Diet and Diabetic Retinopathy: Insights From the Diabetes Control and Complications Trial (DCCT)

Abstract and Introduction

Abstract

Objective

We explore the influence of lifestyle on the progression of retinopathy.

Design

Post hoc statistical analysis.

Subjects/Setting

One thousand forty-one patients with type 1 diabetes from 29 specialty clinics.

Intervention

The Diabetes Control and Complications Trial (DCCT) lifestyle data (diet, exercise, and tobacco use) and retinopathy-related risk factors (mean arterial pressure, the low-density lipoprotein/high-density lipoprotein cholesterol ratio [LDL-C/HDL-C], serum triglycerides, glycosolated hemoglobin [HbA1c] levels, body mass index [BMI], and insulin utilization) were related to the rate of progression of retinopathy.

Main Outcome Measures

Correlation between lifestyle data with progression of retinopathy and retinopathy-related risk factors.

Results

The percentage of calories as total fatty acids at baseline and overall positively correlated with prestudy and overall progression of retinopathy (r = .15, P < .0001 and r = .14, P < .0001, respectively). Average overall percentage of calories as dietary fiber inversely correlated with prestudy and overall progression of retinopathy (r = −.07, P = .0102 and r = −.10, P < .0002, respectively). The progression of retinopathy correlated with mean arterial pressure (prestudy r = .09, P = .0004 and overall r = .20, P < .0001), LDL-C/HDL-C (prestudy r = .13, P < .0001 and overall r = .15, P < .0001), serum triglycerides (prestudy r = .18, P < .0001 and overall r = .26, P < .0001), HbA1c (prestudy r = .10, P < .0001 and overall r = .45, P < .0001), BMI (prestudy r = .08, P < .0034 and overall r = .05, P = .08), insulin utilization (prestudy r = .19, P < .0001 and overall r = .14, P < .0001), tobacco use (prestudy r = .08, P < .0231 and overall r = .09, P < .0011), and the intensive vs conventional therapy study group (on-study r = −.27, P < .0001).

Conclusion

Tobacco use and diet, particularly the consumption of fatty acids and dietary fiber, are significantly associated with the rate of progression of diabetic retinopathy and retinopathy-related risk factors.

Introduction

Previous to 1993, studies had not proven that near-normal glycemia affects the development of retinopathy or other microvascular or macrovascular complications of diabetes.[1–10] In 1993, the study authors of the Diabetes Control and Complications Trial (DCCT) presented the results and conclusions of their 9-year study, supporting intensive glycemic control to slow the progression of retinopathy, nephropathy, and neuropathy. With a randomized, controlled trial of 1441 patients with type 1 diabetes, an intensive treatment program with close monitoring by the patients and medical personnel was associated with reduced progression of retinopathy (65%), nephropathy (54%), and neuropathy (60%) compared with conventional treatment.[11] The intensive treatment group patients received 3–4 insulin injections per day or a continuous infusion insulin pump, whereas conventionally treated patients had 1 or 2 insulin injections per day. The intensive control group averaged 1.9% lower glycosolated hemoglobin (HbA1c) levels than the conventional treatment group throughout the 6.5-year average follow-up period (HbA1c = 7.2% vs 9.1%, respectively, P < .0001).[11]

These results of the DCCT regarding microvascular complications of patients with type 1 diabetes established the role of near-normal glycemia despite the increased burden on the patient[12] and the additional hypoglycemic risks.[13,14] Since the completion of the DCCT, the risk of hypoglycemia has been significantly reduced with the advent of short-acting insulins.[15]

The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure does not refer to blood pressure control reducing the microvascular complications of diabetes.[16] One observational study found a positive correlation between hypertension and retinopathy.[17] The most recent report of the National Cholesterol Education Program does not mention evidence that lipid-lowering strategies reduce the microvascular complications of diabetes.[18] The American Diabetes Association's (ADA's) “Evidence-Based Nutrition Principles and Recommendations for the Treatment and Prevention of Diabetes and Related Complications” does not specifically address the effects of nutrition on the microvascular complications of diabetes or of nutrition influencing HbA1c. This report observes that low-saturated-fat (ie, < 10% of energy), high-carbohydrate diets increase postprandial levels of plasma glucose but do not state what they do to HbA1c. It reports that high monounsaturated fat diets have not been shown to improve fasting plasma glucose or HbA1c values. It also comments about the strong short-term relationship between total carbohydrate intake and insulin consumption but not the long-term relationship of insulin consumption to any of the macronutrient components.[19]

Among the many complications of overweight and obesity mentioned in the “Surgeon General's call to action to prevent and decrease overweight and obesity,” progression of diabetic retinopathy is not mentioned.[20] A position statement by the ADA encourages aerobic exercise in type 1 diabetics but does not reference specific evidence that it slows the progression of diabetic retinopathy.[21] Tobacco has been found to accelerate diabetic retinopathy[17,22,23] and to not be correlated with diabetic microvascular complications.[24] Consequently, the extent of the relationship between diet, exercise, and tobacco use and diabetic retinopathy, either directly or indirectly, via increases in blood pressure, lipids, HbA1c, body mass index (BMI), or insulin utilization has not been determined.

The purpose of this analysis of the DCCT data is to (1) determine the extent to which the intake of macro- and micronutrients of subjects differed by time (ie, baseline vs on-study) and strata (ie, primary prevention and secondary intervention); (2) quantify and compare the influence of retinopathy risk factors (duration of diabetes, blood pressure, serum lipids, HbA1c, treatment group [intensive therapy vs conventional treatment], BMI, insulin utilization, exercise, and tobacco use) on the progression of retinopathy; and (3) calculate the influence of individual dietary variables, exercise, and tobacco use on the progression of retinopathy and on mean arterial pressure, the low-density lipoprotein/high-density lipoprotein cholesterol ratio (LDL-C/HDL-C) ratio, serum triglycerides, HbA1c, BMI, and insulin utilization.

Subjects and Methods

Study Cohort

The DCCT eligibility criteria and screening methods have been reported in detail. At baseline, all subjects had to be between 13 and 39 years of age, C peptide-deficient, and without hypertension, hypercholesterolemia, severe diabetic complications, or other medical conditions.[25–28] Institutional review boards in the 29 diabetes specialty centers approved the trial, so further human subjects research committee approval was not required. The primary prevention arm subjects had type 1 diabetes for at least 1 year but for less than 5 years and showed no evidence of retinopathy on stereoscopic fundus photographs. Secondary intervention stratum patients had background diabetic retinopathy (at least 1 microaneurysm in either eye but less than stage P2, according to the Diabetic Retinopathy Study [ie, very mild-to-moderate nonproliferative retinopathy])[26,29] and type 1 diabetes for 1–15 years.

Investigators recruited a total of 1441 study participants (760 males and 676 females; 96.5% white, 2.0% black, and 1.5% other races) between 1983 and 1989, including 726 subjects in the primary prevention study and 715 in the secondary intervention study.[30] At the end of the trial in 1993, the mean time on-study was 6.5 years (range, 3.8–9.7 years).[11]

Procedures

The DCCT computerized datasets in Statistical Analysis System (SAS) format with documentation purchased from the National Technical Information Service (NTIS) served as the data for this study. These data are in the public domain, and this post hoc analysis did not need to be authorized by the DCCT Study Group.

The procedures used to collect blood and measure the level of HbA1c (each month in the intensive group and every 3 months in the conventional treatment group) and lipids (each year) have been reported in detail.[25,31] DCCT ophthalmologists evaluated the extent of retinopathy by means of 7-field stereoscopic color fundus photographs taken every 6 months for all patients. The progression of retinopathy while on-study (defined as a 3-grade change on the final Early Treatment Diabetic Retinopathy Scale [ETDRS] relative to the baseline level in which the patient did not recover or improve in a subsequent visit) served as the major end point of the DCCT reports on eye disease.[29] Because we hypothesized that lifestyle habits before the trial may influence progression of retinopathy, we used the slope of the progression of retinopathy over time as described below.

Trained dietitians interviewed participants and conducted a modified Burke-type diet history that took approximately 1.5–2 hours to complete. The dietitians collected quantitative and qualitative information representing a usual week of dietary intake over the previous year.[32] Subsequently, the dietitians obtained follow-up diet histories at years 2 and 5 and at the end of the study.[26] By analyzing these diet histories with nutritional component software, the dietitians and statisticians generated a dataset consisting of each subject's consumption of macro- and micronutrients at baseline and while on-study. Of the 1441 patients entered into the DCCT, baseline dietary data are missing from 5 and on-study dietary data are missing from 10. This study reports on the 1436 patients with baseline nutritional data and the 1431 with follow-up diet analyses. This diet analysis instrument had a high reproducibility on repeated administrations of the diet history.[33] Of the 99 nutrient variables that we screened, we included the following:

1. Energy (kilocalories);

2. Protein (percentage of kilocalories);

3. Total carbohydrates (percentage of kilocalories);

4. Starch (percentage of kilocalories);

5. Refined carbohydrates (percentage of kilocalories);

6. Dietary fiber (grams/1000 kcal);

7. Total fatty acids (percentage of kilocalories);

8. Saturated fatty acids (percentage of kilocalories);

9. Monounsaturated fatty acids [MUFA] (percentage of kilocalories);

10. Polyunsaturated fatty acids [PUFA] percentage of kilocalories);

11. Alcohol (percentage of kilocalories);

12. Sodium (milligrams/1000 kcal); and

13. Cholesterol (milligrams/1000 kcal).

We reported “dietary fiber” because it is the sum of insoluble and soluble fiber. Finding correlations of fatty acids according to carbon length with progression of retinopathy, we left for a later exercise. Other micronutrients had no significant correlations with the progression of retinopathy.

The ADA diet (< 35% fat) was recommended for all study subjects in the first year of the trial. After a marked increase in BMI over the first year,[34] the DCCT research team advised all subjects to adhere to the American Heart Association Step 1 Diet (ie, < 30% fat and < 10% saturated fat).[25]

The DCCT investigators evaluated the subjects' exercise levels yearly on a 4-point scale: 1 = sedentary, 2 = mild activity, 3 = moderate activity, and 4 = strenuous activity. In the DCCT database, pretrial tobacco use in pack-years was reported at baseline, and subsequently, each year the study authors reported consumption in pack-years. We related the prestudy rate of progression of retinopathy and retinopathy-related risk factors to the pack-years smoked prestudy and the overall rate of progression of retinopathy and retinopathy-related risk factors to the total lifetime pack-years. To compare the percentage of subjects smoking at baseline with the average percentage smoking while on-study, we used the variable, “currently smoking: yes or no,” at baseline and on-study (average percentage over the course of the study).

Statistical Methods

The influence of diet and other lifestyle factors on the progression of retinopathy would be expected to occur due to the patients' entire exposure to nutrients and other lifestyle habits from the inception of the diabetes. Consequently, we related the baseline nutrient intake with the rate of progression of retinopathy prestudy (rate of progression of retinopathy prestudy: ETDRS – 1/the duration in months of pretrial diabetes) and with risk-factor levels measured at baseline. In the scale used in the DCCT, “ETDRS = 1” means undetectable retinopathy (range of ETDRS scores, 1–18). The overall rate of retinopathy progression is:

$$(\text{ETDRS at completion of the trial}-1)/\text{overall duration of diabetes in months}$$

.

To compare the progression of retinopathy while on-study by group, according to compliance with the ADA Diet, we averaged the on-study nutrient data and correlated it with the rate of progression of retinopathy on-study (slope of retinopathy progression determined by: [ETDRS at completion of the trial – ETDRS score at baseline]/time on-study in months).

To compare dietary intake between subjects, we used the percentage of caloric intake for the macronutrients (eg, percentage of protein = protein [grams] × 4 [kilocalories/gram]/total kilocalories/day), dietary fiber in grams/1000 kilocalories/day, and milligrams/1000 kilocalories/day for sodium and cholesterol.

We used the Spearman correlation function to relate the progression of retinopathy prestudy with the baseline blood pressures, serum lipids, HbA1c values, BMIs, insulin use, exercise levels, and tobacco consumption. Likewise, we used the Spearman correlation function to relate overall progression of retinopathy to the average blood pressures (yearly evaluation), serum lipids (yearly), HbA1c values (intensive group [monthly] or conventional group [every 3 months]), insulin use (every 3 months), and exercise levels (yearly). We related the overall progression of retinopathy to the final BMI and total lifetime tobacco use in pack-years. Two sample t tests were used to compare the means of various groups where appropriate.

We examined the positive or negative associations of the baseline and overall progression of retinopathy with baseline and overall nutrient intake, exercise levels, and tobacco use. Average nutrient intake was also related to the means of the blood pressures, serum lipids, HbA1c scores, and insulin use and to the final BMI.

SAS statistical software (release 8.2, SAS Institute, Cary, North Carolina) was used for the data analysis. As this is hypotheses-generating research, we chose an alpha of P < .05. We present the Bonferroni correction in the tables to guide the reader for interpretation.[35]

Results

Compared with the baseline diet, subjects on-study decreased the intake of food energy and the proportion of calories as refined carbohydrates, saturated fatty acids, MUFA, and cholesterol, whereas the proportion of starch and total carbohydrates increased (Table 1). Physical activity levels, measured on a 4-point scale (1 = sedentary, 2 = mild activity, 3 = moderate activity, and 4 = strenuous activity) showed relatively little exercise. On this scale, 87.2% of subjects averaged </= 2.00, 11.7% of subjects averaged 2.01–3.00, and 1.1% of subjects averaged 3.01–4.00. The percentage of subjects smoking did not change significantly from baseline to on-study (percentage of smokers at baseline: 19%; average percentage of subjects smoking on-study: 21%, P = .53).

Table 1.

Consumption of Nutrients, Exercise Levels, and Tobacco Use at Baseline and On-Study

	Baseline Nutrient Intake (n = 1436)		On-Study Nutrient Intake (n = 1431)
Nutrient	Mean	95% CI	Mean	95% CI	t Test of Means	P Value
Food energy (kilocalories)	2484	±1895	2240	±1073	9.89*	<.0001
Protein (percentage in kilocalories [grams])	18.0 [112]	±5.92 [36.7]	17.9 [100]	±4.56 [25.5]	.20	.40
Total carbohydrates (% kcal [g])	45.3 [281]	±14.4 [89.2]	46.8 [262]	±12.5 [70.0]	6.28*	<.0001
Starch (% kcal [g])	22.5 [139]	±8.83 [54.7]	24.3 [136]	±7.66 [42.9]	11.84*	<.0001
Simple carbohydrates (% kcal [g])	15.9 [98.5]	±10.7 [66.3]	14.5 [81.2]	±11.6 [64.9]	6.80*	<.0001
Total fat (% kcal [g])	37.8 [104]	±13.6 [37.5]	35.7 [98.3]	±11.8 [32.5]	9.05*	<.0001
Saturated FA (% kcal [g])	13.2 [36.4]	±6.25 [17.2]	12.2 [33.6]	±5.17 [14.2]	8.80*	<.0001
MUFA (% kcal [g])	14.3 [39.4]	±6.05 [16.7]	13.4 [33.3]	±5.13 [12.8]	7.93*	<.0001
PUFA (% kcal [g])	7.59 [20.9]	±4.80 [13.2]	7.39 [18.4]	±3.71 [9.23]	2.54	.0110
Dietary fiber (g/1000 kcal [grams])	10.9 [27.0]	±7.70 [19.1]	10.8 [26.8]	±6.33 [15.7]	.83	.40
Alcohol (% kcal [g])	.741 [2.62]	±3.41 [12.1]	.983 [3.48]	±4.19 [14.8]	3.36*	.0008
Cholesterol (mg/1000 kcal)	162	±140	140	±100	9.75*	<.0001
Sodium (mg/1000 kcal)	1849	±756	1895	±611	3.66*	.0003
Exercise, 1-4 scale	1.91	±1.11	1.79	±1.43	5.04*	<.0001
Tobacco use, pack-years	2.86	±13.4	4.06	±17.6	N/A	N/A

CI = confidence interval; FA = fatty acids; MUFA = monounsaturated fatty acids; PUFA = polyunsaturated fatty acids; N/A = not available

^*Bonferroni corrected P <.003

(Table 2) shows that the secondary intervention stratum consumed a significantly higher percentage of kilocalories as fat and cholesterol and a lower percentage as carbohydrates than the primary prevention stratum. Exercise levels and tobacco consumption did not differ significantly by stratum.

Table 2.

Overall Consumption of Nutrients, Exercise Levels, and Tobacco Use by Stratum

	Primary Prevention Stratum (n = 721)		Secondary Intervention Stratum(n = 710)
Overall Average Nutrient Intake	Mean	95% CI	Mean	95% CI	t Test of Means	P Value
Food energy (kcal)	2335	±1490	2283	±1400	1.42	.15
Protein (% kcal [g])	18.0	±4.25	17.8	±4.46	1.39	.16
Total carbohydrates (% kcal [g])	46.8	±11.6	45.7	±11.3	3.64*	.0003
Starch (% kcal [g])	24.1	±7.10	23.3	±6.71	4.48*	<.0001
Simple carbohydrates (% kcal [g])	15.1	±9.64	14.7	±10.3	1.48	.14
Total fat (% kcal [g])	35.8	±10.8	37.1	±10.8	4.57*	<.0001
Saturated FA (% kcal [g])	12.3	±4.72	12.8	±4.93	3.93*	<.0001
MUFA (% kcal [g])	13.4	±4.70	14.0	±4.74	4.82*	<.0001
PUFA (% kcal [g])	7.43	±3.60	7.53	±3.37	1.09	.27
Dietary fiber (g/1000 kcal [grams])	10.9	±5.92	10.7	±6.03	1.57	.12
Alcohol (% kcal [g])	.970	±3.86	.860	±3.32	1.16	.25
Cholesterol (mg/1000 kcal)	141	±91.4	152	±99.7	4.36*	<.0001
Sodium (mg/1000 kcal)	1873	±556	1881	±567	.54	.59
Exercise, 1-4 scale	1.84	±.98	1.81	±.96	.04	.96
Tobacco use, pack-years	3.84	±16.9	4.30	±18.1	.50	.62

CI = confidence interval; FA = fatty acids; MUFA = monounsaturated fatty acids; PUFA = polyunsaturated fatty acids

^*Bonferroni corrected P <.004

Dietary variation in this DCCT cohort was relatively small (Table 3).

Table 3.

Distribution of Saturated Fat, MUFA, and Fiber Intake Among DCCT Subjects (n = 1431)

	Saturated Fatty Acids (% kcal)	MUFA (% kcal)	Dietary Fiber (g/1000 kcal)	Cholesterol (mg/1000 kcal)
Maximum	21.5	26.0	26.7	466
Q1	14.0	15.3	12.5	169
Median	12.5	13.8	10.4	139
Q3	11.0	12.2	8.6	114
Minimum	4.3	5.1	4.8	21.3

MUFA = monounsaturated fatty acids; DCCT = Diabetes Control and Complications Trial

Progression of retinopathy positively correlated with mean arterial pressure, the LDL-C/HDL-C ratio, triglycerides, HbA1c, BMI, insulin use, and tobacco consumption (Table 4). When broken down by strata, the positive associations of the overall progression of retinopathy with final BMI and average insulin use were somewhat stronger (final BMI: r = .10, P < .0064 and r = .10, P < .0120 for primary prevention and secondary intervention, respectively; and average insulin use: r = .14, P = .0002 and r = .24, P < .0001 for primary prevention and secondary intervention, respectively).

Table 4.

Significant Risk Factors vs Rate of Retinopathy Progression

Retinopathy Risk Factors*	Prestudy Progression Rate †† (n = 1436)		Overall Progression Rate ‡‡ (n = 1412)
	r	P Value	r	P Value
Baseline/overall mean BP	.09‖	.0004	.20‖	<.0001
Baseline/overall LDL-C/HDL-C	.13‖	<.0001	.15‖	<.0001
Baseline/overall triglycerides	.18‖	<.0001	.26‖	<.0001
Baseline/overall HbA1c	.10‖	.0001	.45‖	<.0001
Baseline/final BMI	.08	.0034	.05	.08
Baseline/overall insulin use/kg ideal body weight	.19	<.0001	.14	<.0001
Baseline/overall exercise	.00	.89	.00	.99
Baseline/overall tobacco use§	.08	.0281	.09‖	.0011
Group (conventional = +, intensive = −)	−.01	.60	−.27‖	<.0001
Age (older = +, younger = −)	−.05	.09	−.08‖	.0039
Sex (males = +, females = −)	.03	.31	.08‖	.0020

BP = blood pressure; LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; HbA1c = glycosolated hemoglobin; BMI = body mass index

^*The baseline nutrient intake is related to the rate of progression of retinopathy at baseline (degree of retinopathy at baseline on the [ETDRS – 1]/the duration in months of pretrial diabetes) and with risk-factor levels measured at baseline. The overall average of the nutrient data is correlated with the overall rate of retinopathy progression (slope of the retinopathy progression determined by: [ETDRS at completion of the trial – 1]/time on study in months) and with the mean of all values for the risk factors.

^††The rate is the baseline ETDRS retinopathy score – 1/duration of diabetes prestudy.

^‡‡The ETDRS retinopathy score at the closeout of the study – 1/overall duration of diabetes

^§Total lifetime pack-years at baseline/total pack-years before the trial plus on-study: For baseline calculation, only secondary intervention stratum subjects (mild retinopathy at baseline) are included (n = 710).

^‖Bonferroni corrected P <.0055

There was no correlation between exercise and retinopathy progression rate (Table 4) and borderline positive associations of exercise with blood pressure and HbA1c (Table 5). The borderline positive correlation of exercise and HbA1c corresponds to the decline of both exercise and HbA1c with age (r = −.12, P < .0001 and r = −.15, P < .0001 for exercise and HbA1c, respectively). Exercise inversely correlated with lipid levels, final BMIs, and insulin use. In part, the trend toward exercise correlating with increased blood pressure relates to higher exercise levels in males (males: 1.92 ± 1.00 on the 1-4 scale vs females: 1.71 ± .89, t = 4.20, P < .0001) and higher blood pressures (males: 89.0 ± 11.6 mm Hg vs females: 84.7 ± 10.6 mm Hg, t = 7.32, P < .0001). When broken down by sex, exercise is inversely associated with blood pressure in females (r = −.13, P < .0001) and not correlated in males (r = .03, P = .40). Only 1% (n = 15) averaged more than moderate exercise while on-study.

Table 5.

Nutrient Intake, Exercise, and Tobacco Use Correlated With Selected Risk Factors

Average Lifestyle Factors	Mean Systolic Blood Pressure (n = 1431)		Mean LDL-C/HDL-C (n = 1431)		Mean Serum Triglycerides (n = 1431)		Mean HbA1c (n = 1431)		Final BMI		Mean Insulin Use/kg Ideal Body Weight
	r	P Value	r	P Value	r	P Value	r	P Value	r	P Value	r	P Value
Energy (kcal)	.19§	<.0001	.16§	<.0001	.08§	.0020	.10§	.0002	.01	.71	.10	<.0001
Protein*	.01	.80	−.01	.79	−.04	.17	−.04	.11	.03	.24	−.06	.0264
Total carbohydrates*	−.12§	<.0001	−.06	.0251	−.09§	.0006	−.03	.34	−.08	.0040	−.16§	<.0001
Starch*	−.03	.21	.01	.59	−.00	.86	−.01	.68	−.03	.29	−.04	.15
Sugars*	−.08	.0037	−.07	.0108	−.03	.23	.03	.23	−.06	.0145	−.07	.0103
Total fats*	.08	.0043	.08§	.0024	.09§	.0011	.06	.0250	.03	.22	.17§	<.0001
Saturated fats*	.08	.0032	.08§	.0026	.12§	<.0001	.06	.0235	.05	.06	.18§	<.0001
MUFA*	.09§	.0006	.10§	.0001	.10§	.0002	.06	.0358	.04	.15	.18§	<.0001
PUFA*	−.01	.70	.00	.89	−.02	.46	.03	.21	−.01	.83	.04	.18
Fiber††	−.08	.0021	−.12§	<.0001	−.16§	<.0001	−.05	.07	−.05	.0533	−.28§	<.0001
Alcohol*	.10§	.0002	−.06	.0265	.01	.58	−.02	.51	.03	.29	−.04	.18
Cholesterol‡‡	.07	.0099	.06	.0160	.07	.0059	.05	.08	.07	.0126	.09§	.0003
Sodium‡‡	−.02	.39	.03	.24	.08§	.0020	.05	.0532	.06	.0140	.12§	<.0001
Exercise (1-4 scale)	.04	.09	−.07	.0131	−.08§	.0034	.05	.0464	−.13§	<.0001	−.12§	<.0001
Tobacco (pack-yrs)	−.02	.36	.13§	<.0001	.17§	<.0001	.09§	.0009	−.05	.08	.05	.0530

LDL-C = low-density lipoprotein cholesterol; HDL-C = high-density lipoprotein cholesterol; HbA1c = glycosolated hemoglobin; BMI = body mass index; MUFA = monounsaturated fatty acids; PUFA = polyunsaturated fatty acids

^*% of kcal

^††g per 1000 kcal

^‡‡mg per 1000 kcal

^§Bonferroni corrected P <.003

The significantly higher rate of progression of retinopathy in male subjects (.0175 [.152] vs .0156 [.0158] ETDRS step units per month in females, t = 2.30, P = .0218) is associated with greater proportions in males compared with females of dietary total fat [SD] (37.4 [5.14]% vs 35.4 [5.60]%, t = 7.03, P < .0001), saturated fat [SD] (12.9 [2.37]% vs 12.1 [2.42]%, t = 6.17, P < .0001), and monounsaturated fat [SD] (14.2 [2.29]% vs 13.2 [2.34]%, t = 6.03, P < .0001) and less total carbohydrate [SD] (45.1 [5.62]% vs 47.6 [5.62]%, t = −8.57, P < .0001) and dietary fiber [SD] (10.2 [2.75] vs 11.5 [3.09], g/1000 kcal, t = −8.71, P < .0001).

Compared with the conventional treatment group, the intensive therapy group consumed fewer calories [SD] (2326 [742]% vs 2150 [709]%, t = 4.60, P < .0001) and had a lower proportion of calories as simple carbohydrates [SD] (15.3 [5.93]% vs 13.7 [5.27]%, t = 4.58, P < .0001). The conventional treatment group had a lower proportion of kilocalories as saturated fatty acids [SD] (12.1 [2.59] vs 12.4 [2.57]%, t = 2.20, P = .0277 but a higher intake of saturated fat [SD] (32.0 [15.1]g vs 30.2 [13.2]g, t = 2.40, P = .0162).

(Table 6) shows that the correlation between the nutrient variables and the progression of retinopathy from the onset of diabetes to the closeout of the trial corresponded well with the prestudy associations. The proportions of dietary carbohydrates, protein, and dietary fiber are inversely associated with retinopathy progression. The proportion of cholesterol and dietary fat – particularly saturated and monounsaturated fat – positively correlate with retinopathy progression.

Table 6.

Nutrient Intake Correlated With Retinopathy Progression Rate

Nutrient Intake Baseline/On-Study*	Baseline Progression Rate†† (n = 1436)		Overall Progression Rate‡‡ (n = 1412)
	r	P Value	r	P Value
Food energy (kcal)	.01	.65	.07	.0073
Protein (% kcal)	−.08	.0043	−.06	.0188
Total carbohydrate (% kcal)	−.13§	< .0001	−.11§	< .0001
Starch (% kcal)	−.10§	.0001	−.08§	.0014
Simple carbohydrate (% kcal)	−.04	.09	−.05	.0592
Total fatty acids (% kcal)	.15§	< .0001	.14§	< .0001
Saturated fatty acids (% kcal)	.16§	< .0001	.11§	< .0001
MUFA (% kcal)	.13§	< .0001	.12§	< .0001
PUFA (% kcal)	.02	.45	.09§	.0004
Dietary fiber (g/1000 kcal)	−.07	.0102	−.10§	.0002
Alcohol (% kcal)	.03	.19	.03	.26
Cholesterol (mg/1000 kcal)	.13§	< .0001	.10§	.0003
Sodium (mg/1000 kcal)	.02	.50	.02	.47

MUFA = monounsaturated fatty acids; PUFA = polyunsaturated fatty acids

^*The baseline nutrient intake is related to the rate of progression of retinopathy at baseline (degree of retinopathy at baseline on the [ETDRS – 1]/the duration in months of pretrial diabetes) and with risk-factor levels measured at baseline. The overall average of the nutrient data is correlated with the overall rate of retinopathy progression (slope of the retinopathy progression determined by: [ETDRS at completion of the trial – 1]/time on study in months) and with the mean of all values for the risk factors.

^††The rate is the baseline ETDRS retinopathy score – 1/duration of diabetes prestudy.

^‡‡The ETDRS retinopathy score at the closeout of the study – 1/overall duration of diabetes

^§Bonferroni corrected P < .004

The relationship between retinopathy-associated risk factors and nutritional intake (Table 5) is similar to that of the direct relationship between diet and retinopathy progression (Table 6). Alcohol increases mean arterial pressure and decreases the LDL-C/HDL-C ratio. Caloric intake is positively associated with the progression of retinopathy and the retinopathy-associated risk factors only indirectly, because high caloric intake is inversely correlated with the proportion of protein and carbohydrate consumption (r = −.25, P < .0001 and r = −.29, P < .0001, respectively) and directly associated with the proportion of calories as fat (r = .29, P < .0001).

Compared with the conventionally treated group, the intensive therapy group had a 57.2% lower rate of progression of retinopathy while on-study (means [SD] = .0313 [.0350] vs .0134 [.0192] ETDRS step units per month in the conventional and intensive treatment groups, respectively). Conventionally treated subjects who adhered to the ADA dietary recommendations and maintained saturated fat intake at < 10% of calories and total fat at < 30% of calories while on-study (n = 102) compared with conventionally treated subjects who did not adhere to the ADA dietary guidelines (n = 617) showed a 33.2% reduced rate of retinopathy progression on-study (ETDRS [SD] step units per month = .0281 [.0315] vs .0188 [.0228] for those not adhering and adhering to ADA diet, respectively, t test = 3.82, P = .0002). The HbA1c level of the subjects adhering to the ADA diet was correspondingly lower than those not adhering to the diet (HbA1c = 8.65 [1.23] and 9.14 [1.30], respectively, t = 3.70, P = .0003). This effect of adhering to the ADA dietary recommendations was much attenuated in the intensive therapy group (adhered to ADA diet: [n = 78], mean retinopathy progression rate = .0113 [.0128] ETDRS scale steps per month vs those who did not adhere to ADA diet: [n = 619], mean retinopathy progression rate = .0130 [.0181] ETDRS scale steps per month, t = 1.05, P = .30). Those in the intensive therapy group who adhered to the ADA diet had a nonsignificantly lower HbA1c level (HbA1c = 7.10 [.808] and 7.26 [.902], ADA-adherent and nonadherent, respectively, t = 1.63, P = .11).

Discussion

The entire cohort demonstrated significant correlations between diet and retinopathy progression rate in the prestudy period Table 4), whereas when broken down by stratum, the correlations were absent or weaker (data not shown). These relationships probably reflect 2 things: that the DCCT dietary data cover 68% of the time with diabetes in the primary prevention stratum compared with only 44% in the secondary intervention stratum and that the secondary intervention stratum consumed a significantly higher proportion of calories as fat than the primary prevention stratum (Table 2).

The benefits of exercise on lipids, BMI, and insulin utilization (Table 5) are some of the many reasons to encourage people with type 1 diabetes to maintain fitness programs. The low overall level of exercise in the DCCT subjects probably accounted for the absence of a significant inverse correlation between exercise and retinopathy progression rate, blood pressure, and HbA1c (Table 4 and Table 5). The effect of regular, vigorous exercise on the progression of diabetic retinopathy should be prospectively studied, including subjects who are more active.

Smoking tobacco correlated with faster pretrial progression of retinopathy in the secondary intervention stratum and the overall progression of retinopathy in the entire cohort (Table 4). The positive correlation between tobacco use and dietary saturated fat and cholesterol intake documented previously in this database[36] may enhance any direct effect of smoking to promote retinopathy development.

The positive correlations between tobacco use and the progression of retinopathy (Table 4) and of the retinopathy-associated risk factors (Table 5) suggest that smoking accelerates microvascular disease, as it does for macrovascular disease. The absence of an adverse effect of smoking on mean arterial pressure may be due to the trend toward an inverse association of tobacco use with BMI (r = −.05, P = .08) and the direct correlation of BMI with mean arterial pressure (r = .28, P < .0001). Of course, smoking is a very poor method and not recommended to control weight and the associated hypertension.

Despite the strong influence of HbA1c on the progression of retinopathy (baseline r = .10, P = .0001 and on-study r = .45, P < .0001), we did not control for HbA1c when correlating the nutrient variables with progression of retinopathy. Controlling for HbA1c would introduce a bias because the diet significantly influences the HbA1c as well as blood pressure, serum lipids, BMI, and insulin utilization (Table 5). The relationships between saturated fat, MUFA, dietary fiber, and cholesterol and HbA1c as well as the other risk factors (Table 5) correspond well with the correlations between these nutrient variables and the progression of retinopathy (Table 6).

Limitations of this study include a retrospective analysis, in which post hoc analysis of DCCT data can generate hypotheses but not prove them. The single disease cohort, study subjects living only in the United States and Canada, and the narrow age range limit the generalizability of the findings somewhat. The analysis of fatty acids according to carbon length correlated with progression of retinopathy and retinopathy-related risk factors may also yield additional information.

The correlations reported here between dietary fat and carbohydrates are statistically significant but relatively weak (r < .25). The DCCT diet database represents only about 53% of the average subject's time with diabetes. The data do not capture any dietary changes that occurred before the trial began or between diet analyses. The importance of the pretrial diet is illustrated in (Table 1), in which the average subject's diet made a substantial shift from baseline to on-study, decreasing saturated fat and MUFA by 1.0% and .9%, respectively. The small variations in intake of nutrient variables (Table 3) and the sedentary exercise patterns limited the range of risk reductions that could be demonstrated from statistically analyzing diet and exercise.

Ideally, to estimate the clinical effect of saturated fat, MUFA, dietary fiber, cholesterol, and other variables on the progression of retinopathy, computerized diet data repeated several times per year from the onset of diabetes should be analyzed, and the interquartile range of the saturated fat percentage, MUFA percentage, and dietary fiber grams/1000 kcal should all be at least 10. The interquartile range of cholesterol milligrams/1000 kcal should be greater than 150. Consequently, the weak correlation of nutrient variables and retinopathy progression is not deemed to be a serious limitation, as this approach is relatively conservative and the relationships are hypothesized to be stronger with more frequent assessments and greater variation between subjects.

Notwithstanding these limitations, this analysis demonstrates that the percentages of the total caloric intake contributed by various nutrients may partially determine the rate of progression of retinopathy and retinopathy-associated risk factors. Our findings illustrate the need to conduct further cross-cultural research incorporating multiple outcomes. Prospective studies should be designed to test the relationship between diet and the progression of retinopathy and incidence of retinopathy-associated risk factors suggested here. The positive correlation of saturated fat and MUFA and the negative association of carbohydrates and dietary fiber with retinopathy progression and retinopathy risk factors in this study suggest but do not prove that animal-based food products promote retinopathy much more than plant-based products. Animal products are high in saturated fat and MUFA and have no dietary fiber.

This retrospective analysis confirms that people with type 1 diabetes should reduce their total dietary fat to less than 30% and the saturated fat to less than 10%, as was formerly recommended by the ADA. However, further research should also be done to determine whether an even more plant-based diet would be of greater benefit in preventing the progression of diabetic retinopathy and other microvascular complications. Finally, this analysis should be updated when data from the Epidemiology of Diabetes Interventions and Complications (EDIC) study[37] (ie, follow-up to the DCCT) are made available to the public.