Recent trends over time in vascular disease in type 1 diabetes: insights from the Pittsburgh Epidemiology of Diabetes Complications study

Abstract

Aim 

The aim of this study was to evaluate recent trends, by diagnosis year, in the cumulative incidence of vascular type 1 diabetes complications and their risk factors up to 35 years duration.

Participants and methods 

Participants from the Epidemiology of Diabetes Complications study of childhood-onset type 1 diabetes were categorized into three-calendar year onset cohorts: 1965–1969, 1970–1974, and 1975–1980. All-cause, cardiovascular and renal mortality were determined both on the complete cohort (n=440) as well as in the subset with clinical examinations (n=363). Coronary artery disease (CAD, CAD death, myocardial infarction, revascularization, angina, and ischemic ECG abnormalities); lower extremity arterial disease (abnormal ankle brachial index/amputation/intermittent claudication); distal symmetric polyneuropathy (clinical exam); cardiac autonomic neuropathy (abnormal heart rate response to deep breathing); overt nephropathy (albumin excretion rate>200 μg/min); and proliferative retinopathy (fundus photography) were assessed on the subset with clinical examinations.

Results 

While all-cause, cardiovascular and renal mortality improved with a more recent diagnosis in the overall cohort, no univariate differences were observed for complications requiring examination (CAD, distal symmetric polyneuropathy, cardiac autonomic neuropathy, and overt nephropathy), although a borderline trend toward improved complication-free survival with a more recent onset was observed for proliferative retinopathy (P=0.06). Lower extremity arterial disease incidence showed an anomalous pattern, being highest in the 1975–1980 and lowest in the 1970–1974 cohort (P=0.009). Allowing for risk factors over time did not materially change these results.

Conclusion 

Despite declines in all-cause, cardiovascular and renal mortality in the total cohort, no improvements were observed in either mortality or complication incidence with a more recent onset in the examined cohort.

Introduction

From antiquity and until the discovery of insulin in 1921, survival among patients with childhood-onset type 1 diabetes (T1D) hardly exceeded 2.5 years, while an adult diagnosis afforded only an additional 2–7 years ¹. Following the discovery of insulin, treatment with exogenous insulin replacement changed these statistics, dramatically prolonging life expectancy. However, the improved survival revealed a high risk of physical and psychological long-term complications in this condition. More recently, the landmark findings from the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) study have clearly shown that intensive diabetes management can reduce the development and progression of both microvascular ² and macrovascular ³ complications. Nevertheless, the extent to which these significant research advances have led to the delay or elimination of the long-term complications of T1D is unclear.

In this manuscript, we, therefore, assess recent trends over time in mortality, microvascular, and macrovascular complications among individuals diagnosed with childhood-onset T1D. We document successes and failures in curbing the incidence of these adverse health outcomes across three-calendar year subcohorts diagnosed over a 15-year period, and explore reasons underlying similarities or differences in disease trajectories. Our assessment is based on data derived from the Epidemiology of Diabetes Complications (EDC) study, a prospective investigation of individuals diagnosed with T1D before their 17th birthday and who has been shown to be representative of the T1D population of Allegheny County (Pennsylvania, USA) ⁴.

Participants and methods

The EDC study is a historical-prospective investigation of a representative cohort of childhood-onset T1D diagnosed, or seen within 1 year of diagnosis, at Children’s Hospital of Pittsburgh between 1950 and 1980 ⁵. Of 1124 individuals eligible to participate in the study, 145 (12.9%) died before study initiation (1986–1988), 191 (17%) declined participation, 130 (11.6%) provided only survey information and have been followed less frequently, and 658 (58.5%) participated in biennial surveys throughout follow-up and periodic clinical examinations (biennially for 10 years, and subsequently at the 18, 25, and 30-year of follow-up). Death certificates were obtained for 144 (99%) of the 145 predeceased individuals. The University of Pittsburgh Institutional Review Board approved the study protocol.

The present analyses highlight results based on study participants whose T1D diagnosis occurred during 1965–80 (n=440), as they would have had the majority of their diabetes duration in the era of Hba1c and self-monitoring of blood glucose and a majority of their diabetes duration since the DCCT results, which led to intensification of diabetes management. In this subgroup, complication incidence was assessed for up to 35 years of T1D duration, which was achieved by essentially all study participants by the 25-year follow-up examination.

Mortality assessment

Vital status was assessed as of 31 August 2014. In addition to death certificates, medical records, autopsy/coroner’s reports, and/or interviews with next-of-kin regarding the circumstances surrounding the death were also obtained when possible. Data were reviewed by a physician mortality classification committee (chaired by T.J.O.), which classified deaths according to standard procedures ⁶.

Vascular complication assessment

Coronary artery disease (CAD) was determined by study-physician diagnosed angina, myocardial infarction confirmed by Q-waves on electrocardiogram (Minnesota codes 1.1 or 1.2) or hospital records, angiographic stenosis of at least 50%, revascularization, ischemic ECG changes (Minnesota codes 1.3, 4.1–4.3, 5.1–5.3, 7.1), or CAD associated death. Lower extremity arterial disease (LEAD) was defined by a history of amputation for a vascular cause, intermittent claudication (Rose Questionnaire), or an abnormal ratio of ankle/brachial systolic blood pressures (<0.9) at rest. As previously validated by ankle radiography, an ankle–brachial difference of systolic blood pressure of 75 mmHg was taken to indicate medial wall calcification. Ratios were calculated individually for each ankle artery (posterior tibial and dorsalis pedis) in both legs. The presence of an abnormality in any of the resulting four ratios was considered positive. Distal symmetric polyneuropathy (DSP) was defined as experiencing at least two out of three of the following: symptoms consistent with DSP; sensory and/or motor signs; and absent/reduced tendon reflexes based on the DCCT clinical exam protocol ⁷. Cardiac autonomic neuropathy (CAN) was defined as a history of an abnormal heart rate response to deep breathing (expiration–inspiration ratio<1.1). Serum and urinary albumin were obtained through immunonephelometry ^8,9. Albumin excretion rate greater than 200 µg/min (>300 mg/24 h) in at least two of three validated timed biennial urine collections defined overt nephropathy (ON). In 10% of urine collections deemed inadequate based on creatinine excretion, as well as in the 25-year exam where albumin excretion rate was not assessed, albumin to creatinine ratio was used, and ON was defined as more than 30 mg/g ⁹. Proliferative retinopathy (PR) was assessed with stereoscopic fundus photographs of fields 1, 2, and 4 with a Zeiss camera, which were read by the University of Wisconsin-Madison Fundus Photography Reading Center and categorized based on the modified Arlie House system ¹⁰.

Risk factor assessment

Before each clinical visit, participants were sent surveys with regard to demographics, healthcare, diabetes self-care, and medical history information. BMI was calculated as weight (kg) divided by height (m²). Intensive insulin therapy was defined as multiple (i.e. three or more) daily insulin injections or the use of an insulin pump, as well as checking glucose 28 times or more weekly. During the clinical visit, participants underwent evaluations to assess anthropometrics as well as complication status and further provided a fasting blood sample and urine samples. Blood pressure was measured with a random zero sphygmomanometer, according to the Hypertension Detection and Follow-up Program protocol, after a 5-min rest ¹¹ and hypertension was defined as at least140/90 mmHg or use of medications for high blood pressure. During the first 18 months of follow-up, HbA₁ was measured by ion-exchange chromatography (Isolab, Akron, Ohio, USA), after which automated high-performance liquid chromatography was used (Diamat; Bio-Rad, Hercules, California, USA). The two assays were highly correlated (r=0.95). For follow-up beyond the first 10 years, HbA_1c was measured with the DCA 2000 analyzer (Bayer, Tarrytown, New York, USA). The DCA and Diamat assays were also highly correlated (r=0.95). The original HbA₁ values (1986–1998) and HbA_1c values (1998–2004) were converted to DCCT-aligned HbA_1c values via a regression equation derived from duplicate assays [DCCT HbA_1c=(0.83×DIAMAT HbA₁)+0.14 and DCCT HbA_1c=(DCA HbA_1c–1.13)/0.81]. HDL cholesterol was determined by a precipitation technique (heparin and manganese chloride) with a modification ¹² of the Lipid Research Clinics method ¹³. Total cholesterol was measured enzymatically ¹⁴ and nonhigh density lipoprotein (non-HDL) cholesterol was calculated as total minus HDL cholesterol.

Statistical analysis

All analyses were conducted using the SAS statistical software (SAS Institute Inc., Cary, North Carolina, USA). Participant characteristics were compared across diagnosis cohorts (i.e. 1965–1969, 1970–1974, and 1975–1980) using χ²-tests for classification variables and analysis of variance for continuous variables. Procedure LIFETEST was used to generate survival function estimates by diagnosis cohort, with diabetes duration as the time axis, whereas procedure GPLOT was used to produce survival plots. The survival distributions across diagnosis cohorts were compared with the log-rank test. Analyses for the outcomes of all-cause, cardiovascular, and renal mortality are presented both for all participants (i.e. examined, surveyed-only, and predeceased), to reduce the effect of survival bias, and for the examined cohort. Analyses of outcomes for which a clinical exam was required (i.e. CAD, LEAD, DSP, CAN, ON, and PR) were restricted to examined participants, among whom time-dependent Cox models were also constructed to assess whether changes over time in demographic, self-care, and clinical characteristics accounted for any observed differences in incidence across diagnosis cohorts. The co-occurrence of complications was also evaluated among 163 participants with data available for all examined vascular complications studied at a T1D duration of 35 years and χ²-tests were used to assess the divergence between the observed data and values that would be expected under a null hypothesis of no association.

Finally, means and proportions were used to describe trends over time in clinical continuous and categorical risk factor variables, respectively. Mixed models were used to assess significant trends by diabetes duration. To assess whether risk factor trajectories differed across diagnosis cohorts, we tested the presence of effect modification by including an interaction term between diabetes duration and diagnosis cohort in separate, for each risk factor, mixed models.

Results

Descriptive characteristics at study initiation (1986–1988) of the EDC participants included in this investigation overall as well as by diagnosis cohort are presented in Table 1. Of 440 individuals diagnosed within 1965–1980, the majority (82.5%) participated in biennial clinical examinations whereas 14.1% participated only through surveys and 3.4% died before study entry. As expected, a greater proportion of predeceased individuals were diagnosed in 1965–1969, whereas the majority of survey-only and examined participants were diagnosed with T1D after 1970. The mean age at study entry was 24 years and declined across diagnosis cohorts, being 27.9 years in the 1965–1969, 23.9 years in the 1970–1974 and 19.1 years in the 1975–1980 diagnosis cohort (P<0.0001). No differences across diagnosis cohorts were observed in the age at diabetes onset (8.6 years), whereas a slightly lower prevalence of women was observed in those diagnosed between 1970 and 1974. The majority (96.1%) of participants were Caucasian, with no differences noted by diagnosis cohort. Among examined participants, no differences were observed in HbA1c and HDL cholesterol concentrations, though individuals diagnosed after 1975 had a lower BMI and non-HDL cholesterol concentration and were less likely to have hypertension at study entry.

Table 1: EDC participant characteristics overall and by type 1 diabetes diagnosis cohort at study entry (1986–1988)

Survival function estimates by diagnosis cohort for all-cause, cardiovascular and renal mortality (whose assessment does not require a clinical examination and are thus based on the entire EDC cohort) are shown in Fig. 1. All-cause mortality was significantly lower among those diagnosed after 1970 compared with individuals with a T1D diagnosis in 1965–1969 (Fig. 1a, P=0.009). Cardiovascular mortality improved from the earliest to the most recent diagnosis cohort (Fig. 1b, P=0.04), whereas renal mortality was highest in the 1965–1969 and lowest in the 1970–1974 cohort (Fig. 1c, P=0.004). Restricting the analyses to examined participants, however, survival estimates were similar across diagnosis cohorts both for all-cause (P=0.76), as well as for cardiovascular (P=0.56) and renal mortality (P=0.28) (not shown).

Fig. 1: Complication survival across diagnosis cohorts among Epidemiology of Diabetes Complications (EDC) study participants, including examined, survey-only and predeceased. (a) Survival curves for all-cause mortality by diagnosis cohort in the total EDC study population (log-rank P=0.009). (b) Survival curves for cardiovascular mortality by diagnosis cohort in the total EDC study population (log-rank P=0.04). (c) Survival curves for end-stage renal disease mortality by diagnosis cohort in the total EDC study population (log-rank P=0.0004).

Figure 2 displays survival function estimates by diagnosis cohort for vascular complications requiring a clinical examination. As seen, no clear differences by diabetes duration were observed for complication-free survival across diagnosis cohorts for CAD (Fig. 2a, P=0.56), DSP (Fig. 2c, P=0.10), CAN (Fig. 2d, P=0.62), and ON (Fig. 2e, P=0.57), although a marginal improvement from the 1965–1969 to the 1975–1980 diagnosis cohort was observed for PR (Fig. 2f, P=0.06). LEAD-free survival was highest in the 1970–1974 and lowest in the 1975–1980 cohort (Fig. 2b, P=0.009).

Fig. 2: Complication-free survival across diagnosis cohorts among examined Epidemiology of Diabetes Complications (EDC) study participants. (a) Survival curves for coronary artery disease by diagnosis cohort among examined EDC study participants (log-rank P=0.56). (b) Survival curves for lower extremity arterial disease by diagnosis cohort among examined EDC study participants (log-rank P=0.009). (c) Survival curves for distal symmetric polyneuropathy by diagnosis cohort among examined EDC study participants (log-rank P=0.10). (d) Survival curves for cardiac autonomic neuropathy by diagnosis cohort among examined EDC study participants (log-rank P=0.62). (e) Survival curves for overt nephropathy by diagnosis cohort among examined EDC study participants (log-rank P=0.57). (f) Survival curves for proliferative retinopathy by diagnosis cohort among examined EDC study participants (log-rank P=0.06).

Fig. 2: (Continued)

Among examined participants with data available for all complications of interest at 35 years of diabetes duration, only 15.3% were complication free. More than 18.0% were diagnosed as having a single vascular complication, while the majority (66.3%) was diagnosed with two or more complications. In this subgroup, the observed number with any two complications co-occurring was greater than would be expected under a null hypothesis of no association (Supplemental Table 1, Supplemental digital content 1, https://links.lww.com/CAEN/A18). Particularly striking is the high χ²-statistic for the co-occurrence of ON and PR.

In time-dependent Cox proportional hazards models, allowing for baseline duration, sex, and repeated measurements over time of BMI, current smoking status, HbA1c, systolic blood pressure, hypertension medication use, and non-HDL cholesterol, evidence for an improvement in complication-free survival for the 1975–1980 diagnosis cohort was observed only for DSP and PR, although findings did not reach statistical significance (Table 2). Using year of T1D diagnosis as the main independent variable in these analyses produced similar results, with the exception of a marginal improvement in the incidence of ON with a more recent diagnosis (hazard ratio=0.72, 95% confidence interval: 0.51–1.02).

Table 2: Time-dependent Cox proportional hazards models (with repeated covariate measurements over time) for the incidence of type 1 diabetes complications by 35 years duration: examined EDC study cohort

Given the suggestion of a marginal improvement in complication-free survival after adjustment for risk factor trajectories, we further evaluated whether risk factor levels differed post study baseline. Among all examined participants, a significant increase in the adoption of intensive therapy (from 5.1% in 1986–1988 to 63.0% in 2012–14, P_trend<0.0001) was observed, which was accompanied by a decline in HbA1c (8.9% in 1986–1988 vs. 8.2% in 2012–14, P_trend<0.0001). These trends were similar across diagnosis cohorts for HbA1c (P_interaction=0.93), but not for intensive insulin therapy, where a significant time by diagnosis cohort interaction was observed (P_interaction=0.04): intensive therapy was higher in the 1975–1980 diagnosis cohort until the late 1990s, and became similar across cohorts by 2004. Nevertheless, as at each given time point, the more recent diagnosis cohorts would have a shorter diabetes duration compared with their more distantly diagnosed counterparts, assessing differences in risk factor trajectories by follow-time fails to provide a clear picture of what differences may or may not exist at similar diabetes durations. We, thus, further assessed trends in intensive insulin therapy and HbA1c by diabetes duration and diagnosis cohort. Analyses were restricted to those individuals with a diabetes duration between 17 and 37 years, to assure participants across all three cohorts were studied at comparable durations. The overall proportion of adopting intensive therapy increased significantly (P<0.0001), whereas HbA1c declined across diabetes durations postbaseline (Fig. 3a); these trends were, however, similar across diagnosis cohorts (intensive therapy P_interaction=0.36 and HbA1c P_interaction=0.11).

Fig. 3: Risk factor trends by diabetes duration among individuals in the examined Epidemiology of Diabetes Complications (EDC) cohort diagnosed with type 1 diabetes in 1965–1980. (a) Trends in HbA1c by diabetes duration and diagnosis cohort among examined EDC study participants. P _{interaction (time×diagnosis cohort)}=0.11; P _trend<0.0001. (b) Trends over time in BMI by diagnosis cohort among examined EDC study participants. P _{interaction (time×diagnosis cohort)}=0.55; P _trend<0.0001. (c) Trends in non-HDL cholesterol by diabetes duration and diagnosis cohort among examined EDC study participants. P _{interaction (time×diagnosis cohort)}=0.12; P _trend=0.0008. HDL, high-density lipoprotein.

During the same period, levels of BMI increased (P<0.0001), current smoking declined (P<0.0001), the proportion diagnosed with hypertension increased significantly from 9.2% (across all cohorts) in 1986–1988 to 38.5% in 2012–2014 (P_trend<0.0001), and the concentration of non-HDL cholesterol fell (across all cohorts) (P<0.0001). BMI increased by the duration of diabetes postbaseline (BMI P<0.0001; Fig. 3b), as did hypertension (P<0.0001), whereas the proportion of current smokers (P<0.0001) and non-HDL cholesterol concentration declined (P=0.0008; Fig. 3c). Once again, no significant ‘diabetes duration by diagnosis cohort interactions’ were observed (BMI P_interaction=0.55, current smoking P_interaction=0.12, hypertension P_interaction=0.15, and non-HDL cholesterol P_interaction=0.12).

Discussion

In a cohort representative of the T1D population in Allegheny County, Pennsylvania, USA, univariately, all-cause, cardiovascular and renal mortality over 35 years duration significantly declined in those with more recent onset. However, restricting analyses to examined participants, no differences in mortality were observed with a more recent diagnosis. Generally, no significant univariate differences in complication-free survival were observed across diagnosis cohorts, with the exception of LEAD incidence, which was highest in the 1975–1980 yet lowest in the 1970–1974 diagnosis cohort, and a marginal improvement in PR in the more recent diagnosis cohorts. Similar results were obtained when risk factor trajectories were accounted for, with nonsignificant improvements observed in the 1975–1980 diagnosis cohort for DSP, PR, and macroalbuminuria.

The clinical course of cumulative complication incidence by 30 years’ duration was previously compared among the intensive arm of the DCCT/EDIC, its conventional therapy arm, and a subset of the EDC cohort selected to match DCCT entry criteria ¹⁵. Lower cumulative incidences for cardiovascular disease, nephropathy, and PR were observed in the intensive therapy group, with similar complication rates observed between the conventional therapy group of DCCT/EDIC and the EDC subcohort. In addition to their support of intensive therapy as a means to curb complication development, these findings suggest that rates observed within the EDC cohort are not unusual, but rather represent those of the general T1D population in the USA.

The fact that only marginal improvements in complication-free survival with a more recent diabetes onset arose for most of the vascular complications studied after adjusting for risk factor trajectories, suggests that important risk factors may not have improved as expected over time among those diagnosed with the disease more recently. Indeed, no deviations in risk factor levels were observed across cohorts at the same diabetes duration. Thus, diabetes management improved during the study follow-up, with dramatic increases in the adoption of intensive insulin therapy after 1996–1998 and sharp declines in the concentration of HbA1c, a trend similar across diagnosis cohorts. Trends for nondiabetes specific risk factors were also similar by diabetes duration among diagnosis cohorts, and although the proportion of current smokers fell, as did the concentrations of non-HDL cholesterol, both BMI and the proportion with hypertension increased. These findings, therefore, challenge the notion that a more recent T1D diagnosis necessarily denotes improved risk factor management, leading to improved health outcomes. These results may also help explain the lack of a difference across diagnosis cohorts in the incidence of the majority of complications assessed in unadjusted models, despite a more recent onset coinciding with younger chronological age.

There exist few prospective studies, which could offer comparable data, having a similar length of follow-up on a well-phenotyped T1D cohort, and objective assessments for an extensive list of complications. The Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR) has previously shown that the prevalence of PR by 35 years duration declined among those diagnosed in 1975–1980 compared with individuals with a diabetes diagnosis occurring between 1922 and 1974 ¹⁶. These findings, both in terms of the cumulative incidence rate and the observed trends, are similar to our data of a declining incidence of PR, which was also more pronounced in the most recent, 1975–1980, diagnosis cohort. More recently, a comparison of retinopathy severity between the WESDR cohort and another cohort from the same geographic area but diagnosed 8–34 years later showed threefold higher odds of more severe retinopathy in WESDR by 20 years of duration, a relationship which was somewhat attenuated, though still significant, after adjusting for HbA1c ¹⁷. A Danish study of individuals diagnosed with T1D between 1965 and 1984 also suggested a declining cumulative incidence of PR with a more recent diabetes onset by 20 years of duration ¹⁸; once again, these cumulative incidence rates were similar to those we observed in the EDC study by 20 years. Declines in retinopathy incidence were further reported in adolescents with T1D ^19,20 and, as the previously mentioned studies in adults, these studies in youth also suggested that improvements in risk factors, including glycemic control, may be responsible for the reductions in retinopathy in more recently diagnosed cohorts. In the EDC study, taking into account repeated covariate measurements over time reduced the marginally lower risk associated with the 1975–1980 and eliminated the lower risk associated with the 1970–1974 diagnosis cohort.

The incidence of DSP, a highly prevalent and severe complication of diabetes, nonsignificantly declined in the most recent diagnosis cohort, being 63% lower in the 1975–1980 compared with the 1965–1969 diagnosis cohort. This nonsignificant decline in the development of DSP with a more recent diabetes onset emerged only after taking into account risk factor trajectories, suggesting that both strict diabetes control, as well as management of other vascular disease risk factors, such as lipids, blood pressure, body fatness, and smoking are important for the progression of this complication, as previously shown ²¹. Currently, no other studies have published data on trends in DSP incidence in adults with T1D. Nevertheless, among T1D youth, peripheral nerve abnormalities were shown to increase over time, regardless of whether intensification of diabetes management resulted in reductions ²², or not ²³, in the concentration of HbA1c. Notably, rates reported among youth at a median duration of 7.5 years ²³ seem comparable to rates observed among EDC participants diagnosed after 1970 at a duration of ~18 years. Taken together, these data suggest that peripheral nerve abnormalities are highly prevalent even among young patients with T1D and that optimal control of nonglycemic risk factors may be essential in these complications in this population. Interestingly, similar to our findings of no improvement, among T1D youth, cardiac autonomic abnormalities appeared to remain unchanged over time ²³, while their prevalence was reported to be comparable to that of adults with diabetes ²⁴, suggesting that autonomic neuropathy develops early in the natural history of T1D and remains prevalent throughout a patient’s life.

Unlike other diabetes complications, where existing data may be limited, numerous studies have focused on diabetic nephropathy, although trend data are available mostly for more advanced forms of the disease, that is, ESRD. Within the EDC study, we recently showed that the crude cumulative incidence of macroalbuminuria was essentially identical between participants diagnosed in 1950–1964 and those diagnosed in 1965–1980 by 30 and 40 years of diabetes duration, whereas rates of ESRD dramatically declined across diagnosis cohorts ²⁵. Our current analyses focusing on the post-1965 diagnosis EDC cohort provide further support for declining renal disease mortality by 35 years of diabetes duration and a stagnant crude cumulative incidence of macroalbuminuria over time. This picture for macroalbuminuria, only slightly changed when differences in risk factor trajectories across diagnosis cohorts were accounted for, revealing a marginally significant 28% reduction in disease incidence with a more recent diabetes diagnosis year in analyses with onset year used as a continuous variable. Declines in the incidence of increased albuminuria over time were previously reported in youth with T1D ^20,23, although trend analyses in adults are only available for renal replacement therapy, mostly from European countries, where ESRD incidence was reported to be much lower compared with that in the USA ^25–28. Despite this, European investigators observed that ESRD has declined further over time ^27,28.

In multivariable analyses, survival free from LEAD appeared to have improved nonsignificantly only in the 1970–1974 compared with the 1965–1969 diagnosis cohort, while a nonsignificant increase in incidence was observed in the 1975–1980 cohort. This lack of a significant decline in the incidence of LEAD, especially in the most recent onset cohort, likely relates to unfavorable trajectories for nondiabetes related risk factors, as they have been previously shown to contribute to the development of LEAD in this population ²⁹.

The incidence of CAD remained unchanged across diagnosis cohorts, although significant reductions were observed in cardiovascular disease mortality in the overall cohort. Substantial reductions in the incidence of mortality, but less marked declines in hospitalization for CAD, were previously reported from the Swedish National Diabetes Register, during a shorter timeframe (1998–2014) ³⁰, but are consistent with our data suggesting greater improvements with CAD mortality than morbidity. Data from Australia also noted declines in all-cause and cardiovascular disease mortality between 2000 and 2011, with the exception of those 0–40 years ³¹; in the latter group, there were significant increases in all-cause mortality and no change in cardiovascular disease mortality. As risk factor data were not available, however, the increased risk among younger individuals is difficult to explain.

There are many inherent strengths of the present study, including a well-characterized, representative cohort of childhood-onset T1D, objective assessments for an extensive list of complications, and follow-up extending beyond 25 years. However, as with any research study, the present is not free of limitations. In particular, although restricting analyses to a diabetes duration of 35 years assured estimates obtained were comparable across diagnosis cohorts, it reduced the available sample size, likely reducing our power to detect significant results. It is further likely that survival bias would have differentially affected results for individuals diagnosed earlier than would those with a more recent diabetes onset. We, however, sought to address this issue by including data for those who died prior to study baseline for analysis pertaining to mortality outcomes.

Conclusion

It is clear that our efforts to optimize diabetes management have been fruitful, producing dramatic declines in glycemic levels across all diagnosis cohorts. It is also evident, however, that although glycemic control is certainly an important risk factor for vascular diabetes complications and all-cause mortality, it is not sufficient, unaided, to eliminate the excess complication risk associated with T1D. Other risk factors of great consequence for the development and progression of vascular complications, such as BMI, smoking, hypertension, and dyslipidemia continue to afflict the T1D population. Alarmingly, this continues to be the case also for the most recent onset subgroup of the EDC study, although data such as those presented from Australia ³¹ and the similar complication rates in youth as in adults with T1D ²⁴, suggest this observation may not be restricted to the EDC cohort. Thus, the adverse risk factor patterns, present also in the most recent cohort, are counteracting the successes in curbing complication development, which would be expected from overall improved management of dysglycemia and dyslipidemia. It is therefore of utmost importance that efforts are expanded to address other modifiable risk factors whose intense control should occur early in the natural history of this disorder in order to assure not only an increased life expectancy but also a better quality of life for patients with T1D.