Kohner EM, Aldington SJ, Stratton IM, Manley SE, Holman RR, Matthews DR, Turner RC, for the United Kingdom Prospective Diabetes Study. United Kingdom Prospective Diabetes Study, 30Diabetic Retinopathy at Diagnosis of Non–Insulin-Dependent Diabetes Mellitus and Associated Risk Factors. Arch Ophthalmol. 1998;116(3):297-303. doi:10.1001/archopht.116.3.297
To report on the prevalence of retinopathy in patients with newly diagnosed non–insulin-dependent diabetes mellitus (NIDDM) and to evaluate the relationship of retinopathy to clinical and biochemical variables.
A multicenter, randomized, controlled clinical study of therapy in patients with NIDDM.
Setting and Patients
Patients were part of the United Kingdom Prospective Diabetes Study, a 23-center study of 2964 white patients who had both eyes photographed and assessed.
The presence and severity of diabetic retinopathy were evaluated by sex, and the relationship of retinopathy to medical and biochemical parameters was assessed.
Retinopathy, defined as microaneurysms or worse lesions in at least 1 eye, was present in 39% of men and 35% of women. Marked retinopathy with cotton wool spots or intraretinal microvascular abnormalities was present in 8% of men and 4% of women. The severity of retinopathy was related in both sexes to higher fasting plasma glucose levels, higher systolic and diastolic blood pressure, lower serum insulin levels, and reduced β-cell function. In addition, in men, increased alcohol consumption was related to increased severity of retinopathy, while leaner women had more severe eye lesions. Visual acuity was normal in most patients, but in men there was a trend for those with more severe retinal lesions to have worse visual acuity.
Diabetic retinopathy is common in patients with newly diagnosed NIDDM. Careful ophthalmic assessment at diagnosis is important.
THE UNITED Kingdom Prospective Diabetes Study (UKPDS) is a multicenter, randomized, controlled clinical trial of therapy in patients with newly diagnosed non–insulin-dependent diabetes mellitus (NIDDM) that aims to determine whether improved blood glucose control will prevent microvascular and macrovascular morbidity and mortality and whether any specific therapy is advantageous or disadvantageous. Details of the study design and randomization have been described previously.1,2 In brief, patients with newly diagnosed NIDDM were treated for 3 months with diet only, and if the fasting plasma glucose level was not adequately controlled at the end of this time, were randomized into treatment by diet only, oral agents, sulfonylureas (chlorpropamide, glyburide, or glipizide), or insulin. In addition, obese patients (≥120% of ideal body weight) were also randomized to be treated with the biguanide agent metformin. The study started in 1977, and retinal photography was included during 1983.
This report describes the retinal status of the patients at entry into the UKPDS and relates this to various clinical and biochemical variables. This gives insight into possible risk factors for the development of retinopathy and highlights the importance of full medical and eye examination at diagnosis of NIDDM.
Patients with newly diagnosed NIDDM (fasting plasma glucose level >6 mmol/L [>108 mg/dL] on 2 occasions) between the ages of 25 and 65 years were recruited into the study between 1977 and 1991. Exclusion criteria were severe vascular disease, accelerated hypertension, preproliferative or proliferative or previously photocoagulated retinopathy, renal failure (serum creatinine level >175 µmol/L [>2.0 mg/dL]), life-threatening disease, disease requiring steroid treatment, or occupation precluding treatment with insulin (eg, bus driver). The patients gave informed consent, and the study was approved by the ethics committee of each participating center.
A total of 5102 patients were recruited into the study. Of these, 4177 were white, while the others were mostly Asian from the Indian subcontinent or Afro-Caribbean. The present analysis deals only with the 2964 whites who entered after 1983 (when retinal photography started) and who had both eyes photographed and assessed.
Best corrected visual acuity was measured using the Snellen chart and, since 1986, the Early Treatment Diabetic Retinopathy Study logMAR charts.3 Earlier Snellen chart data were corrected to their equivalent logMAR values. After pupillary dilation, eye examination was carried out using direct ophthalmoscopy. Four standard 30° fields of each eye were photographed in stereo pairs (lateral to macula, macula, disc, and nasal) and analyzed centrally, as described by Aldington et al.4 All identifying information was masked, and each photograph was allocated a unique identification number. Assessments were carried out in duplicate, with independent adjudication of differences between graders.
Photographs were initially assessed for quality and adherence to protocol, then were graded in comparison to Early Treatment Diabetic Retinopathy Study standard photographs5 for severity of retinopathy features. Subsequently, a computer algorithm assigned a numerical level to each eye (the modified Wisconsin level),6 summarizing the severity of the lesions found. A simplified description of the levels allocated is shown in Table 1. As a modified Wisconsin level was assigned to each eye individually, for the purpose of this report, results are categorized based on the worse eye/better eye system.7 Thus, 10/10 indicates absence of any retinopathy in both eyes; 20/10, microaneurysms in 1 eye only; and 20/20, both eyes with microvascular abnormalities only. Similarly, 31/<31 indicates that the worse eye has some retinal hemorrhages and/or hard exudates in addition to microvascular abnormalities, with the better eye having less retinopathy, while 31/31 also indicates those features, but in both eyes. This yields a possible 21-point scale (10/10 to 75/75) for each patient.
At entry, a full medical history was obtained, including details of physical activity. Patients were categorized as sedentary (rarely participates in physical activity), moderately active (on their feet less than half the day, weekend exercise only), active (on their feet more than half day, bicycle to work, regular exercise), or fit (manual worker, or regular and vigorous exercise at least 3 times each week). The history also included smoking (never, ex-smoker, current smoker) and alcohol consumption (none, occasional or social, regular, heavy). A full medical examination was carried out, including measurement of height and weight. Body mass index was calculated by dividing the weight in kilograms by the square of height in meters.
Each participating center measured fasting plasma glucose levels. These measurements were monitored by the UKPDS glucose quality assurance scheme, with an interlaboratory coefficient of variation of approximately 4%.1 All other samples were transported to the coordinating laboratory overnight at 4°C. Assays were monitored by trilevel quality-control serum samples with Westgard rules8 for acceptance and by external quality-control schemes where possible.
Analytical methods were upgraded during the study, and data were realigned according to new methods after relevant laboratory comparisons.9 Data in this report are presented using laboratory methods previously described.9
Glycosylated hemoglobin levels were measured by high-pressure liquid chromatography (Biorad Diamat automated glycosylated hemoglobin analyzer, Biorad Laboratories Ltd, Hemel Hempstead, England) (normal range, 0.045-0.062 [n=145], 2.5-97.5 percentiles). Urine albumin concentrations were expressed relative to the mean creatinine concentration of 8 mmol/L (90 mg/dL) in women and 11 mmol/L (125 mg/dL) in men to allow for urine dilution.9 Fasting insulin levels were measured by double antibody radioimmunoassay (PhRIA 100, Pharmacia Ltd, Milton Keynes, England) with 100% cross-reactivity to proinsulin (normal range, 21-111 pmol/L [2.9-15.5 µU/mL]). β-cell function and insulin sensitivity were assessed using fasting plasma glucose and insulin levels by homeostasis model assessment.10 This assesses, in a structural model of glucose-insulin relationships, the degree of impairment of β-cell function and insulin sensitivity required to produce each patient's fasting plasma glucose and insulin levels. The model was calibrated to give β-cell function and insulin sensitivity of 100% in a group of normal subjects aged 18 to 25 years. Plasma lipid and lipoprotein, urine albumin, and urine creatinine levels were measured by standard laboratory techniques.9
All data were analyzed using software from SAS Institute Inc11 and BMDP Statistical Software Inc.12 Continuous variables are described in the form mean (SD) if approximately normally distributed or in the form geometric mean (1-SD range) if approximately log-normal. We used a polychotomous logistic regression model to identify important variables and to estimate odds ratios. In detail, this entailed examining the association between retinopathy level and covariates 1 by 1 in a univariate model (procedure PR in the BMDP software) to investigate, for example, whether those with more severe retinopathy at diagnosis had higher levels of plasma glucose. Since the variables measured at diagnosis were not independent, this model was then used, with stepwise selection of variables, to arrive at a subset of variables that together best explained the retinopathy level. The continuous variables chosen by this method were split into quartiles, and the PR procedure was used to calculate the odds ratios. P<.02 was considered statistically significant.
Of the 4177 white patients, 3315 had baseline photographs, and 3185 (96.1%) of these were taken within the allowed window of 6 months before to 18 months after randomization. In 2972 patients (93.3%), photographs of both eyes could be graded. Photographs of only 1 eye could be graded in 112 patients (3.5%), and in 101 patients (3.2%) neither eye was gradable. Eight patients had photocoagulation before randomization; these patients were excluded from the analyses. The data presented, therefore, are for the 2964 patients in whom both eyes were graded. The quality of the photographs was high; only 1.7% of photographs temporal to the macula, 3.5% of the macula photographs, 2.5% of the disc photographs, and 5.5% of the nasal photographs could not be graded. In these patients, retinopathy level was assigned on the basis of the remaining fields.
Lesions other than diabetic retinopathy were present in some eyes. The most common of these were extensive drusen, present in 270 eyes; a nevus in 52 eyes, pigment epithelial defects in 55 eyes, and retinal branch vein occlusion in 5 eyes. None of these conditions or other nondiabetic conditions excluded grading for diabetic retinopathy.
Retinopathy was present in 39% of the men and 35% of the women. However, 19% of the men and 20% of the women had microaneurysms in 1 eye only, with no other features of diabetic retinopathy. In most of these patients (97%) there were only 3 or fewer microaneurysms. More severe forms of retinopathy were significantly more common in men (χ2 for trend, P<.001), with a retinopathy grade of 41/<41 or more severe in 7.9% of men and 4.5% of women (Table 2).
Demographic and clinical results are shown in Table 3. Women were older and more obese at diagnosis and had higher blood pressure than men. They had higher fasting plasma glucose levels (P<.001) and slightly higher glycosylated hemoglobin levels (P<.02). Plasma cholesterol levels and both high-density lipoprotein and low-density lipoprotein cholesterol levels were higher in women (P<.001 for all). Women also had higher insulin levels and were more insulin resistant (P<.001 for both). Women were less active but smoked and drank less than men; more women were treated with diuretic agents, and more were receiving antihypertensive therapy (P<.001 for all).
In both sexes the presence and severity of retinopathy and demographic variables were associated univariately (Table 4), with higher fasting plasma glucose levels (P=.02 for men, P=.002 for women) and higher systolic (P=.006 for men, P<.001 for women) and diastolic (P=.009 for men, P=.004 for women) blood pressures. More severe retinopathy was also associated with lower plasma insulin levels (P=.005 for men, P=.002 for women) and reduced β-cell function (P<.001 for men and women). In addition, in men, higher high-density lipoprotein cholesterol levels (P=.009) and increased alcohol consumption (P=.005) were also associated with more severe retinopathy, while in women there was a significant trend for those who were leaner (P=.005) and who had higher glycosylated hemoglobin levels (P=.005) to have more severe lesions. Triglyceride levels, low-density lipoprotein cholesterol levels, smoking, physical exercise, treatment with antihypertensive or diuretic therapy, and, in women, hormone replacement therapy did not appear to be related to severity of retinopathy.
When patients (both men and women) with microaneurysms in only 1 eye (level 20/10) were compared with those who had no retinopathy (level 10/10), they were not significantly different in any of the clinical variables. When they were compared with patients who had more severe retinopathy (levels ≥20/20), there were some differences (Table 4), suggesting that patients with a retinopathy level of 20/10 resembled those without retinopathy.
Table 5 presents the results of a polychotomous logistic regression analysis that associates increasing severity of retinopathy with variables commonly measured in routine clinical practice, blood pressure and fasting plasma glucose level. After adjusting for systolic blood pressure and fasting plasma glucose level, female subjects presented with less severe retinopathy at diagnosis. The risk of increasing severity of retinopathy increased with higher fasting plasma glucose level and was about 30% higher for those in the top half of the distribution. The odds ratio for systolic blood pressure rose steadily with each quartile of systolic blood pressure: 11% higher in the second quartile, 23% higher in the third quartile, and 31% higher for the top quartile.
A stepwise procedure was used to determine which variables found to be significant in the univariate model were significant in the multivariate model. After adjusting for sex, 3 variables remained in the model: β-cell function, fasting plasma insulin level, and systolic blood pressure (Table 6). Systolic blood pressure was strongly associated with steadily increasing risk of retinopathy. Fasting plasma insulin level below the top quartile was associated with higher levels of retinopathy. β-Cell function in the lower half of the distribution was associated with increased risk of retinopathy.
The great majority of patients had normal visual acuity of 0.0, 0.1, or less than 0.2 on the logMAR chart (6/5 or better and 6/6 on the Snellen chart) in their worse eye at diagnosis (Table 7). Among the men, there was a significant trend for worsening visual acuity with increasing retinopathy severity (P=.005). There was no similar relationship among the women. Visual acuity was worse in the women than in the men (P<.001).
Among white patients with newly diagnosed NIDDM referred from primary care physicians, diabetic retinopathy was present at diagnosis in 39% of the men and 35% of the women. This figure is consistently higher than in other studies, which have recorded prevalences of retinopathy at diagnosis of 18% to 22%.13,14 Most previous studies used ophthalmoscopy for the diagnosis of diabetic retinopathy. Even when it is performed by skilled observers, 1 to 3 microaneurysms may well be missed on ophthalmoscopy. It could be argued that a few microaneurysms without any other lesions do not indicate diabetic retinopathy. Klein et al15 reported finding microaneurysms among subjects without diabetes. The fact that the clinical characteristics of patients with microaneurysms in only 1 eye resembled those of patients without retinopathy suggests the possibility that the microaneurysms were an incidental finding. However, follow-up of these patients16 indicates that microaneurysms alone are important predictors of progression to more severe retinopathy. Furthermore, the 4- and 10-year follow-up data of Wisconsin patients with microaneurysms only17,18 also emphasizes the importance of microaneurysms as a risk factor for both proliferative retinopathy and macular edema. It is unlikely that these patients had undiagnosed diabetes for a longer period than patients in other studies,19 as the National Health Service enables patients to present early without financial disincentives. Furthermore, primary care physicians were encouraged to send all their newly diagnosed patients to the hospital in the districts of the participating centers.
Increased systolic and diastolic blood pressures were associated with the severity of retinopathy. This might signify that patients with microvascular disease in the retina may also have renal involvement and secondary higher systolic blood pressure. However, increased urine albumin concentrations showed only a weak association with retinopathy in women and no association in men, and albuminuria was not included in the multivariate model. The data therefore suggest that higher systolic blood pressure per se is probably an important determinant of retinopathy. This is in accordance with the theory that hypertension, by increasing retinal blood flow, is important in the evolution of diabetic retinopathy.20,21
Retinopathy was also associated with increased fasting plasma glucose levels in women and showed a similar trend in men. This would have been anticipated from the Diabetes Control and Complications Trial22 and the Kumamoto study.23 Measurement of β-cell function from fasting glucose and insulin levels by homeostasis model assessment10 showed an even stronger relationship of retinopathy with impaired β-cell function. Since the increasing hyperglycemia of NIDDM is associated with progressive deterioration of β-cell function,24 it is possible that the assessment of β-cell function provides a better guide to the severity of diabetes and the hyperglycemic exposure over the previous years than the measurement of fasting plasma glucose and glycosylated hemoglobin levels at a single clinic visit. The lesser association of retinopathy with these indices of glycemia might in part be because some patients had already restricted their diet between the time when the diagnosis was suspected by their primary care physician and their first clinic visit.
Retinopathy was more prevalent in men than women, but this could not be explained by the major risk factors, since women had higher blood pressure and glucose levels than men. Men were less obese than women, but body mass index was not related to the presence or severity of retinopathy in men, while in women the leaner subjects had more severe retinopathy. Retinopathy was not associated with greater impairment of insulin sensitivity, although this has been suggested in a small study of NIDDM patients using clamp studies.25
Other factors thought to be of importance in some previous studies, such as smoking26,27 and exercise, were not found to be important. These results emphasize the importance of large studies for obtaining definitive data.
In conclusion, diabetic retinopathy was common in patients with newly discovered NIDDM, although much of this was restricted to just a few microaneurysms in 1 eye only. This high prevalence underlines the importance of detailed ophthalmic investigation of newly diagnosed patients. The major associated variables were higher systolic blood pressure and hyperglycemia. Ongoing large-scale trials—the Hypertension in Diabetes Study,28 in which patients have been allocated to improved blood pressure control, and the UKPDS, in which patients are allocated randomly to improved glucose and blood pressure control—will determine whether careful attention to these risk factors will reduce the incidence of related clinical end points, such as reduced visual acuity and need for photocoagulation.
Accepted for publication December 1, 1997.
This work was supported by grant 2 UO1 EY07049-09 from the National Eye Institute, Bethesda, Md; by grants from the British Diabetic Association, London, England; Medical Research Council, London; National Institutes of Digestive Disorders and Kidney Disease, National Institutes of Health, Bethesda, Md; United Kingdom Department of Health, London; British Heart Foundation, London; by grants from pharmaceutical companies, including Bayer AG, Leverkeusen, Germany; Bristol-Myers Squibb Co, Princeton, NJ; Hoechst-Roussel Pharmaceuticals Inc, Somerville, NJ; Eli Lilly & Co, Indianapolis, Ind; Novo-Nordisk Pharmaceuticals Inc, Princeton, NJ; Lipha, Lyons, France; and Farmitalia Carlo Erba, Milan, Italy; and by grants from additional companies, including Securicor Plc, Sutton, England; Eastman Kodak Co, Rochester, NY; and Cortecs Diagnostics, Clwyd, Wales.
Participating centers included the Royal Infirmary, Aberdeen, Scotland; City Hospital, Belfast, Northern Ireland; Royal Victoria Hospital, Belfast; General Hospital, Birmingham, England; St Helier Hospital, Carshalton, England; Derbyshire Royal Hospital, Derby, England; Ninewells Hospital, Dundee, Scotland; Royal Devon and Exeter Hospital, Exeter, England; Ipswich Hospital, Ipswich, England; Leicester General Hospital, Leicester, England; St George's Hospital, London; Hammersmith Hospital, London; Whittington Hospital, London; Royal Infirmary, Manchester, England; Northampton General Hospital, Northampton, England; Norfolk and Norwich Hospital, Norwich, England; Radcliffe Infirmary, Oxford, England; Peterborough Hospital, Peterborough, England; Hope Hospital, Salford, England; Lister Hospital, Stevenage, England; North Staffordshire Royal Infirmary, Stoke-on-Trent, England; Torbay Hospital, Torbay, England.
Reprints: Eva M. Kohner, MD, Department of Endocrinology, Diabetes & Metabolic Medicine, Division of Medicine, UMDS Guy's and St Thomas's Medical and Dental School, Fourth Floor, North Wing, St Thomas's Hospital, Lambeth Palace Road, London SE1 7EH, England.