Abstract
Objective: The purpose of this study was to assess the relation of retinal microvascular signs and incident stroke and stroke mortality in an older population.
Methods: The authors took retinal photographs on baseline participants (3,654 patients aged 49+ years) of the Blue Mountains Eye Study (1992 to 1994). They assessed the presence of retinopathy (microaneurysms, retinal hemorrhages) in participants without diabetes and retinal arteriolar signs in all participants using standardized grading protocols. Incident stroke/TIA/cerebrovascular death (combined stroke events) were identified at follow-up examinations during 1997 to 1999.
Results: During a 7-year period, 859 participants died, 97 (11.3%) of which died of cerebrovascular causes. Of survivors, 24 had confirmed incident stroke, and 11 had incident TIA. Combined stroke events were more frequent in participants with retinopathy (5.7%), with moderate/severe arteriovenous nicking (4.2%), or with focal arteriolar narrowing (7.2%) compared with those without (1.9%). After controlling for age, sex, systolic blood pressure, smoking, and self-rated health, retinopathy was significantly associated with combined stroke events (relative risk [RR] 1.7, 95% CI 1.0 to 2.8) in persons without diabetes. This association was stronger in those without severe hypertension (RR 2.7, CI 1.2 to 6.2) or in persons with two or more retinal microvascular signs (RR 2.7, CI 1.5 to 5.2). Generalized or focal arteriolar narrowing or arteriovenous nicking was not independently associated with combined stroke events after multivariate adjustment.
Conclusions: In older Australians without diabetes, retinopathy signs predict stroke or stroke-related death independent of traditional stroke risk factors.
The retinal arterioles share similar anatomic, physiologic, and embryologic characteristics with cerebral arterioles.1 Therefore, pathologic changes seen in the retinal circulation, such as microaneurysms, retinal hemorrhages, arteriovenous (AV) nicking, and arteriolar narrowing, may be risk markers for concomitant or subsequent cerebrovascular disease.2–4 We sometimes refer to these signs as hypertensive retinopathy.5
Recent population-based studies have provided fresh data linking retinal microvascular signs, as identified from retinal photographs, and risk of stroke. In the Atherosclerosis Risk in Communities (ARIC) Study, persons aged 51 to 72 years with retinopathy (microaneurysms, retinal hemorrhages), AV nicking, and generalized arteriolar narrowing were more likely to have developed incident stroke than persons without these retinal signs, even after adjustment for blood pressure, diabetes, cigarette smoking, and other stroke risk factors.6 This suggests that a retinal assessment, possibly via photography, may provide information for stroke risk prediction7 independent of that provided by traditional risk factors.
However, further studies are needed to confirm these findings in other populations and settings. For example, we do not know whether the ARIC findings are applicable to older people, who generally have higher levels of blood pressure and a higher risk of stroke and may have different stroke risk factors.8,9 Further, we do not know the predictive value of combinations of these retinal microvascular signs because this question has not been investigated in previous studies.7
We aimed to assess the relationship between the presence and combination of retinal microvascular signs and incident stroke events in the Australian Blue Mountains Eye Study.
Methods.
Study population.
The Blue Mountains Eye Study is a population-based cohort study of vision, common eye diseases, and other health outcomes in an urban population aged 49 years or older. Baseline participants (n = 3,654) represented 82.4% of eligible residents living in two postal code areas in the Blue Mountains, Australia.10 This community is representative of the older Australian urban population.11 The study was approved by the Western Sydney Area Human Ethics Committee. Written, informed consent was obtained from all participants, who also had a detailed medical history questionnaire administered by trained interviewers. A comprehensive eye examination was performed, including measurement of visual acuity and refraction together with retinal, optic disk, and lens photography.
Retinal photography.
At the baseline examinations, dilated 30-degree stereoscopic retinal photographs of the macula, optic disk, and other retinal fields of one or both eyes were taken in 98% (n = 3,583) of the study population, using a Zeiss FF3 fundus camera (Carl Zeiss, Oberkochen, Germany), and were graded for macular and retinal lesions.12,13 Assessment of retinal microvascular signs has previously been described.13 In brief, retinopathy lesions (microaneurysms, hemorrhages in subjects without diabetes), focal retinal arteriolar narrowing, and AV nicking were graded from 35-mm transparencies using a light box and a Donaldson viewer. Only arterioles located at least one half-disk diameter away from the optic disk margin were assessed for focal arteriolar narrowing. Standard photographs for retinal vascular changes were selected by a retinal specialist (P.M.) from the standard photographic set developed for the Modified Airlie House Classification of Diabetic Retinopathy14 (Standards 2A, 8A, 10A, and 14) and the Wisconsin Age-Related Maculopathy Grading System.15 Focal arteriolar narrowing was identified and graded as absent/questionable (none or less severe than the standard photograph) or present (equal to or more severe than the standard). AV nicking was defined as a decrease in venular width on both sides of the venule where crossed by an arteriole. AV nicking was graded as absent/questionable, mild, or severe.
For assessment of generalized retinal arteriolar narrowing, a computer-assisted grading method was used to measure retinal vessel diameters.16 In brief, digitised retinal images of one eye of each participant were displayed on a computer screen, and all vessels greater than 25 μm in diameter and completely passing through the region 0.5 to 1.0 disk diameters from the optic disk margin were measured using a software program (RetinalAnalysis software, University of Wisconsin-Madison, Madison, WI). Average arteriolar or venular diameter was calculated and summarized as the arteriolovenular ratio, with lower arteriolovenular ratios indicating greater severity of generalized arteriolar narrowing. Reliability of this method has been assessed previously,16 with κ values ranging from 0.80 to 0.93 for intergrader and intragrader reliability.
Incident stroke definition.
During 1997 to 1999, we invited baseline participants to attend follow-up examinations. All 2,334 participants (75% of survivors) seen at the 5-year follow-up examinations were interviewed using standard questionnaires, which included medical history of the past 5 years. Persons who reported having a “stroke” during this period were cross-checked with medical records obtained from local hospitals and local doctors’ offices (participants’ family doctors or specialists). The diagnosis of stroke was based on MONICA criteria, and a majority of this group had either CT or MRI performed. Incident definite stroke was defined if typical clinical symptoms were reported, confirmed by CT or MRI examination. Incident TIA was defined if typical clinical symptoms of a transient nature were reported, with negative CT scan results. We used strict diagnostic criteria to avoid misclassification but could have missed a few cases, particularly those without CT or MRI records.
To identify and confirm persons who died after the baseline examination, demographic information including surname and first and second names, sex, and date of birth of the 3,654 participants were used to cross-match with Australian National Death Index (NDI) data for deaths up to the end of 2001. In the NDI database, cause of death is collected from death certificates and recorded using the International Classification of Diseases (ICD) code system. Codes for cerebrovascular death were either from ICD 9 (4309, 4319, 4321, 4332, 4340, 4341, 4349, 4369, 4370, 4373, 4379, 4389, and 4409) or ICD 10 (I619, I633, I634, I639, I64, I694, and I698). For the end point of cerebrovascular death, persons with a history of stroke before baseline were not excluded. For the end point of incident stroke/TIA or for the combined end points (stroke/TIA/cerebrovascular death), termed combined stroke events, we excluded subjects with self-reported history of stroke at baseline (n = 193), because they were not then considered at risk of stroke. Two incident stroke patients died months or a year or two later and were included in both end points of incident stroke/TIA and cerebrovascular death, when analyses were performed separately for these two end points.
Risk factors.
At the baseline examination, blood pressure was measured once using a single mercury sphygmomanometer with an appropriate cuff size, after participants had been seated for at least 10 minutes. Severe hypertension was defined as systolic blood pressure of 160 mm Hg or greater or diastolic blood pressure of 95 mm Hg or greater at the time of examination, or current use of antihypertensive medications. Diabetes was diagnosed by either medical history or fasting blood glucose of 7.0 mmol/L or greater at the baseline examination. Current smokers were defined as participants who reported that they currently smoked or had stopped smoking less than 1 year before the examination. Self-rated health was assessed using a standard question.17
Statistical methods.
Statistical Analysis System (SAS, SAS Institute, Cary, NC) was used for statistical analysis. We calculated cumulative incidence rates for incident stroke events in subjects with and without specific retinal microvascular signs. Cox proportional hazards models were used to assess the relative risk (RR) of stroke and TIA, or stroke events, in association with retinal signs, adjusting for age, sex, systolic blood pressure, smoking, self-rated health, and diabetes. Survival time was calculated as days survived from the baseline examination to the first stroke event and death for cases with an event, or to December 31, 2001, for persons with no end-point events by that time.
Results.
Of the 3,654 Blue Mountains Eye Study baseline participants, 3,583 (98.1%) had photographs gradable in at least one eye for focal arteriolar narrowing or AV nicking, and 3,355 (91.8%) had photographs of at least one eye gradable for generalized arteriolar narrowing. Of the 3,583 participants, focal retinal arteriolar narrowing was detected in 280 (7.7%), mild AV nicking was detected in 1,415 (38.8%), and moderate or severe AV nicking was detected in 316 (8.7%). Among persons without diabetes and with gradable retinal photographs (n = 3,343), 376 (11.3%) were found to have retinopathy lesions.
As at the end of 2001, there were a total of 859 deaths, of which 97 (11.3%) were due to cerebrovascular causes, and among surviving participants, 24 had confirmed incident stroke and 11 had incident TIA at the follow-up examination. Table 1 shows the baseline characteristics of at risk participants with and without incident combined stroke events during the 5-year period, after excluding 193 participants who reported a history of stroke at baseline. One person had a cerebral hemorrhage from a cerebral aneurysm and so was excluded from analysis of incident combined stroke events. A further 256 persons with diabetes and 71 persons without retinal photographs taken in either eye were excluded (16 were overlapping in these two groups, leaving 55 persons without diabetes who had no retinal photographs taken). In general, persons who had development of incident stroke events were older and, after adjustment for age, had higher systolic and diastolic blood pressure. Nondiabetic persons with retinopathy were older, and, after age adjustment, were more likely to have hypertension, higher systolic and diastolic blood pressures, and higher body mass index. Persons with retinopathy at baseline had slightly higher mean serum fibrinogen levels than persons without these lesions (4.2 vs 4.1 g/L; p = 0.026).
Table 1 Baseline characteristics* of participants at risk of combined stroke events (incident stroke/TIA/cerebrovascular death) or presence of retinopathy†
Table 2 shows that in persons without diabetes, incident combined stroke events were greater in subjects with than without retinopathy. For all persons with and without diabetes, the presence of AV nicking or focal arteriolar narrowing at baseline was associated with a higher incidence of incident combined stroke events. After adjusting for age, retinopathy and focal arteriolar narrowing were significantly associated with an increased risk of incident stroke/TIA. These associations remained significant after further adjustment for other stroke risk factors, including sex, systolic blood pressure, smoking, self-rated health, and diabetes. When the outcome included cerebrovascular death (combined stroke events), the association between retinopathy and incident combined stroke events became weaker but remained significant, whereas the associations between focal arteriolar narrowing and incident combined stroke events became nonsignificant (see table 2). Generalized arteriolar narrowing was not independently associated with combined stroke events after age and multivariate adjustment.
Table 2 Retinal microvascular signs at baseline and 5-year incidence of stroke/TIA, cerebrovascular deaths or combined stroke events
We also examined the association between retinopathy signs and the risk of stroke/TIA, or the risk of combined stroke events in nondiabetic persons with and without hypertension. The associations seemed stronger in nondiabetic persons without severe hypertension (table 3). We were unable to investigate incident stroke and incident TIA separately, because of small numbers.
Table 3 Multivariable-adjusted relative risk* of combined stroke events (incident stroke/TIA/cerebrovascular death) by retinal microvascular signs in persons without diabetes, stratified by severe hypertension
Finally, we found that nondiabetic persons with two or more retinal microvascular signs at baseline were more likely to have development of incident stroke/TIA or combined stroke events than persons with no signs (OR 6.4, 95% CI 2.3, 17.3 for incident stroke/TIA and OR 2.7, 95% CI 1.5, 5.2, for combined stroke events).
Discussion.
Our study shows that older nondiabetic persons with retinopathy (microaneurysms, retinal hemorrhages) have an increased risk of stroke events, even after controlling for traditional risk factors. The magnitude of this risk was present and possibly stronger in those without than with severe hypertension. The presence of focal arteriolar narrowing or moderate to severe AV nicking was also associated with an increased risk of incident stroke, TIA, or stroke-related death, although the associations were nonsignificant after adjusting for other cardiovascular risk factors. The presence of two or more of these three retinal vessel signs was strongly associated with incident stroke events. Generalized arteriolar narrowing was not related to incident stroke or cerebrovascular deaths.
Our findings are compatible with several previous studies linking retinal microvascular signs, as detected by clinical examination, and stroke risk,18–22 and more recent studies that show associations between retinopathy signs, as graded from retinal photographs, and incident stroke,6 prevalent stroke,23 and subclinical MRI cerebral lesions.24
A more detailed examination of our study and the ARIC study6 provides further insights into the association of retinopathy signs and stroke risk. First, in these two studies, microaneurysms and retinal hemorrhages were stronger predictors of stroke events than retinal arteriolar signs (generalized or focal retinal arteriolar narrowing and AV nicking). This may reflect slightly different pathophysiologic processes associated with different retinal microvascular signs. Experimental and histopathologic studies suggest that generalized and focal arteriolar narrowing and AV nicking represent microvascular changes occurring earlier in the course of hypertension. Retinal hemorrhages, however, may indicate more severe microvascular disease, because hemorrhages often occur in the presence of retinal ischemia.1,25 The stronger association of retinopathy lesions and stroke event risk may reflect the link of more recent and severe vascular damage to cerebrovascular disease.26
Second, we found overall weaker associations between these retinal microvascular signs and incident stroke events in our older population (mean age 65.8 years) than were found in the younger ARIC study participants (mean age 53.6 years).6 We note that the association of retinopathy with prevalent stroke in the Cardiovascular Health Study population (aged 69 to 97 years)23 was also weaker than the ARIC study findings.6 This “age-related” decrease in the predictive value of retinopathy seems similar to that seen for other cardiovascular risk factors, such as blood pressure and cholesterol.9,27,28 It may reflect selective mortality, in which persons who manifest retinopathy signs and who are susceptible to the pathologic effects associated with these abnormalities may be more likely to die before they reach age 75 years, leaving a select group of older persons whose biologic or genetic makeup provides relative protection from processes associated with these retinal signs.29 Alternatively, retinopathy signs in older persons without diabetes may represent different and more complex processes than in younger people.23,30
Third, we note the weaker associations when we included cerebrovascular deaths in combined stroke event end point. This may be related to a higher prevalence of comorbid conditions in older people, which may result in increased misclassification of this outcome (e.g., cause-of-death misreporting).
Our study has several important limitations. First, insufficient numbers of incident stroke cases did not allow more precise analysis of incident stroke and TIA separately, nor of stroke subtype. Second, the effects of hypertension and diabetes on these retinal associations are likely to be substantial and may have not been completely adjusted for. We had only a single blood pressure measurement to define hypertension status and a single fasting blood glucose measurement to define diabetes status. Therefore, we cannot completely exclude residual confounding effects. Third, some of the retinal signs might be related to conditions not specified in this study (e.g., hematologic conditions), although these conditions are rare in the general population. Fourth, in Australia, we do not routinely perform autopsies for deaths occurring in the general population and therefore have no data available to validate the cause of death. Finally, our study may have had insufficient power to detect a smaller magnitude of association between focal arteriolar narrowing or AV nicking and subsequent stroke events. Selective survival, as noted above, is also likely to have biased these associations toward the null.
The clinical significance of these findings needs further research. In this study, trained graders performed the retinal photographic assessment in a standardized setting. This assessment protocol needs to be translated to clinical practice using a more simplified grading system before a retinal assessment for stroke risk stratification could be considered to have practical clinical utility.7
Footnotes
-
Disclosure: The authors report no conflicts of interest.
Received October 28, 2004. Accepted in final form June 15, 2005.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.