Age- and sex-adjusted survival curves by diabetic retinopathy status (P<.001) at baseline examination in younger-onset persons.
Age- and sex-adjusted survival curve by visual impairment status (P<.001) at baseline examination in younger-onset persons.
Age- and sex-adjusted survival curves by diabetic retinopathy status (P<.001) at baseline examination in older-onset persons.
Age- and sex-adjusted survival curve by visual impairment status (P<.001) at baseline examination in older-onset persons.
Klein R, Klein BEK, Moss SE, Cruickshanks KJ. Association of Ocular Disease and Mortality in a Diabetic Population. Arch Ophthalmol. 1999;117(11):1487-1495. doi:10.1001/archopht.117.11.1487
Copyright 1999 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1999
To investigate the association of ocular disease with all-cause and cause-specific mortality in a diabetic population.
Geographically defined population-based cohort study.
An 11-county area in Wisconsin.
Participants were all younger-onset diabetic persons (diagnosed as having diabetes at <30 years of age and taking insulin) and a random sample of older-onset diabetic persons (diagnosed as having diabetes at ≥30 years of age). Diabetic retinopathy, macular edema, visual acuity, and cataract were measured using standardized protocols at baseline examinations from 1980 to 1982, in which 996 younger-onset and 1370 older-onset persons participated. Participants were followed up for 16 years.
Main Outcome Measure
All-cause and cause-specific mortality as determined from death certificates.
In the younger-onset group, after controlling for age and sex, retinopathy severity, macular edema, cataract, history of cataract surgery, and history of glaucoma at baseline were associated with all-cause and ischemic heart disease mortality. In the older-onset group, after controlling for age and sex, retinopathy and visual impairment were related to all-cause, ischemic heart disease, and stroke mortality. No ocular variable under study was related to cancer mortality in the older-onset group. After controlling for systemic risk factors, visual impairment was associated with all-cause and ischemic heart disease mortality in the younger-onset group. In the older-onset group, retinopathy severity was related to all-cause and stroke mortality, and visual impairment was related to all-cause, ischemic heart disease, and stroke mortality.
Presence of more severe retinopathy or visual impairment in diabetic patients is a risk indicator for increased risk of ischemic heart disease death. Presence of these ocular conditions may identify individuals who should be under care for cardiovascular disease.
PERSONS WITH diabetes have poorer survival rates than those without, mainly because of cardiovascular disease.1- 19 The relation of diabetes to death from cancer varies.2,3,11,14,15,19,20 Decreased survival rates in persons with diabetes have been associated with cataract, diabetic retinopathy, and visual impairment.21- 29 However, few studies24 have examined the associations between specific eye conditions and cause-specific mortality in people with diabetes. The purpose of the present investigation was to examine the association of retinopathy and other eye conditions with all-cause and cause-specific mortality in a large population-based study: the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR).
The population has been described in detail in previous reports.30,31 The study area is composed of 11 counties in southern Wisconsin (1980 population of 839,324). Diabetic persons were identified by reviewing the records of 452 of the 457 physicians providing primary care to diabetic persons from July 1, 1979, through June 30, 1980. Two methods of identifying patients were used: daily lists with names entered by the physician or his or her staff and computer retrieval of records of patients with a coded diagnosis of diabetes. Chart reviews were completed for 9841 of the 10,135 patients so identified. Of these persons, 338 were confined to nursing homes, 157 had died before July 1, 1979, 45 did not have diabetes (incorrect computer coding), and 18 had moved before July 1, 1979, or had gestational diabetes. These 558 patients were excluded from further analysis.
A 2-part sample of 2990 of the remaining 9283 patients was selected on July 1, 1980, for the examination phase of the study. The first part consisted of all persons diagnosed as having diabetes before 30 years of age and taking insulin, referred to as the younger-onset group (n=1210). The second part consisted of a probability sample stratified on duration of diabetes of persons diagnosed by a physician as having diabetes, at or after the age of 30 years and confirmed by a random or postprandial serum glucose level of at least 11.1 mmol/L (200 mg/dL), or a fasting serum glucose of at least 7.8 mmol/L (140 mg/dL) on at least 2 occasions, referred to as the older-onset group (n=1780). The specifics of the sampling strategy and exact proportions appear in an earlier publication32 but were approximately 25% for eligible persons with 0 to 4 or 5 to 14 years' duration, and 100% for persons with 15 or more years of diabetes. Otherwise, to ensure the representativeness of the samples, selection was made without regard to any patient characteristic. Of the older-onset group, 824 were taking insulin and 956 were not.
Subjects were invited to participate in the examination phase of the study, which was conducted in a mobile examining van from August 21, 1980, to July 30, 1982. Pertinent parts of the examination included obtaining informed consent, measuring blood pressure according to the Hypertension Detection and Follow-up Program Protocol,33 performing a refraction and measurement of visual acuity using a modified Early Treatment Diabetic Retinopathy Study protocol,34 administering a medical history questionnaire, performing a slitlamp examination of the anterior segment and lens, taking stereoscopic color fundus photographs of 7 standard fields,34 determining urine protein level, and determining glycosylated hemoglobin value from a finger-stick capillary blood sample.35 Height and weight were also measured.
Baseline examinations were obtained for 996 younger-onset persons (82.3%) and 1370 older-onset persons (77.0%). Reasons for nonparticipation in the examination and comparisons between participants and nonparticipants have been presented elsewhere.31,32 Among younger-onset persons, nonparticipants had a significantly shorter duration of diabetes than participants, whereas current age, age at diagnosis, sex, and blood pressure were similar in both groups.31 In older-onset persons, nonparticipants were older, were older at diagnosis, had longer duration of diabetes, and were more likely to be female than participants, whereas blood pressure was similar in both groups.32
To examine the association of the development of visual loss and progression of retinopathy to survival, we also used data collected at the 4-year follow-up of the cohort. Survivors from the baseline examination were invited to participate in a 4-year follow-up examination from 1984 to 1986.36,37 Of the 2366 persons examined from 1980 to 1982 in the baseline study, 1878 (79.4%) participated in the follow-up. Of the original cohort, 404 (17.1%) died before the follow-up examination, 9 could not be located, 28 permitted an interview only, and 47 refused to participate. Reasons for nonparticipation and comparison between participants and nonparticipants are presented elsewhere.36,37 The follow-up examination followed similar protocols as the baseline examination.
An original aim of the WESDR was to examine mortality in the study population. Thus, all sampled persons are contacted annually by telephone to determine vital status. In addition, designated contact persons, relatives, and physicians are contacted, and newspaper obituaries are reviewed daily. In all cases, an attempt is made to obtain an exact or approximate date of death. Annually, a request is made to the Wisconsin Center for Health Statistics, Section of Vital Statistics, for death certificate information, including cause of death of these persons. In addition, persons who are not known to be dead but have been lost to follow-up are submitted for matching against death records. Wisconsin death records through December 1996 have been searched. Finally, information on persons who have moved out of Wisconsin and are suspected of being deceased and persons who are lost to follow-up is submitted to the National Death Index for matching against national death data. When a match is made, a copy of the death certificate is obtained from the appropriate state.
All medical conditions on the Wisconsin death certificates were coded by trained nosologists in the Wisconsin Division of Health using the International Classification of Diseases, Ninth Revision.38 Out-of-state certificates were coded and processed in the same manner. Underlying cause of death was selected by the Automated Classification of Medical Entities computer program.39,40 Cause-specific mortality analysis of the present investigation was based on both the underlying cause of death and any mention of the cause on the death certificate. Any mention of a cause could be the immediate cause of death, underlying cause, secondary cause, or any other significant condition affecting the chain of events leading to the death listed on the death certificate.
The survival interval began on the date of the initial examination. Follow-up has been 16 years, with a median of 16 and 8.5 years in younger- and older-onset persons, respectively. In addition, only 14 younger-onset and no older-onset persons have been lost to follow-up. Only deaths that have been confirmed by a death certificate were included in the analysis. If a death certificate was not located for a suspected death, the person was considered to be alive as of the last contact date.
To determine the severity of retinopathy, all fundus photographs were graded using a modification of the Early Treatment Diabetic Retinopathy Study adaptation of the modified Airlie House classification of diabetic retinopathy.41,42 This scheme specifies 13 levels of retinopathy. For tabular analyses of baseline data, the level in the worse eye was used and was grouped into 4 categories of retinopathy: none (level 10), mild nonproliferative (levels 21, 31, and 37), moderate nonproliferative (levels 43, 47, and 53), and proliferative (levels 60, 61, 65, 71, 75, and 85). Progression to proliferative retinopathy was estimated from all persons who were free of this complication at baseline examination but developed it at the 4-year follow-up examination. To determine the progression of retinopathy during the 4-year period, retinopathy level for a participant was derived by concatenating the levels for the 2 eyes and giving the eye with the higher level greater weight. This scheme provided a 15-step scale (10/10, 21/<21, 21/21, 31/<31, 31/31, 37/<37, 37/37, 43/<43, 43/43, 47/<47, 47/47, 53/<53, 53/53, 60+/<60+, and 60+/60+) when all levels of proliferative retinopathy were grouped as 1 level. For persons with no (level 10/10) or only nonproliferative retinopathy (levels 20/10 through 53/53), progression was defined as an increase in the severity of retinopathy by 2 steps or more from baseline level at the 4-year follow-up examination.
Macular edema was defined as thickening of the retina, with or without partial loss of transparency within 1 disc diameter from the center of the macula,43 or the presence of focal photocoagulation scars in the macular area associated with a history of development of macular edema, as documented by stereoscopic fundus photographs. If macular edema could not be graded in an eye, the individual was assigned the score of the other eye. Incidence of macular edema was estimated from data for all persons who had no macular edema, had not been treated previously with photocoagulation at baseline examination, and participated at the 4-year follow-up.
For each eye, the best corrected visual acuity was recorded as the number of letters read correctly from 0 (≤20/250) to 70 (20/10).44 For eyes with visual acuity worse than 20/250, 1 of 6 levels of visual acuity was recorded: 20/320, 20/400, 20/800, hand motions, light perception, and no light perception. Levels of impairment in visual function were defined as the best-corrected visual acuity for a participant based on the better eye, as follows: none (>20/40), mild (20/40 to 20/63), moderate (20/80 to 20/160), and severe (≤20/200). A doubling of the visual angle was defined as a loss of 15 letters (ie, a change from 55 to 40 letters corresponded to a visual acuity change from 20/20 to 20/40) at the 4-year follow-up.
Nuclear sclerosis was considered present at baseline if an increase in optical density of the lens nucleus, with or without a color change, was found on slitlamp biomicroscopy through a dilated pupil. Severity of the lens change was judged against a standard slitlamp camera photograph. Posterior subcapsular cataract was defined as a granular-appearing opacity of the posterior subcapsular region found on slitlamp examination. Severity was judged against a standard red reflex photograph. For the analysis reported herein, cataract was defined as the definite presence of either type of cataract.
The presence of glaucoma at baseline was defined by a history of glaucoma, a history of taking antiglaucoma medications, or both. Rubeotic glaucoma was defined by the presence of neovascularization of the iris, as determined by slitlamp examination (without gonioscopy) and a history of glaucoma, a history of taking antiglaucoma medications, or both.
Current age was defined as age at the time of baseline examination. Mean systolic blood pressure at baseline was the average of 2 systolic blood pressure measurements, and mean diastolic blood pressure was the average of 2 diastolic blood pressure measurements. Hypertension was defined as systolic blood pressure of 160 mm Hg (140 mm Hg if younger than 25 years) or greater, diastolic blood pressure of 95 mm Hg (90 mm Hg if younger than 25 years) or greater, or a history of hypertension with use of antihypertensive medications. A person was defined as having a positive history if he or she responded positively to the questions regarding cardiovascular disease, angina, myocardial infarction, and stroke. Primary care physicians were consulted whenever the participant was unsure of the diagnosis. Body mass index was defined as body weight (in kilograms) divided by the square of the height (in meters). Urine protein was measured by dipstick (Labstix; Ames, Elkhart, Ind) and was considered to be present if greater than or equal to 0.30 g/L. Cigarette smoking status was determined as follows. Subjects were classified as never having smoked if they had smoked fewer than 100 cigarettes in their lifetime, as being ex-smokers if they had smoked more than this number of cigarettes but had stopped smoking before the baseline examination, or as currently smoking if they had not stopped. Pack-years smoked was defined as the number of packs (20 cigarettes) smoked per day times the number of years smoked.
Multivariate analyses were performed by Cox proportional hazards regression.45 This permits the evaluation of the effect of specific ocular diseases and visual impairment on survival while controlling for other risk factors. Age and sex were included in every model, because these are generally regarded as important factors for mortality. Additional risk factors were selected in stepwise fashion, remaining in the final models if significant at the .05 level.46 Finally, each specific ocular condition was added to separate models to determine its independent effect on mortality. Hazard ratios for mortality were computed as exp(β), in which β is the coefficient of a variable in the model. The 95% confidence intervals for the hazard ratio were computed as exp(β±1.96 × SEβ), in which SEβ is the SE of the coefficient. Because of concerns that older people may not have been included in the original cohorts or were more likely to be lost to follow-up—resulting in a healthy cohort effect—interactions between age and the ocular variables were investigated. Age as a continuous variable was used in the computation of interaction terms.
Baseline characteristics of the population are presented in Table 1. Males comprised 51.4% and 46.4% of the younger- and older-onset groups, respectively. Mean (±SD) age at diagnosis was 14.6 (±7.6) and 54.8 (±12.4) years in the younger- and older-onset groups, respectively. Mean (±SD) age at examination was 29.3 (±13.3) and 66.6 (±11.3) years in the younger- and older-onset groups, respectively. Prevalence and severity of retinopathy were higher in the younger-than in the older-onset group, whereas visual impairment, cataract, cataract extraction, and glaucoma were more frequent in the older- compared with the younger-onset group.
There have been 214 (21.5%) and 996 (72.7%) confirmed deaths in the younger- and older-onset groups, respectively. There were an additional 17 deaths in the younger-onset group and 39 deaths in the older-onset group that could not be confirmed by death certificates. Table 2 presents all-cause mortality and ischemic heart disease, stroke, and cancer mortality listed as underlying and any mentioned cause of death in the younger- and older-onset groups. There were too few deaths due to stroke or cancer in younger-onset persons to examine associations with ocular conditions. For further analyses of specific causes of death, "any mention" of cause of death was used.
Hazard ratios, all-cause mortality, and mortality from a specific cause after adjusting for age and sex only or age, sex, and other factors listed in Table 3 are presented in Table 4 and Table 5. Most of the specific ocular conditions were statistically significantly associated with all-cause mortality (Table 4, Figure 1, and Figure 2) and ischemic heart disease mortality in the younger-onset group while controlling for age and sex. Hazard ratios varied from 1.58 (presence of cataract at baseline) to 11.02 (presence of proliferative retinopathy at baseline). Including other risk factors associated with all-cause and ischemic heart disease mortality in the models resulted in diminished magnitudes of the associations in the younger-onset group; only the associations of visual impairment at baseline with all-cause and ischemic heart disease mortality were statistically significant (Table 4). Adding blood pressure or gross proteinuria to the models had the greatest impact on reducing the hazard ratios in most models (data not shown).
In the older-onset group, the associations between ocular conditions and all-cause (Table 5, Figure 3, and Figure 4) and cause-specific mortality were mostly positive but weaker than those found for the younger-onset group (Table 4). After controlling for other factors, including age and sex, retinopathy severity at baseline was associated with all-cause mortality, and proliferative retinopathy at baseline was associated with stroke mortality (Table 5). Older-onset persons with visual impairment at baseline were more likely to have ischemic heart disease or stroke mentioned on the death certificate than those without visual impairment. Cataract, cataract extraction, or glaucoma at baseline were not associated with mortality. No ocular condition was associated with cancer in the older-onset group.
We examined the data for evidence of interaction of age with the ocular variables with respect to overall mortality. The only significant interaction was between age and visual impairment in the older-onset cohort; the effect of visual impairment was found to weaken with older age (data not shown).
We also examined whether progression of retinopathy, progression to proliferative retinopathy, incidence of macular edema, or doubling of the visual angle from baseline to the 4-year follow-up was associated with an increased risk of mortality. In the younger- and older-onset groups, respectively, 4-year rates of progression of retinopathy were 44.2% and 31.6%; for the 4-year incidence of proliferative retinopathy, the rates were 10.4% and 4.6%; for macular edema, the rates were 9.0% and 5.3%; and for the 4-year doubling of the visual angle, the rates were 5.9% and 9.6%. Of 148 younger- and 635 older-onset persons who died after the 4-year follow-up, 65 younger- and 276 older-onset persons had ischemic heart disease listed as underlying and any mentioned cause of death. In addition, in the older-onset group, stroke was listed as underlying and any mentioned cause of death in 111 persons and cancer in 56 persons. In the younger-onset group, after controlling for age, sex, and other factors, the association of doubling of the visual angle with ischemic heart disease death remained statistically significant (Table 6). In the older-onset group, progression of retinopathy to proliferative disease was associated with all-cause mortality, and progression to proliferative retinopathy was associated with ischemic heart disease mortality (Table 6). Older-onset persons who had a doubling of the visual angle during the first 4 years of the study were more likely to have either ischemic heart disease or stroke mentioned on the death certificate than those without these conditions.
The WESDR offered a unique opportunity to examine the relation of specific ocular conditions to survival. Most information about these relations has been derived from studies of select groups of patients. The WESDR is unique in that it consists of a large cohort of people with both type 1 and type 2 diabetes mellitus who were followed up for 16 years and measured eye disease according to defined protocols.31,32,47 In addition, the large number of confirmed deaths—214 in the younger-onset group and 996 in the older-onset group—provided sufficient power to determine associations with most of the ocular conditions under study.
In the WESDR, while controlling for only age and sex, retinopathy was strongly associated with all-cause and ischemic heart disease mortality in both the younger- and older-onset groups. These associations were expected, since hyperglycemia,47- 51 hypertension,47,52 and dyslipidemia53- 55 are related to retinopathy development and cardiovascular disease complications in people with diabetes. For ophthalmologists and optometrists, probably the more important analysis is after adjusting for only age and sex; the presence of diabetic retinopathy or visual impairment indicates a greatly increased risk of death in the next 16 years.
The findings of the WESDR are consistent with other reports. In a prospective study of 709 patients taking insulin who were diagnosed as having diabetes before the age of 50 years and followed up for up to 13 years, Davis et al22 reported a 5-year survival rate of 56% for those with proliferative retinopathy compared with 99% for those with no or minimal retinopathy at baseline. This relation remained after controlling for duration of diabetes and sex. Similarly, based on a retrospective review of 128 diabetic patients followed up at the Radcliffe Infirmary, Oxford, England, Caird et al21 reported a 5-year survival rate of 55% for those with proliferative retinopathy compared with 92% for those with only microaneurysms and 92% for those without any retinopathy. Systemic factors were not controlled for in these studies, and the association of retinopathy with specific causes of death was not examined. Rates in both studies are similar to 5-year survival rates in the WESDR in younger-onset persons (97% in those with no or minimal nonproliferative retinopathy, and 76% in those with proliferative retinopathy) and older-onset persons (72% in those with no or minimal retinopathy and 52% in those with proliferative diabetic retinopathy) (R.K., unpublished data, March 2, 1999). In a population-based cohort of 249 persons with known type 2 diabetes mellitus in Oxford, England, after controlling for age, sex, and systemic factors, retinopathy severity was associated with increased 6-year all-cause mortality with a relative risk of 3.4.28 After correcting for age, duration of diabetes, and sex in a cohort27 of 353 Mexican Americans with type 2 diabetes mellitus followed up for a mean of 8 years, retinopathy severity was associated with mortality. This association was weakened while adjusting for cholesterol status in women. Together, these data suggest that severe retinopathy is an important risk indicator for cardiovascular disease mortality. However, it is not known whether careful monitoring and earlier detection and treatment of cardiovascular disease and its risk factors in diabetic persons with severe retinopathy would result in decreased cardiovascular mortality.
After correcting for age, sex, and other factors, data from the WESDR showed an association between proliferative retinopathy and stroke mortality in the older-onset group. This is consistent with our earlier finding of an age-adjusted relative risk of proliferative retinopathy of 2.9 for incident stroke (95% confidence interval, 1.2-6.8) in older-onset persons taking insulin, and a relative risk of 6.0 (95% confidence interval, 1.1-32.6) for older-onset persons not taking insulin.56 The same risk factors for retinopathy, glycemia50 (R.K., unpublished data, July 24, 1998), and blood pressure52 are associated with an increased risk of stroke in people with diabetes. Thus, the presence of retinopathy may be an indicator of small cerebral blood vessel disease. The presence of severe retinopathy in persons with type 2 diabetes mellitus should alert the physician to the need for assessment and possible treatment of risk factors to prevent the development or recurrence of stroke.
After controlling for age, sex, and other factors, neither glaucoma nor cataract was related to all-cause, ischemic heart disease, or stroke mortality in the WESDR. This is consistent with data from the Beaver Dam Eye Study,29 in which cataract was not associated with all-cause mortality in people with type 2 diabetes mellitus. On the other hand, in the Framingham Eye Study, after adjusting for systemic risk factors, cataract, regardless of type, was associated with decreased survival rates only in people with diabetes.24 These investigators concluded that lens changes in persons with type 2 diabetes mellitus reflected general health status. The reasons for the differences among studies are not apparent.
In the WESDR, visual impairment was consistently associated with all-cause and cardiovascular disease mortality in both groups. This is consistent with earlier reports by Davis et al22 in persons with type 1 diabetes mellitus. Visual impairment in persons with diabetes is associated with proliferative retinopathy and macular edema,57 complications associated with poor glycemic control, hypertension, dyslipidemia, and renal failure.48,54,56
Mortality from most cancers has not been found to be increased in diabetic people,1,3,11,14,19,20 with one exception.15 Thus, one would not expect ocular conditions to be a risk indicator of cancer death among diabetic persons. Indeed, we found this to be true in the older-onset group. The number of cancer deaths was too small in the younger-onset group to evaluate.
The present study may have some limitations. Incomplete enumeration of all cases of diabetes in the study area could introduce selection biases in the results. Indeed, the exclusion of institutionalized persons, which includes nursing home residents, would reduce our mortality estimates, since these people are predominantly older. However, this need not affect the relation of ocular conditions with mortality. Indeed, aside from an age–visual impairment interaction in the older-onset group, we did not find other significant age–ocular condition interactions with respect to overall mortality. Misclassification of cause of death might also result in reducing the strength of the associations found.
In summary, in the WESDR, after controlling for only age and sex, strong and statistically significant associations were found between most of the ocular conditions studied and all-cause and ischemic heart disease or stroke mortality. Even though controlling for other factors diminished many of these associations, these findings show that these ocular conditions, especially the presence of severe retinopathy or visual impairment, serve as risk indicators for death from vascular diseases in people with diabetes. Because vascular disease is involved in most deaths in people with diabetes, there should be a public health benefit accrued from identifying such individuals and monitoring them for heart disease.
Accepted for publication June 9, 1999.
This research was supported by grants EY03083, EY12198, and HL59259 (Drs R. Klein and B. E. K. Klein) of the National Institutes of Health, Bethesda, Md, and in part by a Senior Scientific Investigator Award of Research to Prevent Blindness Inc, New York, NY (Dr R. Klein).
We thank the 452 Wisconsin physicians and their staff members who participated in and supported this study; our collaborators, Matthew D. Davis, MD, and Mari Palta, PhD; Terry Spennetta and Earl Schrago, MD, who provided laboratory support; and Wisconsin Center for Health Statistics, Madison, for assistance in obtaining death certificate information.
Reprints: Ronald Klein, MD, MPH, Department of Ophthalmology and Visual Sciences, University of Wisconsin-Madison, 610 N Walnut St, 460 WARF, Madison, WI 53705-2397 (e-mail: email@example.com).