Prolonged Effect of Intensive Therapy on the Risk of Retinopathy Complications in Patients With Type 1 Diabetes Mellitus:  10 Years After the Diabetes Control and Complications Trial FREE

The Diabetes Control and Complications Trial (DCCT) was designed to determine whether intensive therapy with the aim of maintaining glycemic levels as close to the nondiabetic range as possible would prevent or delay the long-term complications of type 1 diabetes mellitus.¹The DCCT demonstrated substantial reductions in the risk of development and progression of the early microvascular complications of diabetes over an average of 6.5 years of intensive therapy as compared with conventional therapy. At the close of the DCCT in 1993, patients in the conventional therapy group were offered intensive therapy and instructed in its use. All patients subsequently returned to their health care providers for further diabetes care and 97% of the original DCCT cohort (n = 1394) was enrolled in the Epidemiology of Diabetes Interventions and Complications (EDIC), a long-term observational study.²An earlier report showed that the ongoing risk of all levels of retinopathy remained significantly reduced in the intensive compared with the conventional group during the first 4 years of EDIC, despite similar glycated hemoglobin (HbA_1c) levels over this period (called “metabolic memory”).³Determining the duration of metabolic memory is important to quantify the long-term clinical effects of intensive diabetes therapy. The current report describes the continuing differences between the 2 original treatment groups in retinal complications 10 years after the close of the DCCT.

At baseline, the 1441 patients enrolled in the DCCT during 1983-1989 were 13 to 39 years of age, had type 1 diabetes mellitus for 1 to 15 years, and were in generally good health. The primary prevention (726 patients with no retinopathy, albumin excretion rates <28 μg/min [<40 mg/24 hours], and 1-5 years' diabetes duration) and secondary intervention (715 patients with diabetes duration of 1-15 years, minimal to moderate nonproliferative diabetic retinopathy [NPDR], and urinary albumin excretion rates <139 μg/min [≤200 mg/24 hours]) cohort participants were randomly assigned to either intensive therapy, with the goal of achieving glycemic levels as close to the nondiabetic range as safely possible, or to conventional therapy, as previously described.¹Intensive therapy included at least 3 injections of insulin daily or continuous subcutaneous insulin infusion with pumps, with insulin dose adjustments based on frequent self-monitoring of capillary glucose levels, meal size and composition, and physical activity levels. Mean duration of follow-up in the DCCT was 6.5 years.

During EDIC, retinopathy was assessed by 7-field stereo fundus photography in approximately one-quarter of the cohort each year and in the entire cohort at years 4 and 10. Photography was not conducted if a patient had previously undergone panretinal photocoagulation in both eyes. Retinopathy status was determined in 1211 patients at EDIC year 10, 1045 based on fundus photography and 166 living patients with a known history of panretinal photocoagulation in either eye during DCCT (35 patients) or EDIC (131 patients). All photographs were graded centrally, with graders masked to therapy assignment, using the final Early Treatment Diabetic Retinopathy Study (ETDRS) grading scale⁴and DCCT methods.⁵The primary outcome was the time to the first occurrence of further retinopathy progression during EDIC, defined as a 3-step or more progression from the level of retinopathy at DCCT closeout,³representing a reproducible measure of clinically important worsening. The secondary retinopathy outcome was the time to the first occurrence of proliferative diabetic retinopathy (PDR) or worse during EDIC. Other retinopathy outcomes were the prevalence of a 3-step or more progression from DCCT entry, severe NPDR (ETDRS level 53/<53) or worse, clinically significant macular edema (CSME),⁶and photocoagulation therapy (focal or scatter). Patients who received panretinal scatter photocoagulation (laser) therapy in either eye were counted as having worsened retinopathy for all of these outcomes thereafter, and patients who received focal photocoagulation for macular edema were counted as having CSME thereafter. Visual acuity was assessed by ETDRS methods.⁷

Interreader reliability during EDIC was evaluated by having different graders reread the same 50 fundus photographs at each EDIC year and comparing the results with the primary double reading at DCCT closeout. The individual weighted κ measure⁸of interrater agreement beyond chance ranged from 0.82 to 0.92 for ordinal ETDRS scores and from 0.71 to 0.90 for ordinal CSME scores over 10 years of measurements. The overall weighted κ⁹stratified for EDIC year was 0.91 for ETDRS scores and 0.84 for CSME scores.

Hemoglobin A_1cwas measured annually in a central laboratory by high-performance liquid chromatography.¹⁰The mean HbA_1cvalue was calculated as the time-weighted average during the DCCT and EDIC.¹¹

To test for differences between groups, the Wilcoxon rank sum test was used for quantitative or ordinal observations¹²and the χ²test, for categorical data.¹³Generalized estimating equations with an unstructured working correlation matrix¹⁴were used to assess the aggregate HbA_1clevel difference between groups over EDIC years and to test for differences in odds reduction in further 3-step or more progression and PDR between EDIC years 4 and 10.

Analyses of progression of retinopathy were stratified by, or included adjustment for, retinopathy severity at the end of DCCT, defined as no retinopathy (ETDRS grade 10/10), microaneurysms only (ETDRS grade 20), mild NPDR (ETDRS grade 30), moderate NPDR or greater (≥ETDRS grade 40), and any previous laser therapy (focal or scatter). The Mantel-Haenszel method provided a stratified adjusted odds ratio,¹⁵with test-based confidence intervals (CIs). Logistic regression models assessed the effects of covariates on the prevalence (odds) of a particular retinopathy outcome at a specific EDIC year.¹⁵Pvalues were obtained from likelihood ratio tests. The percentage of reduction in the odds with intensive vs conventional therapy was computed as (1 − odds ratio) × 100.

The Weibull proportional hazards regression model for interval-censored data¹⁶evaluated the treatment group effects on the cumulative incidence of further retinopathy progression during EDIC adjusted for other covariates. The model used all photographs in all patients. The Weibull assumption was verified by empirical estimation of the survival function.¹⁷Risk (hazard) reduction with intensive vs conventional therapy was calculated as (1 − hazard ratio) × 100. Pvalues were obtained from likelihood ratio tests. The proportion reduction in −log likelihood (R²_") was used to describe the proportion of variation in risk explained by the HbA_1clevels.¹⁵All analyses were performed using SAS (SAS Institute Inc, Cary, North Carolina).

Table 1shows the characteristics at DCCT baseline and at the end of the DCCT (EDIC baseline) of the 1211 subjects with retinopathy status determined at EDIC year 10. At DCCT baseline, there were no significant differences between the intensive and conventional treatment groups. However, treatment group differences reflecting the beneficial effects of intensive therapy were seen at DCCT end for HbA_1clevel, prevalence of hyperlipidemia, retinopathy level, and need for photocoagulation.

During 6.5 years of treatment in DCCT, intensive and conventional therapy groups adhered to their assigned therapies 98% and 97% of the time, respectively. At EDIC year 4, 95% of the former intensive therapy group were still being treated with multiple daily injections of insulin or an infusion pump, compared with 75% of the former conventional therapy group (P < .001). At EDIC year 10, the differences between the 2 groups had narrowed further with regard to insulin therapy and self-monitoring (Table 1).

At entry to the DCCT, the mean HbA_1clevel in each treatment group was 9% (to convert to proportion of total hemoglobin, multiply by 0.01) (Table 1). Following 6.5 years of DCCT follow-up, the mean HbA_1clevels were 7.3% and 9.0% in the intensive and conventional therapy groups, respectively. At the first EDIC evaluation, 1 year after DCCT end, HbA_1cvalues in the 2 groups had converged (Figure 1). Over 10 years in EDIC, the mean HbA_1clevels in the 2 former treatment groups were almost the same (8.07% in the conventional therapy group vs 7.98% in the intensive therapy group; P = .20).

The prevalences of various levels of retinopathy and CSME were lower in the former intensive therapy group than in the former conventional therapy group at the end of DCCT and also at years 4 and 10 of EDIC (Table 2). The likelihood (odds) of a 3-step or more progression in retinopathy from DCCT baseline, the principal DCCT outcome, was 76% lower in the intensive than in the conventional therapy group at the end of DCCT (10.7% vs 33.2%). After 4 years of follow-up in EDIC, 48.9% of the former conventional therapy group had a 3-step or more progression in retinopathy from DCCT baseline compared with 17.8% of the former intensive therapy group;after 10 years, 60.6% had progressed in the conventional group vs 35.8% in the intensive group.

The overall prevalences during EDIC reflect, in part, retinopathy differences associated with intensive vs conventional therapy during DCCT. To eliminate the carryover of the treatment group differences at the end of the DCCT into EDIC, we performed logistic regression analysis adjusted for the level of retinopathy at the end of the DCCT. The adjusted odds of retinopathy progression from DCCT entry were reduced by 74% with intensive vs conventional therapy at 4 years of EDIC and 57% at 10 years (each P < .001). Continued significant reductions at 4 and 10 years of EDIC follow-up were also observed in the adjusted odds of severe NPDR or worse, PDR or worse, CSME, and photocoagulation. However, the odds reductions at 10 years were less than that observed at 4 years, except for those for photocoagulation (Table 2).

To assess metabolic memory further, we examined the prevalence of further 3-step or more progression of retinopathy from the level of retinopathy at DCCT closeout, adjusted for the level at closeout, among those free of panretinal scatter laser therapy during the DCCT. There was a 71% (95% CI, 56%-81%) odds reduction (P < .001) with intensive vs conventional therapy at EDIC year 4 (6.6% and 21.8% prevalence, respectively) and 50% (95% CI, 35%-62%) reduction (P < .001) at EDIC year 10 (24.2% and 40.8% prevalence for intensive and conventional treatment groups, respectively). Generalized estimating equations analysis showed that the beneficial treatment effect in further 3-step or more progression waned (P = .003). We also examined the prevalence of PDR or worse among those free of PDR during the DCCT. The odds reduction with intensive therapy after adjustment for the retinopathy levels at DCCT closeout was 76% (95% CI, 45%-89%) at EDIC year 4 (P < .001), with prevalences of 1.5% and 8.9% in the intensive and conventional treatment groups, respectively. At year 10, the odds reduction of PDR or worse was 59% (95% CI, 37%-73%; P < .001), with prevalences of 6.5% and 19.2% in the intensive and conventional treatment groups, respectively. However, for PDR or worse, the generalized estimating equations analysis showed that the odds reduction was not significantly different between years 4 and 10 (P = .12).

An additional analysis examined the cumulative incidence of further 3-step or more progression during EDIC from the level at DCCT closeout in multivariate Weibull proportional hazards regression models using evaluations at all years in subjects, after excluding 36 participants who had scatter photocoagulation in either eye during DCCT (Table 3) (n = 1349). The Weibull model revealed a highly significant beneficial effect of DCCT intensive therapy up to 10 years after the end of DCCT. Figure 2A presents the estimated cumulative incidence of retinopathy further progression in each group derived from the Weibull model, reaching 51% at 10 years in the former conventional and 29% in the intensive treatment groups. The risk (hazard) of further retinopathy progression over the 10 years of EDIC was reduced by 53% (P < .001; 95% CI, 43%-61%). However, this beneficial effect was attenuated compared with the results over the first 4 years after the end of DCCT³(Figure 2B), when there was a 70% risk (hazard) reduction (95% CI, 59%-79%; P < .001) with intensive therapy. The Weibull model was further fit for the interval between EDIC year 4 and EDIC year 10 (Figure 2B) among those patients who were free of further 3-step progression from DCCT closeout as of EDIC year 4 (n = 1105). The risk reduction between EDIC years 4 and 10 verified the persistent and highly significant, albeit attenuated, beneficial effect of the former intensive therapy over this 6-year period (38% risk reduction; 95% CI, 22%-51%; P < .001).

A Weibull model analysis of the cumulative incidence of PDR or worse among patients who were free of PDR or worse during DCCT had similar results (Figure 3). The risk (hazard) of PDR during the 10 years of EDIC follow-up was reduced by 56% (95% CI, 37%-70%; P < .001), 71% during the first 4 years of EDIC (95% CI, 44%-85%; P < .001), and 46% from EDIC year 4 to 10 (95% CI, 16%-65%; P < .001).

Table 3presents separate Weibull models of the incidence of further 3-step progression and PDR over the 10 years of EDIC follow-up within the strata defined by retinopathy levels at DCCT closeout (EDIC baseline). For all DCCT closeout retinopathy levels, there was an overall benefit over the 10 years, but as the severity of retinopathy increased at DCCT closeout, the relative benefits of DCCT intensive therapy decreased. Whereas Table 3examines the prolonged protective effect of intensive vs conventional treatment at each retinopathy level, Table 4examines whether the inclusion of other risk factors attenuates the effects of intervention group. After adjusting for other DCCT baseline and DCCT closeout covariates, the differences between DCCT treatment groups in the risk of further progression of retinopathy remained highly significant (Table 4) (P < .001). Risk of further progression of retinopathy increased significantly with higher HbA_1clevel at DCCT baseline (19% increase in risk per 1% increase in HbA_1clevel; P < .001), higher mean blood pressure at DCCT closeout (11% increase in risk per 5 mm Hg–increase in the mean blood pressure; P < .001), and hyperlipidemia at DCCT closeout (70% increase in risk for those with hyperlipidemia vs those without; P = .001).

Another Weibull model assessed the effect of the combined DCCT and EDIC mean HbA_1clevels on the risk of further progression of retinopathy among those free of scatter photocoagulation during DCCT. Within each former therapy group, the hazard of further 3-step or more progression of retinopathy during EDIC increased as the mean HbA_1cvalues during the DCCT and EDIC increased, adjusting for cohort, diabetes duration, HbA_1clevel at DCCT entry, and the level of retinopathy at the end of the DCCT. In the former conventional and intensive therapy groups, there was a 1.9 times greater risk and 2.0 times greater risk, respectively, of further progression of retinopathy for every 10% increase in HbA_1clevel (eg, from 8.0% to 8.8%) during the DCCT and EDIC (95% CI, 1.8-2.2; P < .001 and 95% CI, 1.8-2.3; P < .001, respectively). The HbA_1clevel effects on further progression of retinopathy were not significantly different for the 2 former DCCT treatment groups (P = .40).

In additional models that combined both treatment groups and adjusted for mean HbA_1clevels during DCCT or for mean HbA_1clevels during EDIC separately, 89% of the prolonged effect (R²) of DCCT intensive therapy on further retinopathy progression was explained by the differences in the DCCT mean HbA_1clevels, whereas the EDIC mean HbA_1clevels explained only 1.6% of the prolonged intensive therapy effect.

After 10 years of EDIC follow-up, 4 former intensive therapy patients had a visual acuity worse than 20/200 in 1 eye. None was so affected in both eyes. Only 1 of these 4 patients lost vision owing to diabetic retinopathy. One former conventional therapy group patient had a visual acuity worse than 20/200 in 1 eye at EDIC year 10 owing to PDR.

During the first 10 years of follow-up in the EDIC, the level of glycemic control in the former DCCT therapy groups converged. Based on previous epidemiologic assessments, the small difference in HbA_1cvalues between the former therapy groups would be expected to reduce the relative benefit of intensive therapy that occurred during the DCCT.¹⁸However, progression of retinopathy during the first 4 years of post-DCCT follow-up remained markedly less frequent in the former intensive therapy group, despite an increase in median HbA_1cvalue from 7.2% during the DCCT to 7.9% during the EDIC, than in the former conventional therapy group. Conversely, in the former conventional therapy group, the risk of progression of retinopathy during the first 4 years of EDIC³remained about the same as during the first 4 years of the DCCT,¹⁹despite a decrease in the median HbA_1cvalue from 9.1% during DCCT to 8.2% during EDIC. The continued separation in retinopathy of the former treatment groups was not merely a reflection of the differences in the severity of retinopathy between the 2 groups at the end of the DCCT, since the reductions in risk of further progression persisted after adjusting for the differences in complications between the 2 therapy groups at DCCT end. DCCT/EDIC has shown a similar prolonged effect of prior intensive therapy on microalbuminuria and albuminuria¹¹and neuropathy.²⁰

The likelihood of further progression of retinopathy in both groups was strongly associated with the mean HbA_1cvalue during the DCCT and EDIC combined, with a stronger effect of the mean HbA_1cvalue during the DCCT. In the Stockholm Diabetes Intervention Study, the prevalence of severe retinopathy after 7.5 years of follow-up was related to the mean HbA_1cvalue during the first 5 years of follow-up.²¹

Intensive therapy that maintains near-normal glycemic levels for an average of 6.5 years has a beneficial impact on long-term complications that extends at least 10 years beyond the actual period of such therapy. Moreover, therapy that maintains higher HbA_1clevels has adverse effects on complications that persist beyond the period of high HbA_1clevels. However, the DCCT/EDIC results should not be interpreted to mean that intensive therapy need only be applied for a limited period. Rather, the results support the implementation of intensive treatment as early in the course of the disease as possible.Stratified Weibull models fitted separately in the 4 retinopathy strata at DCCT closeout (Table 3) reveal that the metabolic memory is waning faster in patients with more severe retinopathy than in those with milder retinopathy, which reinforces the importance of implementing intensive glycemia control as early in the course of the disease as possible.

One potential limitation of the current study is that we did not adjust for other medication use, which might have confounded the results. However, we have shown in previously published analyses that the use of other medications, such as angiotensin-converting enzyme inhibitors and aspirin, was not significantly different between the treatment groups.²²

The persistent adverse effects of hyperglycemia and the long-term beneficial effects of lowering glycemia on the development and progression of complications, also shown in animal models of diabetes,²³has been termed metabolic memory. One possible explanation for this phenomenon is the slow accumulation, and subsequent slow degradation, of advanced glycation end products (AGEs).²⁴DCCT patients in the intensive therapy group had lower concentrations of these substances in skin collagen than did patients in the conventional therapy group.²⁵The levels of skin collagen AGEs were also associated with the subsequent incidence of progression of retinopathy (and nephropathy) over the first 10 years of EDIC.²⁶Although the metabolic memory effect is present 10 years after the DCCT, the apparent waning of metabolic memory (“metabolic amnesia”) between EDIC years 4 and 10 may be secondary to a combination of clearance of the long-lasting AGEs in the former conventional group and the accumulation of AGEs in the former intensive treatment group. There are currently no direct data to prove this speculation, and alternative explanations include epigenetic effects of hyperglycemia or a combination of effects.

Correspondence:David M. Nathan, MD, Diabetes Unit, Massachusetts General Hospital, Bulfinch 408, 55 Fruit St, Boston, MA 02114 (dnathan@partners.org).

Submitted for Publication:February 4, 2008; final revision received April 23, 2008; accepted May 31, 2008.

Financial Disclosure:None reported.

Funding/Support:DCCT/EDIC is sponsored through research contracts from the Division of Diabetes, Endocrinology, and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institutes of Health. Additional support is provided by the General Clinical Research Centers Program, National Center for Research Resources, National Institutes of Health, and by Genentech, Inc, through a Cooperative Research and Development Agreement with the National Institute of Diabetes and Digestive and Kidney Diseases.