was different (i.e., now stronger in the former intensive group) (OR 1.23 vs. 0.89), though none of these potential group interactions were significant. Furthermore, in the secondary cohort, no ORs were significant for any complication in either treatment group.

Table 3.

Log SIF association with most recent complication status by original DCCT cohort assignment and treatment group

When analyzed continuously, in the primary prevention cohort, log AER was significantly associated with log SIF in the conventional group. In the secondary intervention cohort, significant associations were seen for log AER in both groups and for CAC in the conventional group.

CONCLUSIONS

These results demonstrate an association of SIF with complications of T1D in the well-characterized DCCT/EDIC cohort. Generally, fairly strong univariate correlations were seen for all microvascular complications included with weaker associations for CAC. However, on adjustment for total glycemic exposure, the microvascular complication associations were totally eliminated in the former intensive therapy group and remained significant only for CAN and sustained AER in the former conventional group. The weaker CAC associations were not further affected by controlling for glycemic exposure. These differences by DCCT treatment group appear complex, but fairly small and significant interactions were found only for CAN. Finally, it should be noted that there is no current difference between the former DCCT intensive and conventional groups in SIF, which was measured 16 to 17 years after the large separation in HbA_1c between the groups ended.

SIF is not simply a function of glycemic exposure and AGE formation, but may reflect many other factors, some of which we are able to control, such as skin fluorescence/pigment and hemoglobin levels. Given these multiple determinants, SIF may potentially reflect more than glycemic exposure and AGE formation. However, the current data suggest that any added information is largely linked to renal disease as marked by sustained AER >30 mg/24 h and possibly CAN, marked by reduced R-R variation. The former association might be postulated to reflect reduced renal function, for AGEs that are not bound to long-living proteins, like collagen, may be retained and thus promote greater tissue damage, including greater binding to long-living proteins, and thus contributing to SIF. However, it should be noted that the analyses in Tables 2 and 3 did control for an estimate of GFR (which did indeed reduce the associations), and thus, it is likely log SIF is truly associated with renal damage and autonomic neuropathy beyond renal function per se and glycemic exposure.

The observation that intensive and conventional do not currently differ in terms of SIF is not surprising given the long recent period during which these two groups have had similar HbA_1c levels. This similarity in SIF between conventional and intensive groups agrees with similar levels of specific AGE products and fluorescence between the two groups in a much smaller set of skin biopsies from DCCT/EDIC obtained from 1997 to 1998 (5). However, biopsies done by Monnier et al. (8) near DCCT closeout in 1991 to 1992 did show significantly higher AGE product levels in skin collagen from the conventional versus the intensive group. The observations that SIF have moderately strong associations with complications have clinical implications, despite their being largely, but not totally, eliminated by controlling for lifetime glycemic control. Firstly, it suggests that an SIF measure may be helpful in quantifying prior glycemic exposure—for example, when a patient is being assessed without a lifetime of available HbA_1c measures. We have previously reported that SIF shows significant correlations with mean HbA_1c over prior periods up to 15 years (37). Secondly, because SIF still retains some additional, albeit borderline, associations after glycemic exposure adjustment, further investigation may permit potential new pathways to complications to be identified and quantified. It is also likely that the associations are being underestimated, as there are imperfections in the current SIF measures, including the superimposition of epidermal fluorescence from NADH and flavin adenine dinucleotide (9,38,39) on dermal AGE fluorescence and interday, within-subject, measurement variance. We have previously tested to see if instrument bias was a significant factor in explaining SIF variance using two different methods. Neither test showed instrument bias to be a significant contributor to the overall variance in SIF (37). As NADH and flavin adenine dinucleotide are not as stable as AGEs and fluctuate in response to changes in epidermal cell metabolism and oxidative stress (9,38), their superimposition on the dermal AGE signal could reduce the quantitative relationship of glycemic exposure to SIF because the SIF algorithm only takes into account the overall intensity of the fluorescence and not the spectral shape.

AGE formation also reflects oxidative stress, and thus, there is further potential for SIF to contribute to our understanding of the pathogenesis of diabetes complications if it also reflects this important component. In previous studies (5,7), we have shown with advanced chronological age, the rate of accumulation of specific AGE products in skin collagen varies markedly even in the absence of diabetes. Since there is little variation in average glycemia among nondiabetic people, this variation has been attributed to variations in oxidative stress or antioxidant defenses (5). SIF thus has the potential to provide an integrated measure of stresses imposed by aging, glycemic exposure, and free-radical oxidation, all of which are interrelated.

There are few studies to which these results can be compared. However, the Pittsburgh EDC study of T1D has, using the first generation of the SCOUT device, examined the association of SIF with both coronary calcification and neuropathy (19,20). In the former reports, SIF was calculated from a subset of fluorescence wavelengths to avoid a spectral artifact that was present in the first-generation device but subsequently corrected in the second-generation device used in this study. A subsequent comparison of SIF measured by the first- and second-generation SCOUT algorithms found them to be highly correlated, with an R² of 0.97 (J.M., personal communication).

In the EDC study, strong SIF associations with CAC were noted cross-sectionally, and meaningful improvements in the receiver operating characteristic analyses were obtained. However, it was noted that the associations became stronger as CAC advanced, and generally, this population had a somewhat longer duration, though similar age, reflecting their childhood onset. Strikingly, their CAC scores were much higher, with >70% having some measurable CAC as opposed to only 30% in DCCT/EDIC. It thus is possible that the SIF–CAC association may be stronger in those with more advanced duration and CAC, as seen in EDC.

In terms of neuropathy, EDC also studied both CAN and distal symmetric polyneuropathy and again found strong associations with both measures independent of 18-year mean HbA_1c. Indeed, for CAN, the receiver operating characteristic for SIF was 0.8, while for mean HbA_1c, it was only 0.57. Similar disparity was seen for distal symmetric polyneuropathy (0.78 vs. 0.59). Our CAN results are thus consistent with these EDC findings and with prior data showing AGE accumulation in neural tissue (39). Though analytic measures are different, it does seem SIF associations are weaker in DCCT/EDIC than in EDC when assessed in terms of incremental association after accounting for historic glycemic control. A major factor beyond the differing duration is the far greater detailed assessment of historic control (HbA_1c), both in terms of number of measures and the proportion of diabetes duration covered by HbA_1c values. There are a number of strengths and weaknesses in the current analyses. While the large sample size of DCCT/EDIC and the detailed glycemic assessments are great strengths, the relatively mild complication status of the population at present means that there is low representation of advanced stages of complications in which stronger associations may be seen. Another major limitation is the cross-sectional nature of the analysis, particularly as some complications were assessed up to 3 or 4 years earlier than SIF (e.g., CAC), and thus, their status may have changed. Prospective follow-up of this valuable cohort should partially address these concerns.

In conclusion, these analyses have shown moderately strong associations between SIF and a number of markers of diabetes complications in the DCCT/EDIC cohort despite the SIF measures not being determined, in some instances, at the same time as complication assessment. The associations appear weaker in the former intensive therapy group. Though history of glycemic exposure largely eliminates these associations in the former intensive therapy group, they remain significant for CAN and sustained AER in the former conventional group. Prospective studies are, however, needed to fully evaluate the predictive value of SIF for complications. Further investigations to understand the pathophysiologic basis for these associations and the suggestive treatment group differences are also warranted.