The prevalence of hypertension by age and sex in the Blue Mountains Eye Study, 1992 to 1994.
Examples of retinopathy lesions graded in subjects without diabetes, including microaneurysms (A), flame hemorrhages (B), and blot hemorrhages (C and D).
Age-adjusted prevalence of any retinopathy lesions (hemorrhages or microaneurysms) by systolic blood pressure quartiles and sex. An explanation of the systolic blood pressure quartiles for women and men is provided in the second footnote of Table 4.
Effect of 3 fasting plasma glucose (FPG) cut points for the diagnosis of diabetes on the relationship between retinopathy (either hemorrhages or microaneurysms) and systolic blood pressure quartiles in women. The odds ratio is adjusted for age and fasting plasma glucose level. The error bars indicate 95% confidence intervals. An explanation of the systolic blood pressure quartiles for women and men is provided in the second footnote of Table 4.
Effect of 3 fasting plasma glucose (FPG) cut points for the diagnosis of diabetes on the relationship between retinopathy (either hemorrhages or microaneurysms) and systolic blood pressure quartiles in men. The odds ratio is adjusted for age and fasting plasma glucose level. The error bars indicate 95% confidence intervals. An explanation of the systolic blood pressure quartiles for women and men is provided in the second footnote of Table 4.
Yu T, Mitchell P, Berry G, Li W, Wang JJ. Retinopathy in Older Persons Without Diabetes and Its Relationship to Hypertension. Arch Ophthalmol. 1998;116(1):83-89. doi:10.1001/archopht.116.1.83
To assess the prevalence and relationship of retinopathy lesions in older subjects without diabetes to systemic hypertension.
Three thousand six hundred fifty-four people aged 49 years or older attending the Blue Mountains Eye Study underwent a detailed eye examination, including medical history, blood pressure measurement, and fasting blood collection. Retinopathy lesions (hemorrhages and microaneurysms) were assessed from masked grading of stereoretinal photographs. Subjects with a history of diabetes or an elevated blood glucose level were excluded. Hypertension was defined as the current use of antihypertensive medications or an elevated blood pressure measurement on examination.
Retinopathy was present in 325 subjects without diabetes, a prevalence of 9.8% (95% confidence interval [CI], 8.9%-10.9%), which increased with age. An increased age-adjusted relative risk (RR) for retinopathy was found for women (RR, 1.67; 95% CI, 1.26-2.21) and men (RR, 1.47; 95% CI, 1.07-2.00) with hypertension. In people using antihypertensive medications, retinopathy prevalence was higher for uncontrolled compared with controlled blood pressure but was not related to hypertension duration. Significant (P=.02 to P=.001) trends were found between increasing blood pressure quartiles and age-adjusted retinopathy prevalence, a relationship that was maintained after adjusting for fasting plasma glucose level at 3 diagnostic cut points for diabetes.
This study supports the Beaver Dam Eye Study findings that retinal hemorrhages and microaneurysms are relatively frequent lesions in older people without diabetes and are significantly related to the presence and severity of hypertension.
SEVERE SYSTEMIC hypertension has long been recognized as a cause of various changes in the retinal vasculature of the eye, including flame and blot hemorrhages, microaneurysms, cotton-wool spots, hard exudates, arteriolar narrowing, light reflex changes, and arteriovenous nicking (nipping).1- 6 The importance of these signs as a prognostic indicator of damage to other organs, such as the kidney, heart, or brain, was initially provided by the grading system introduced by Keith et al7 in 1939. However, other reports have shown that these retinal signs may also be observed in older people without hypertension and may be related to arteriosclerosis,8- 10 carotid disease,11,12 and other less frequent causes such as anemia, vasculitis, irradiation,13 or human immunodeficiency virus infection.14 Extravascular retinopathy lesions of hemorrhages and microaneurysms are best known as the typical signs of diabetes, with the severity of retinopathy related to diabetes duration and the level of glycemic control.15 In the Rotterdam Study, retinopathy was not only associated with blood glucose level in people with diabetes, but a significant association was also found in people without diabetes over the normal range of blood glucose levels.16
Most past studies suggested that retinopathy consisting of hemorrhages, microaneurysms, or both is relatively infrequent in people without diabetes. For example, retinopathy was found in only 0.8% of subjects without diabetes in the Framingham Eye Study,17 in 0.4% of men in Göteborg, Sweden,2 and in 0.8% of white men and 2.3% of white women in Evans County, Georgia.12 In contrast, recent reports from the Beaver Dam Eye Study, Beaver Dam, Wis,18- 20 indicated a much higher prevalence of retinopathy (7.8%) in older adults without diabetes. This discrepancy is likely due to differences in ascertainment because of increased sensitivity of photographic grading. In the Framingham, Evans County, and Göteborgstudies, direct ophthalmoscopy was the primary examination used, while masked grading of stereoscopic fundus photographs by trained graders was used in the Beaver Eye Dam Study. Several clinical studies suggested that the extent of retinal vascular changes, including retinopathy, may reflect the severity and duration of hypertension.5,21- 24 Population-based studies (in Rotterdam,16 Göteborg,2 and Evans County12) have found that a higher frequency of retinopathy is associated with a higher blood pressure and a history of hypertension. The Beaver Dam Eye Study18,20 has provided the most comprehensive assessment of the relationship between hypertension and retinopathy lesions in people without diabetes or with known retinal vascular diseases, such as retinal vein occlusion.
As the Blue Mountains Eye Study incorporated detailed grading of retinal photographs similar to the procedure used in the Beaver Dam Eye Study, an opportunity existed to similarly assess the relationship between hypertension and retinopathy in people without diabetes. This study (1) describes the age- and sex-specific prevalence of retinopathy (hemorrhages, microaneurysms, or both) in people without a history or biochemical evidence of diabetes or signs of known retinal vascular disease, (2) assesses the relationship of hypertension and elevated blood pressure to these individual retinopathy lesions and explores any interaction with fasting blood glucose level, and (3) compares the findings with those from the Beaver Dam Eye Study18,20 and other studies.
The Blue Mountains Eye Study was a population-based survey of common eye diseases in 2 urban postal codes in the Blue Mountains, west of Sydney, Australia,25,26 conducted from 1992 to 1994. The study was approved by the Western Sydney Human Ethics Committee and informed consent was obtained from all participants. After a door-to-door census, permanent residents aged 49 years or older were invited to attend a detailed examination at a local hospital clinic. Of 4433 eligible subjects, 3654 (82.4%) participated. Trained interviewers completed a comprehensive questionnaire covering systemic and eye disease history and medication use. Questions defined history of physician-diagnosed hypertension and use of antihypertensive medications. Blood pressure was measured after participants had been comfortably seated for at least 5 minutes, while being questioned. A single measure of systolic and diastolic blood pressure using a mercury sphygmomanometer was recorded from the first and fifth Korotkoff sounds. All subjects were asked to return for fasting blood glucose tests, and 89% complied. Venous plasma glucose level, hemoglobin level, and other test variables were determined from this blood sample.
All participants underwent a detailed ocular examination, which included dilated 30° stereoretinal photographs (Zeiss FF3, Carl Zeiss, Oberkochen, Germany) of diabetic retinopathy study fields 1 (disc) and 2 (macula), nonstereophotographs of fields 3 (temporal) and 4 and 5 (upper and lower vascular arcades), as well as a field nasal to the optic disc.27 Fields 4 and 5 were modified to place the vascular arcade slightly closer to the center of each photograph. The photographs were read in a masked manner by 2 trained graders, and the presence of any retinal hemorrhages or microaneurysms was recorded. Red dots were graded as either hemorrhages or microaneurysms, depending on their size and shape. Sharp round red dots with diameters less than 125 µm were defined as microaneurysms, while others were graded as hemorrhages. Blot- and flame-shaped hemorrhages were graded together, and all questionable lesions were adjudicated. Retinopathy lesions were graded as (1) absent or questionably present and (2) definitely present. In grading hemorrhages or microaneurysms, cases of retinal vein occlusion were excluded. These cases were characterized by the presence of widespread retinopathy lesions (central vein occlusion) or a vascular sectorial distribution (branch vein occlusion), associated with retinal venous dilation in early cases or with venous narrowing or sheathing in late cases.28 No retinopathy cases were identified as resulting from anemia, vasculitis, irradiation, or human immunodeficiency virus infection, assessed from hemoglobin level and medical history.
A subsample of retinal photographs for 202 right and left eyes was regraded by both graders in a masked manner. The sample was randomly selected from cases graded with questionable or definite retinopathy lesions and was assessed as having retinopathy lesions in 46% and 48% of eyes by the 2 graders, respectively. Exact agreement between graders for grading hemorrhages and microaneurysms separately was found for 94% of the cases (κ, 0.90; SE, 0.05). Intragrader reliability was assessed by regrading 104 eyes with exact agreement found for 88% of them (κ, 0.82; SE, 0.07). κ Scores were calculated using the Fleiss method29 and were in the strong agreement category.30
An elevated blood pressure was defined as a systolic blood pressure of 160 mm Hg or higher or a diastolic blood pressure of 95 mm Hg or higher. Subjects with hypertension were defined as having a history of hypertension, using antihypertensive medications, or having an elevated blood pressure. Others were defined as normotensive. Subjects with hypertension were further categorized as (1) using antihypertensive medications and still having an elevated blood pressure (uncontrolled hypertension), (2) using antihypertensive medications without an elevated blood pressure (controlled hypertension), or (3) having an elevated blood pressure and not using antihypertensive medications. A software program (SAS, SAS Institute, Cary, NC) was used for statistical analyses, including χ2 tests, age-adjusted relative risks (RRs), age-glucose–adjusted odds ratios, and 95% confidence intervals (CIs). Trends in age-adjusted proportions were tested for significance by the Cochran-Mantel-Haenszel statistic. Age-adjusted prevalence rates were calculated by the direct method,29 using all nondiabetic Blue Mountains Eye Study participants as the standard population. Statistical differences between age-adjusted rates were tested using a modified Cochran test.31
Of 3654 participants, 217 subjects with a history of diabetes and 43 with a diagnosis of diabetes from an elevated fasting venous plasma glucose level of 7.8 mmol/L (140 mg/dL) were excluded (7.1% of the population). Of the other 3394 subjects, 66 subjects with ungradeable photographs in both eyes and 53 with retinal vein occlusion (prevalence, 1.6%; CI, 1.2%-2.0%) were also excluded, leaving 3275 subjects. Of these subjects, 1828 (55.8%) were normotensive and 1447 (44.2%) were hypertensive, including 993 subjects (30.3%) with a history of hypertension who used antihypertensive medications and 454 subjects (13.9%) with newly diagnosed hypertension who had an elevated blood pressure when examined.
The prevalence of hypertension was significantly related to increasing age in women (χ2 for trend, 115.2; df=1; P=.001) and men (χ2 for trend, 31.8; df=1; P=.001), as shown in Figure 1. Prior to the age of 60 years, hypertension was more frequent among men; however, after this age, it was more frequent among women. After adjusting for age, the prevalence of hypertension in women was slightly higher than in men (46.0% vs 41.7%, P=.01; RR, 1.11; CI, 1.03-1.20).
In persons without diabetes or signs of retinal vein occlusion, retinal hemorrhages were found in 152 subjects (4.6%; CI, 3.9%-5.4%) and retinal microaneurysms were present in 208 subjects (6.4%; CI, 5.5%-7.2%). Either lesion was found in 325 subjects (9.9%; CI, 8.9%-10.9%). Among women, the prevalence of both retinopathy lesions significantly increased with increasing age. In men, retinal hemorrhage prevalence was significantly age related, while microaneurysm prevalence declined slightly after the age of 70 years (Table 1).
After adjusting for age, the prevalence of retinal microaneurysms was significantly higher in men than women (7.5% vs 5.5%, P=.02; RR, 1.36; CI, 1.04-1.77), while the prevalence of retinal hemorrhages was higher in women than men (5.0% vs 4.2%, P=.29; RR, 1.19; 95% CI, 0.87-1.63). The age-adjusted prevalence of either retinopathy lesion was similar in women compared with men (9.6% vs 10.3%, P=.53; RR, 0.94; CI, 0.76-1.15). Figure 2 shows examples of retinopathy lesions, including microaneurysms (A), flame hemorrhages (B), and blot hemorrhages (C and D).
Prevalence rates for retinopathy lesions in women and men with and without hypertension are shown in Table 2 for each age group. The age-adjusted prevalence of both lesions was significantly higher in hypertensive compared with normotensive subjects, for both sexes. Compared with normotensive subjects, an increased age-adjusted RR for either retinopathy lesion was found for women (RR, 1.67; CI, 1.26-2.21) and men (RR, 1.47; CI, 1.07-2.00) with hypertension. For subjects using antihypertensive medications, the prevalence of both retinopathy lesions was higher in people with uncontrolled compared with controlled blood pressure. In subjects with an elevated blood pressure who were not taking medication, the prevalence of retinal hemorrhages was higher than in normotensive subjects (Table 3). Attributable risk calculations indicated that among hypertensive subjects, 42% of retinopathy lesions may be explained by hypertension.
Significant associations were found between increasing quartiles of systolic blood pressure and the age-adjusted prevalence of both retinopathy lesions (Figure 3 and Table 4). This association remained after adjusting for age and fasting blood glucose level in a logistic regression. Similar associations at a lower magnitude were also found for diastolic blood pressure (data not shown). Analyses were initially performed after excluding subjects with known diabetes or a fasting plasma glucose level of greater than 7.8 mmol/L (>140 mg/dL). Reducing the fasting plasma glucose cut point for undiagnosed diabetes to either 6.7 or 5.6 mmol/L (120 or 100 mg/dL) did not appreciably change the retinopathy association with systolic blood pressure quartiles in either women (Figure 4) or men (Figure 5).
No consistent trends for increased prevalence of retinopathy lesions with increasing duration of hypertension were found (data not shown). Neither the direction nor magnitude of the associations between hypertension and the prevalence of microaneurysms or hemorrhages changed if subjects with retinal vein occlusion were included.
Most previous studies estimating the prevalence of retinopathy lesions in subjects with hypertension have been performed in highly selected patient groups, including many people with severe hypertension, and frequently without considering the confounding effect of diabetes. The Beaver Dam Eye Study18,20 and the Blue Mountains Eye Study provide precise age- and sex-specific prevalence estimates of retinopathy (hemorrhages, microaneurysms, or both) in representative, older population samples without diabetes using similar ascertainment (grading of stereoscopic fundus photographs). Although the public health significance of these retinopathy lesions is not yet known, provision of these baseline data is an important first step. Limited population-based data on retinopathy prevalence in people without diabetes using ophthalmoscopy were previously provided by the Framingham Eye Study.17
Two other population-based studies using ophthalmoscopy, in Evans County12 and Göteborg,2 included subjects with diabetes, so they cannot be compared directly. Our population shares many characteristics with the population studied in the Beaver Dam Eye Study. Both included older, largely northern European white subjects. Individuals with a history or biochemical evidence of diabetes were also excluded in the Beaver Dam Eye Study retinopathy reports,18,20 as were cases of retinal vein occlusion. Similar methods of grading the presence of retinopathy lesions were used, making these 2 studies highly comparable. Quite similar retinopathy prevalence rates were found in the 2 studies (9.9% in our study and 7.8% in the Beaver Dam Eye Study). This small difference could have resulted from the higher mean age of participants in our study. It seems unlikely that misclassification of retinopathy status was responsible as there was a high level of interobserver and intraobserver agreement and the grading was masked.
In our study, retinopathy was more than twice as prevalent as in the Rotterdam Study report16 (which included persons with diabetes) in women (9.7% vs 4.5%) and men (10.3% vs 4.6%). This discrepancy might be explained by methodological differences. In the Rotterdam Study, each eye had 2 nonstereoscopic 35° Topcon photographs taken, centered on the macula, compared with the 6 Zeiss 30° fields taken in our study, including 2 stereoscopic fields. However, in a study of diabetic retinopathy, the sensitivity of grading only 2 photographic fields (disc and macula) compared with 7 standard fields was 87%,27 which suggests that this photographic difference may only explain part of the discrepancy observed. An age-related increase in the prevalence of retinal hemorrhages, but not microaneurysms, was found in our study for men and women. Although microaneurysm prevalence was slightly lower in older age groups of men, as also found in the Beaver Dam Eye Study,20 it is quite possible that increasing media opacities could increase the difficulty of grading subtle lesions, such as isolated microaneurysms. After controlling for age, men had a higher prevalence of retinal microaneurysms alone than women in both studies. This difference was statistically significant in the Beaver Dam Eye Study (6.0% in men and 4.4% in women; RR, 1.36; CI, 1.05-1.76)20 and was similar in our study (6.1% in men and 4.7% in women; RR, 1.32; 95% CI, 0.99-1.76).
As in the Beaver Dam Eye Study, we observed a markedly positive relationship between retinopathy and hypertension, which was strongest in the younger age groups. The association persisted after adjusting for age and fasting blood glucose level. In contrast to the Rotterdam Study,16 we found no association between retinopathy and fasting blood glucose level in this nondiabetic population sample. Although a fasting plasma glucose level of 7.8 mmol/L (140 mg/dL) has been shown to have a poor sensitivity for detecting undiagnosed diabetes compared with formal glucose tolerance testing,32 the relationship between blood pressure and retinopathy was affected only marginally by lowering the cut point for diabetes diagnosis to either 6.7 or 5.6 mmol/L (120 or 100 mg/dL). This suggests that our retinopathy findings cannot be explained by the inclusion of some people with undiagnosed diabetes.
The increased age-adjusted prevalence of retinopathy associated with increasing quartiles of systolic and diastolic blood pressure in our study was of a similar magnitude to that observed in the Beaver Dam Eye Study20 (Table 4). The classification of elevated blood pressure in our study was based on only a single blood pressure measurement. If misclassification had occurred, it is likely that the positive association found between hypertension and retinopathy would be weaker.
In both studies, no consistent trends were found indicating increased prevalence of retinopathy with increasing duration of hypertension,18 which could be due to a survivor cohort effect. Klein et al18 suggested that more severe hypertension of short duration might cause more retinopathy than milder systemic hypertension over a longer period. The failure to demonstrate a difference in the prevalence of hemorrhages between hypertensive and nonhypertensive subjects older than 80 years could also be due to a survivor cohort effect, if hypertensive subjects with retinal hemorrhages had increased mortality. It is possible that a higher frequency of arteriosclerotic vascular disease in the older age groups has contributed to a higher prevalence of retinal vascular changes in hypertensive and normotensive subjects, as suggested by Klein et al.18 However, this potential confounding influence was not evaluated or adjusted for in our study. Finally, in interpreting our study findings, it should be stressed that the analyses are based on cross-sectional data, so that a direct causal relationship cannot be inferred.
The presence of typical retinopathy lesions in people without diabetes was previously thought to be uncommon, although few past studies used detailed grading of high-quality retinal photographs to ascertain retinopathy presence. The Blue Mountains Eye Study has found a surprisingly high overall retinopathy prevalence (9.9%), although with similar age-specific prevalence rates for retinopathy, compared with the Beaver Dam Eye Study. Our study also provides strong support for the Beaver Dam Eye Study finding that retinopathy is significantly more frequent in hypertensive compared with normotensive subjects. Further, in this group without diabetes, higher or less-well controlled blood pressure is associated with a higher frequency and severity of retinopathy, independent of blood glucose level.
Despite the strong relationship found between retinopathy and hypertension, our findings indicate that hypertension explains less than half of the retinopathy cases observed. It seems likely that other arteriosclerotic risk factors may also be associated with retinopathy, as well as undetected carotid arterial disease, in some participants.11,12 Long-term follow-up studies of this cohort will be used to assess the relationship further, and 5-year examinations are under way. In such longitudinal studies, it will be useful to evaluate the predictive value of retinopathy lesions for other eye diseases, as well as the risk of hypertensive complications such as renal disease, vascular events, and death.
Accepted for publication July 15, 1997.
This study was supported by grants from the Ophthalmic Research Institute of Australia, Sydney; and the Australian Department of Health and Human Services, Canberra.
Reprints: Paul Mitchell, MD, FRACO, FRCOphth, Department of Ophthalmology, University of Sydney, Hawkesbury Road, Westmead, NSW 2145, Australia (e-mail: firstname.lastname@example.org).