点击显示 收起
1 From the Department of Ophthalmology and Visual Sciences (AEM, RK, and JAM) and the Departments of Population Health and Biostatistical Medical Informatics (MP), University of Wisconsin Medical School, Madison; the Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis (ARF); and the Department of Nutrition, University of North Carolina, Chapel Hill (JS)
2 Supported by the National Institutes of Health grant EY11722 (to JAM), a Lions SightFirst grant from the American Diabetes Association, and the Research to Prevent Blindness. 3 Address reprint requests to JA Mares, University of Wisconsin, Department of Ophthalmology and Visual Sciences, 610 North Walnut Street, 405 WARF Building, Madison, WI 53705-2197. E-mail: jmarespe{at}facstaff.wisc.edu.
ABSTRACT
Background: The potential protective effect of vitamins C and E against the development of diabetic retinopathy has not been thoroughly evaluated in epidemiologic studies.
Objective: The objective was to study the association between prevalent diabetic retinopathy and intake of vitamins C and E in participants of the Atherosclerosis Risk in Communities Study.
Design: A total of 1353 subjects with type 2 diabetes diagnosed between 1993 and 1995 or before were included. Nutrient intake was assessed with a food-frequency and supplement questionnaire administered between 1987–1989 and 1993–1995. Prevalent retinopathy (n = 224) was determined in 1993–1995 from graded fundus photographs.
Results: No association of retinopathy with intake of vitamin C or E from food alone or from food and supplements combined was observed. The odds ratios and 95% CIs for retinopathy for quartile 4 compared with quartile 1 of vitamins C and E intakes from food and supplements combined were 1.1 (0.7, 1.9) and 1.3 (0.8, 2.2), respectively, after adjustment for diabetes treatment and serum glucose. There was a significant interaction of the observed relations with serum glucose concentration (P < 0.05). Additionally, a decreased odds of retinopathy was found among users (reported use =" BORDER="0">3 y before 1993–1995) of vitamin C or E supplements or multisupplements compared with reported use of no supplements: 0.5 (0.3, 0.8), 0.5 (0.2, 0.8), and 0.4 (0.2, 0.9), respectively.
Conclusion: No significant overall associations were observed between risk of retinopathy and intake of major dietary antioxidants. The observed association between risk of retinopathy and supplement use may reflect nondietary factors or a possible benefit of supplementation.
Key Words: Diabetes mellitus • type 2 diabetes • diabetic retinopathy • antioxidants • epidemiology
INTRODUCTION
The intake of vitamins C and E may protect against the development of diabetic retinopathy by affecting presumed pathogenic factors: protein glycosylation, insulin sensitivity, retinal blood flow, and oxidative stress (1-20). However, only 2 population-based studies have investigated the relation between retinopathy and these antioxidant nutrients, and the conclusions from each differ (21, 22). Additional research is needed to assess whether these micronutrients protect against retinopathy over the long term in the general population of persons with diabetes.
No epidemiologic data support a protective effect of dietary micronutrient intake on retinopathy. However, in vitro studies, animal studies, and short-term vitamin supplementation trials suggest that intake of vitamins C and E may protect against the development of retinopathy through many different mechanisms (1-20). These micronutrients have been shown to decrease or to be inversely associated with concentrations of glycated hemoglobin (1-9), and vitamin E has been shown to improve insulin sensitivity in short-term supplementation trials (10, 11). Vitamin C has been shown to act as an aldose reductase inhibitor of the hyperglycemia-induced polyol pathway, affecting retinal blood flow (23, 24) in animal models and human supplementation trials (12-15). However, the results of a clinical trial of an aldose reductase inhibitor did not prevent retinopathy (25). Vitamin E has been shown to inhibit the hyperglycemia-induced diacylglycerol protein kinase C pathway in both human retinal tissue and diabetic rat retinal tissue (16, 17). Increased activity of the diacylglycerol protein-kinase C pathway causes a decrease in retinal blood flow that is thought to promote microvascular diseases such as retinopathy (26, 27). Last, as antioxidants, vitamins C and E defend against the damaging effects of high oxidative stress owing to nonenzymatic glycosylation, autoxidative glycosylation, and metabolic stress in persons with diabetes (18). Oxidative stress has been shown to be positively associated with diabetic complications in some studies (19, 20).
The purpose of this research was to investigate the relation between intake of vitamins C and E at visit 1 (1987–1989) and the risk of prevalent retinopathy, as determined at visit 3 (1993–1995), in the Atherosclerosis Risk in Communities (ARIC) Study. We hypothesized that greater intake of vitamins C and E 6 y before the photographic assessment of retinopathy would be associated with a lower risk of retinopathy.
SUBJECTS AND METHODS
The Atherosclerosis Risk in Communities Study design
The ARIC Study is a prospective study designed to investigate the etiology of atherosclerosis (28). Four thousand adults aged 45–64 y were sampled from each of 4 US communities: Forsyth County, NC; Jackson, MS; suburban Minneapolis; and Washington County, MD. During the first visit from 1987 to 1989, a total of 15 792 persons participated. Participants were recruited to return 3 additional times after visit 1: visit 2 (1990–1992), visit 3 (1993–1995), and visit 4 (1996–1998). Dietary data were collected from participants at visits 1 and 3. Eye photographs of the participants were taken only at visit 3.
Sample selection
Any participant with a diagnosis of type 2 diabetes before or at visit 3 was considered to be at risk of retinopathy, because research has shown that 20% of persons with older-onset diabetes have retinopathy at diagnosis. This is most likely because of the delay in diagnosing type 2 diabetes or the increased sensitivity of retinal tissue to the damaging effects of glucose in older persons (29).
Participants were considered to have diabetes at visit 3 if their fasting blood glucose concentration was =" BORDER="0">126 mg/dL, their nonfasting glucose concentration was =" BORDER="0">200 mg/dL, they self-reported a medical history of diabetes (ie, stated that they had diabetes or "sugar in the blood"), or they used medications in the previous 2 wk for diabetes or high blood sugar. There were 1985 persons identified as having prevalent type 1 or 2 diabetes at visit 3. Participants were queried at visit 3 about their age at diagnosis of diabetes. If the age at diagnosis was missing but the participants were determined to have diabetes at visits 1, 2, or 3 on the basis of blood glucose values, not on the basis of medical history reports, then their age at diagnosis was defined as their age at that corresponding visit.
The sample selection process is outlined in Figure 1. Subjects were excluded if they had type 1 diabetes (n = 25), defined as an age at diagnosis of <30 y and use of insulin, or if their race was other than black or white (n = 9). Of the remaining 1951 subjects identified as having type 2 diabetes, 126 were excluded from the analyses because their age at diagnosis was missing and they were unable to be classified, as described above, as being diagnosed with diabetes at visits 1, 2, or 3. Other subjects were excluded because they had missing eye photo data (n = 328), missing dietary data from visit 1 (n = 36), or a dietary interviewer score that reported unreliable dietary data at visit 1 (n = 11). Additionally, persons were excluded if data for potential confounders measured at visit 1 were missing (n = 85) or if the vitamin supplement data at visit 1 were missing (n = 12). The final sample size included 1353 subjects. This was 69% of the starting sample of black and white subjects with type 2 diabetes at visit 3.
FIGURE 1.. Sample selection process.
Dietary data
Some persons with diabetes could have developed retinopathy years before visit 3 (1993–1995). For this reason, our primary focus was on the diet 6 y before the determination of retinopathy, visit 1 (1987–1989), as opposed to diet measured concurrently with the determination of retinopathy at visit 3, to better ensure that the diet exposure preceded the measurement of disease. Previous research has also supported the importance of analyzing diet and disease relations in the appropriate temporal sequence (30).
Dietary intake was assessed by using the Willett semiquantitative food-frequency questionnaire (FFQ) (31). The reliability of this questionnaire for intake of nutrients from food only, over a mean period of 3 y, was tested previously in the ARIC population (32). All of the participants’ total energy intakes fell within the following energy limits: >600 kcal and <4200 kcal for males and >500 kcal and <3600 kcal for females. Trained ARIC interviewers were used to administer the questionnaire, which was modified from its original form that contained 62 food items to one that contained 66 food items. The intake of daily nutrients and specific food items were calculated at the Channing Laboratory, Harvard Medical School, from the responses of servings per day of each food item on the questionnaire. The nutrient content of food items was determined from database provided by the US Department of Agriculture (33), although dietary fiber and vitamin E were determined by using other sources (34, 35). Estimates of intakes from food sources combined with supplement sources were used to compute total nutrient intakes at visits 1 and 3.
Vitamin supplement data
Supplement use was determined from a questionnaire administered at visits 1, 2, and 3 that queried for the brand name of supplements used. Because detailed information regarding the frequency, dose, and duration of specific supplements was provided only at visit 3, the estimated intake of supplements at visit 1 was imputed from visit 3 data (AE Millen, unpublished observations, 2000). The duration of supplement use was determined by the response to a question asked at visit 3. If this question was missing, the consistency of reported supplement use at visits 1, 2, and 3 was used to estimate the duration of supplement use. Persons who reported consuming supplements consistently between visits 1, 2, and 3 were assigned a dose for that supplement equal to the dose of that supplement that they reported taking at visit 3. If a person reported taking a supplement at visit 1 but not at visit 3, they were assigned the most commonly consumed dose for that supplement at visit 1, as determined by the previously imputed doses for supplement use at visit 1. For multivitamin supplements, the reported intakes of such supplements at visits 1, 2, and 3 defined the duration of use, rather than the self-reported use at visit 3.
Retinal photographs
The presence of retinopathy was assessed from retinal photographs taken at visit 3. A 45-degree nonmydriatic retinal photograph was taken of one randomly selected eye of each study participant. The photos were taken with a Canon CR-45UAF nonmydriatic camera (Canon USA, Itasca, IL), a camera that does not require pharmacologic dilation of the pupil. The slides were graded by trained graders at the ARIC Retinal Reading Center (Department of Ophthalmology, University of Wisconsin, Madison) in a masked fashion by using a modification of the Airlie House classification scheme (36). This procedure provided the following data for retinopathy: severity of retinopathy, number of microaneurysms, number and type of retinal hemorrhages, and presence of hemorrhages or microaneurysms, hard or soft exudates, macular edema, intraretinal microvascular abnormalities, or venous beading. The outcome, prevalent retinopathy, was defined as retinopathy of any of these specific lesions in the absence of other nondiabetic causes (eg, branch retinal vein occlusion).
Medical and demographic data
The duration of diabetes was calculated by subtracting the participant’s age at diagnosis from their reported age at visit 3. Persons diagnosed with diabetes at visit 3 were assigned a duration of diabetes of 1 mo. Other variables were defined as follows: 1) age groups: 50 y, >50 to 55 y, >55 to 60 y, and >60 y as determined at visit 1; 2) diabetes treatment group: exclusive dietary treatment of diabetes, use of oral hypoglycemic agents but not insulin, and insulin use; 3) hypertension: a diastolic blood pressure =" BORDER="0"> 90 mm Hg or a systolic blood pressure =" BORDER="0"> 140 mm Hg or the reported use of antihypertensive medications.
Statistics
The intake of vitamins C and E from food and from food and supplements combined were categorized by quartiles, with quartile 1 being the lowest intake and quartile 4 the highest. Logistic regression was used to estimate odds ratios (ORs) and 95% CIs for prevalent retinopathy in quartiles 2, 3, and 4 for intake of vitamins C and E compared with the lowest quartile, ie, the reference category. Tests for linear trends were done by using logistic regression models with continuous nutrient intakes from food and from food and supplements combined. When we evaluated the associations for vitamin C or E from food alone, we restricted the analysis to those persons who were not consuming supplemental sources of these nutrients.
To adjust for the effect of dietary energy, intake of both vitamins C and E were log transformed and regressed on total energy intake to compute residuals by using the sample population (n = 1353). The residuals were added to the nutrient intake in the sample population at the mean energy intake (37). Quartiles were based on energy-adjusted nutrient values. Total energy intake was added to all crude and adjusted models as a covariate.
Next, specific variables were evaluated as potential confounders: race, duration of diabetes, serum glucose (94% of persons in the sample had fasted for =" BORDER="0">8 h before the serum measurements were made), hypertension, body mass index, waist-to-hip ratio, smoking status, alcohol intake, drinking status, plasma triacylglycerols, plasma cholesterol, hematocrit value, prevalent coronary heart disease, diabetes treatment group, use of oral hypoglycemic agents, or use of insulin. Potential confounders were entered singly into the logistic regression model. If the addition of the factor changed the OR by =" BORDER="0">10%, it was considered to be a confounder. Because treatment group was determined to be a significant confounder in these analyses, no further adjustment for the covariates of insulin use or oral hypoglycemic use—to avoid the effects of multicollinearity—were made.
The consistency of the relations of nutrient intake to retinopathy was evaluated after stratification of the overall analyses by the hypothesized potential effect modifiers of race, duration of diabetes, glycemic control, and diabetes treatment group. Previous evidence for different profiles of risk for diabetic retinopathy in different racial groups (38, 39) suggests that the relation of diet to risk of retinopathy might vary by race. Evidence that the severity of diabetes influences other relations of diet to retinopathy (21, 40) suggests the possibility of similar nutrient and severity interactions in the present dataset. These potential interactions were explored by using the surrogate markers for severity: disease duration, glycemic control, and treatment group. A serum glucose concentration <140 mg/dL was chosen as the cutoff for good glycemic control. The likelihood ratio test was used to compare the logistic regression models with and without interaction terms. An level of 0.10 was used to define significance for interactions.
Additionally, the ORs and 95% CIs for retinopathy were calculated for persons who reported long-term use of supplements containing vitamins C or E. Long-term use was defined as supplement use for =" BORDER="0">3 y leading up to visit 3, as reported at visit 3, or use at both visits 1 and 2 compared with persons who reported no use of supplements containing vitamins C or E at visits 1, 2, or 3. Persons who did not fall into either of these 2 categories were excluded from these analyses. All analyses were conducted by using SAS software (version 8.2; SAS Institute Inc, Cary, NC).
RESULTS
Comparison of subjects included and excluded from the analyses
To understand the differences between the sample selected for this study (n = 1353) and those persons who were excluded (n = 598) from the original sample of subjects with type 2 diabetes (n = 1951), we compared demographic, lifestyle, and clinical characteristics of these 2 groups as reported at visit 1.
The excluded persons were more likely to be older (56 ± 0.23 compared with 55 ± 0.15 y; P = 0.0001), female (57% compared with 51%; P = 0.03), and black (48% compared with 32%; P < 0.0001) than were the persons who were included in the analyses. The excluded persons had a longer duration of diabetes (8 ± 0.40 compared with 6 ± 0.21 y; P < 0.001), had higher serum glucose values (162 ± 3.2 compared with 149 ± 1.7 mg/dL; P = 0.0001), had a greater incidence of hypertension (61% compared with 52%; P = 0.0008), and were less likely to be current drinkers (37% compared with 48%; P < 0.0001). The persons excluded from these analyses were more likely to use oral hypoglycemic agents (18% compared with 16%) or insulin (15% compared with 8%) at visit 1 than just diet to control their diabetes (P < 0.0001). Additionally, there was a greater prevalence of retinopathy (25%) at visit 3 among the excluded persons who also provided retinal photographs (n = 241) than among the persons who were included in the analyses (17%; P = 0.001). The 2 groups were not significantly different (P =" BORDER="0"> 0.05) with respect to body mass index, smoking status, plasma triacylglycerol concentrations, plasma cholesterol, hematocrit, prevalent coronary heart disease at visit 1, supplement use of vitamin C or E at visit 1, and intake of vitamin C, vitamin E, carbohydrates, protein, and total fat (data not shown).
Relations of retinopathy to intake of vitamins C and E
In a comparison of participants in quartile 4 with participants in quartile 1 for the amount of vitamin C intake from food alone or from food and supplements combined, participants in quartile 4 were more likely to be female, to be black, to be hypertensive, to have never smoked or drank, to have used oral hypoglycemic agents or insulin, to have prevalent retinopathy, and to have greater body mass indexes (Table 1). A higher intake of vitamin C from food and from food and supplements combined was also related to the intake of other dietary components. Compared with participants in quartile 1, participants in quartile 4 were also more likely to consume a higher daily percentage of energy from carbohydrates and a lower daily percentage of energy from total fat at visit 1. Seventeen percent (n = 230) of the participants included in this study used vitamin C supplements at visit 1.
View this table:
TABLE 1. Characteristics of the Atherosclerosis Risk in Communities Study participants with type 2 diabetes in quartile (Q) 1 and Q4 for daily median intakes of energy-adjusted vitamins C and E, 1987–19891
In a comparison of participants in quartile 4 with participants in quartile 1 for the amount of vitamin E intake from food alone or from food and supplements combined, participants in quartile 4 were more likely to be female and black, less likely to be current drinkers, more likely to use oral hypoglycemic agents and insulin, more likely to have prevalent retinopathy, more likely to have had diabetes for a longer duration of time, and more likely to have lower hematocrit values (Table 1). A higher intake of vitamin E from food and from food and supplements combined was also related to the consumption of a greater daily percentage of energy from protein. Fifteen percent (n = 205) of the participants included in this study used vitamin E supplements at visit 1.
The overall relations between intake of vitamins C and E and retinopathy are shown in Table 2. In crude analyses, intake of both vitamins C and E from food in quartile 4 compared with quartile 1 was related to a 2-fold higher odds of retinopathy. Intake of vitamins C and E from food and supplements combined in quartile 4 compared with quartile 1 was also directly related to retinopathy in crude analyses. However, after adjustment for the confounders of treatment group and serum glucose, these associations were not statistically significant. Further adjustment of the vitamin C analyses for the percentage of energy as carbohydrates also did not influence the OR (data not shown). Additionally, adjustment of the nutrient analyses for other potential confounders also did not influence these ORs.
View this table:
TABLE 2. Crude and adjusted odds ratios (ORs) and 95% CIs for prevalent retinopathy among Atherosclerosis Risk in Communities Study participants with type 2 diabetes in quartile (Q) 1 compared with Q2, Q3, and Q4 for vitamins C and E intakes at visit 1, 1987–19891
Analyses were also conducted to investigate the relation between retinopathy and servings of reported food at visit 1 that are rich sources of vitamin C and E (data not shown). There was no statistically significant relation observed between retinopathy and intake of fruit and vegetables or intake of citrus fruit, a concentrated food source of vitamin C. There was no relation between retinopathy and servings of the vitamin E-rich food products margarine, peanut butter, cereals, grains, and nuts.
Further analyses were conducted to see whether the associations between these nutrients and retinopathy varied by race, duration of diabetes, glycemic control, or treatment group. The association between retinopathy and intake of vitamins C and E varied by glycemic control (Table 3). The ORs for retinopathy in quartile 4 compared with quartile 1 for vitamin intake were always <1.0 for persons with better glycemic control (serum glucose < 140 mg/d) and >1.0 for persons with poor glycemic control (serum glucose =" BORDER="0"> 140 mg/dL). There was a statistically significant increased risk of retinopathy with increasing intake of vitamin C from food and supplements combined in persons with poor glycemic control (P for trend = 0.04). For all other associations, no statistically significant trends across quartiles were observed.
View this table:
TABLE 3. Crude and adjusted odds ratios (ORs) and 95% CIs for prevalent retinopathy stratified by serum glucose (< or =" BORDER="0">140 mg/dL) as a marker for glycemic control among Atherosclerosis Risk in Communities Study participants with type 2 diabetes in quartile (Q) 1 compared with Q2, Q3, and Q4 for vitamins C and E intakes at visit 1, 1987–19891
The relation between retinopathy and the intake of vitamin E also varied by race (P 0.01; Table 4). There was a direct relation between retinopathy and intake of vitamin E from food and supplements combined among whites (P for trend = 0.04). The same relation among blacks was inverse, but not statistically significant. For all observed associations, the ORs showed no clear indication of a dose response with respect to risk of retinopathy.
View this table:
TABLE 4. Crude and adjusted odds ratios (ORs) and 95% CIs for prevalent retinopathy stratified by race among Atherosclerosis Risk in Communities Study participants with type 2 diabetes in quartile (Q) 1 compared with Q2, Q3, and Q4 for vitamin E intake at visit 1, 1987–19891
There was also a statistically significant interaction by treatment group (test for interaction, P = 0.003) in analyses of vitamin E from food and supplements combined. There was no relation between retinopathy and vitamin E intake in persons who exclusively used diet or insulin, but there tended to be an increased odds of retinopathy with increasing intake of vitamin E in persons using oral hypoglycemic agents for quartile 4 compared with quartile 1 for intake of vitamin E (OR: 2.8; 95% CI: 1.1, 6.9). However, this relation was not statistically significant over the full range of intakes (P = 0.19). There were no interactions between intake of vitamin C and retinopathy by race, duration of diabetes, or treatment group and no interaction between intake of vitamin E and duration of diabetes.
Odds of prevalent retinopathy and intake of vitamins C and E at visit 3
The results of analyses conducted to determine whether the intake of vitamins C and E at visit 3 was similarly or differently related to retinopathy than was the intake of these micronutrients at visit 1 are shown in Table 5. There was no overall statistically significant relation between prevalent retinopathy and intake of vitamin C or E.
View this table:
TABLE 5. Adjusted odds ratios (ORs) and 95% CIs for prevalent retinopathy among Atherosclerosis Risk in Communities Study participants with type 2 diabetes in quartile (Q) 1 compared with Q2, Q3, and Q4 for intake of energy-adjusted vitamins C and E intakes at visit 31
Vitamin supplement use
Analyses were conducted to investigate the association of supplement use with risk of retinopathy. ORs were computed for retinopathy in long-term supplement users compared with nonusers. Compared with nonusers, long-term users of supplements containing vitamins C or E or long-term users of multivitamin supplements had a 2-fold lower risk of prevalent retinopathy (Table 6).
View this table:
TABLE 6. Adjusted odds ratios (ORs) and 95% CIs for prevalent diabetic retinopathy among Atherosclerosis Risk in Communities Study participants with type 2 diabetes using supplements =" BORDER="0">3 y before visit 3 or at both visits 1 and 2 compared with persons who reported no use of supplements at visits 1, 2, and 31
DISCUSSION
We observed no relation between the overall intake of vitamins C and E from food and supplements combined at visit 1 and prevalent retinopathy at visit 3 in persons with type 2 diabetes. Other than the current study, only 2 other epidemiologic studies have investigated the relation between micronutrient intake and retinopathy (21, 22). Mayer-Davis et al (21) found that the intake of antioxidants was directly associated with retinopathy in the San Luis Valley Diabetes Study. In the third National Health and Nutrition Examination Survey, we previously found no association of serum ascorbic acid and -tocopherol concentrations with retinopathy in persons with type 2 diabetes (22). These 2 previous studies had limitations that may explain the inconsistencies in their results. Both the third National Health and Nutrition Examination Survey and the San Luis Valley Diabetes Study were limited by cross-sectional designs. Additionally, the San Luis Valley Diabetes Study was limited by the collection of only 24-h dietary recall data, which does not reflect long-term dietary intake.
We speculate that the inconsistency in overall associations observed between these nutrients and retinopathy in the current and previous studies might reflect either or both of the following: persons may modify their diet and supplement use the longer they have diabetes or may change their diets subsequent to receiving a diagnosis of diabetes. However, there was no interaction found between risk of retinopathy and vitamin C or E intake by duration of diabetes in this study (although the statistical power to investigate this was limited because of the small number of cases of prevalent retinopathy among individuals with newly diagnosed diabetes).
We also observed an increased odds of retinopathy with an increasing intake of both vitamins C and E in persons with poor glycemic control. This may reflect an adverse effect of high vitamin C or E intakes in persons with poor glycemic control. Vitamin C may compete with glucose for binding sites to tissues, thus exacerbating hyperglycemia (18). Alternatively, the direct association between vitamin C and retinopathy may have been observed because those with more severe diabetes treated their hypoglycemic episodes with orange juice. However, there was no observed association between citrus fruit consumption, which included orange juice, and retinopathy. The direct relation between intake of these vitamins and retinopathy in persons with poor glycemic control may be spurious, could reflect physiologic actions of these nutrients to promote retinopathy, or could reflect recent diet changes or supplement use as a result of diagnosed retinopathy or other diabetic complications.
It has been shown that relations between diet and the odds of retinopathy may differ across racial groups because of differences in risk factors (38, 39). We observed different associations between vitamin E intake and the odds of retinopathy among blacks and whites. Associations were direct in whites and inverse, although not statistically significant, in blacks. We speculate that these different associations may be explained by differences in risk factors, lifestyle changes, or a combination of both. However, these observations may have occurred by chance.
In contrast with the lack of an observed relation between retinopathy and intake of vitamins C and E from food and supplements combined, the use of supplements containing vitamins C or E or the use of multivitamin supplements for =" BORDER="0">3 y before 1993–1995 was inversely associated with prevalent retinopathy. This contrast may reflect a high degree of transient supplement use by persons with diabetes, particularly if symptomatic, that would result in a diminished ability to observe a protective association in analyses of retinopathy with intake of these vitamins from food and supplements combined as reported at visit 1. The statistically significant associations observed between retinopathy and the use of vitamin C and E supplements are consistent with the possibility that the use of these types of supplements long term, or other unmeasured attributes or lifestyle characteristics associated with supplement use, may protect against retinopathy.
Our analyses of sources of vitamins C and E from food and supplements combined at visit 1 were limited by the lack of information with regard to supplement dose at visit 1. Intakes of supplements at visit 1 were imputed from data reported at visit 3. Persons may have been misclassified with regard to their intake of vitamins C and E supplements at visit 1. This may have biased our estimates toward the null.
Additionally, misclassification of nutrient intake from the FFQ may have limited our ability to detect associations. We recognize that the energy-adjusted intake of vitamins C and E estimated with the use of the 66-item FFQ, on average, is lower than the absolute intake (31). The misclassification may have been exacerbated when combining nutrient contents in food with nutrient contents in supplements.
Our analyses may have been biased and were reduced in power by the exclusion of subjects with data missing from our analyses. Excluded persons were older and appeared to have more severe diabetes and a higher prevalence of retinopathy. However, included and excluded persons had similar intakes and similar uses of vitamin C and E supplements. This modestly supports the notion that, had these persons been included in these analyses, we still would have not seen the hypothesized association between nutrient intake and risk of retinopathy.
Another limitation of this study may have been the misclassification of the specific lesions of retinopathy, which would have reduced the power to detect an association. In ARIC, only one photograph of one randomly selected eye of each participant was obtained. Additionally, only nonmydriatic fundus photographs, which do not require the dilation of the pupil, were obtained and may have resulted in the misclassification of retinopathy status. It is estimated that 25–30% of retinopathy is missed when only one eye is included in the evaluation (41).
We may have been limited by a lack of power to detect an association when the effect of vitamin C and E on retinopathy was modest. There was 82% power to detect an OR of 0.60 for prevalent retinopathy in quartile 4 compared with quartile 1 of nutrient intake. However, this power calculation does not take into consideration the inevitable influence of error in measuring diet on statistical power, which is difficult to estimate.
In conclusion, we observed no relation between retinopathy and the intake of vitamins C and E from food or from food and supplements combined in the sample of ARIC participants with type 2 diabetes studied. Dietary antioxidant therapy did not appear to be an effective treatment for the prevention of diabetic complications in this sample and is associated with a higher risk of retinopathy in some subsamples of this population. However, there was a protective effect against retinopathy with the intake of vitamins C and E in supplements, which warrants further study. The large body of evidence from in vitro studies, animal studies, and short-term supplementation trials suggests that the intake of vitamins C and E may protect against the development of retinopathy; however, available epidemiologic data do not support a protective effect of dietary micronutrient intake against retinopathy. Changes in diets and supplement use among persons with diabetes may make it difficult to observe protective or adverse relations with the risk of retinopathy in epidemiologic studies.
ACKNOWLEDGMENTS
We thank the staff and participants in the ARIC Study for their important contributions and the ARIC Steering Committee, specifically Moyses Szklo, Maria Ines Schmidt, and David Couper, for their careful review of the manuscript.
AEM and JAM designed the study and analyzed and reported the data. RK, ARF, and JS helped with the study design, data interpretation, and manuscript preparation. MP participated as the primary statistical consultant and helped with manuscript preparation. The authors had no conflicts of interest.
REFERENCES