Study inclusion diagram. RVO indicates retinal vein occlusion.
Incidence of myocardial infarction (MI) in the retinal vein occlusion (RVO) and control cohorts. Error bars represent 95% confidence intervals.
Relative risk of myocardial infarction (MI) for patients with central (CRVO) or branch (BRVO) retinal vein occlusion. Error bars represent 95% confidence intervals.
Incidence of cerebrovascular accident (CVA) in patients with retinal vein occlusion (RVO) and control patients. Error bars represent 95% confidence intervals.
Relative risk of cerebrovascular accident (CVA) for patients with central (CRVO) or branch (BRVO) retinal vein occlusion. Error bars represent 95% confidence intervals.
Werther W, Chu L, Holekamp N, Do DV, Rubio RG. Myocardial Infarction and Cerebrovascular Accident in Patients With Retinal Vein Occlusion. Arch Ophthalmol. 2011;129(3):326-331. doi:10.1001/archophthalmol.2011.2
Copyright 2011 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2011
To compare the incidence rates of myocardial infarction (MI) and cerebrovascular accident (CVA) in hospitalized patients with and without branch or central retinal vein occlusion (RVO).
In this retrospective cohort study, a US population-based health care claims database was used to identify patients with RVO and control patients, matched for age and sex. Events of MI, CVA, and covariates were identified for patients with and without RVO. Incidences of MI or CVA events prompting hospitalization and adjusted rate ratios (RRs) were calculated; RRs were adjusted for covariates consistent with risk factors for outcomes.
Of 4500 patients with RVO and 13 500 controls, the event rates for MI were 0.87 per 100 person-years and 0.67 per 100 person-years, respectively. The adjusted RR for MI was 1.03 (95% confidence interval [CI], 0.75-1.42; P = .85 for RVO vs controls). Event rates for CVA were 1.16 and 0.52 per 100 person-years for RVO and controls, respectively. The adjusted RR for CVA was 1.72 (95% CI, 1.27-2.34; P = .001) for RVO vs controls.
This study provides quantitative data on the incidence of cardiovascular and cerebrovascular outcomes in patients with RVO in a large US population-based health care claims database. Event rates for MI were similar in patients with RVO and controls; however, the event rate for CVA in patients with RVO was almost 2-fold that observed in controls.
Retinal vein occlusion (RVO) is a retinal vascular disease in which a retinal vein is compressed by an adjacent retinal artery, resulting in blood flow turbulence, thrombus formation, and retinal ischemia. Retinal vein occlusion at the optic nerve is defined as central RVO (CRVO), and RVO where an artery crosses the vein is defined as branch RVO (BRVO). Although RVO is a significant cause of severe visual impairment in adults, it can occur at any age.1,2 Risk factors identified for RVO include older age, hypertension, vascular disease, and diabetes mellitus.3- 8 Several risk factors for RVO are also risk factors for arterial thromboembolic events (ATEs), such as myocardial infarction (MI) and cerebrovascular accident (CVA).9,10 Therefore, it is clinically relevant to assess whether patients with RVO are at increased risk for ATEs.
Several studies11,12 have demonstrated that the prevalence, morbidity, and mortality of cardiovascular and cerebrovascular diseases were not increased in patients with CRVO compared with matched control patients; however, 1 study11 showed an increased prevalence of hypertension and diabetes mellitus in these patients. More recently, Cugati and colleagues5 showed a 2-fold increase in cardiovascular mortality in patients with RVO aged 43 to 69 years. Finally, in 2 recently published database cohort studies of a Taiwanese population,13,14 no increased risk of MI or stroke was found in patients with RVO. A few of these studies are limited because of small patient populations.11,12 Furthermore, 2 of the studies regarding the incidences of cerebrovascular and cardiovascular diseases have focused on Asian populations with RVO.13,14 Thus, we performed a cohort analysis using a large US population-based health care claims database, with a demographic distribution approximately equivalent to that of the US population as a whole, to establish rates of MI and CVA in patients with and without RVO.
In this retrospective cohort study, we separately compared annual incidence rates of MI and CVA events prompting hospitalization in patients with and without RVO between January 1, 2002, and December 31, 2005. We used an administrative health care claims database (i3 InVision Data Mart; Ingenix, Eden Prairie, Minnesota) that captures medical and pharmacy services for more than 30 million persons in the United States. The database consists of adjudicated and paid claims among enrollees in 11 health plans or public insurance programs. The medical, pharmacy, and laboratory claims records were aggregated into the database using patient identification numbers to protect the privacy of patients while allowing for longitudinal analysis of claims. Complete pharmacy and medical claims data (including Medicare and Medicaid) were available for all the patients in this study. The database patient population has a demographic distribution approximately equivalent to that of the US population as a whole, although this database has a larger proportion of persons older than 18 years and younger than 65 years and a relatively larger proportion from the South and Midwest.
A diagram of the inclusion and exclusion schema for patients with RVO and controls without RVO in this study is presented in Figure 1. Patients with RVO had at least 2 claims that included International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for CRVO (code 362.35) or BRVO (code 362.36). Two claims on separate days were required to minimize misclassification of RVO for patients with 1 claim for diagnostic testing or “rule out” diagnoses. The index date for each patient was the date of the first RVO claim.
Controls were those without evidence of ICD-9-CM coding for RVO (as defined) during the study period. From this pool of potential controls, 3 patients were randomly selected for each identified patient with RVO. Patients were matched by sex and age (±3 years) to create the final cohort of controls. Each control was assigned the same index date as the corresponding patient with RVO.
Acute MI events were identified by ICD-9-CM codes 410.xx. Ischemic or hemorrhagic CVAs were identified by ICD-9-CM codes 431.xx, 432.xx, 433.x1, 434.x1, and 436.xx. Both MI and CVA events were identified using claims consistent with “inpatient hospital” place of service or health care category. Outcome events with admission and discharge on the same day were excluded because they are not reliable indicators of true events and most likely represent claims for diagnostic procedures. Adjacent hospital claims with the same provider identification were counted as a single hospitalization if they were not separated by at least 1 day. Each outcome (acute MI and CVA) was analyzed separately. Only the first qualifying claim identified after a patient's index date was counted as an event.
For each patient, outcome events were identified along with total observation time expressed to the nearest day. Person-time was calculated for each patient as the time between the index date and the date of censoring (the loss of patients during follow-up). Rules for censoring were defined to accurately determine the period during which patients were at risk for MI or CVA events. Patients were censored at the first occurrence of an event, disenrollment from the insurance plan, or the end of the study period, whichever came first. The times to MI and CVA events were analyzed separately. Event rates and 95% confidence intervals (CIs) were calculated for each cohort and for each outcome by age and sex as the number of events per 100 person-years (PYs). Statistical significance was determined by lack of overlapping 95% CIs.
Cox regression analysis was used to evaluate time-to-event rates in both cohorts using a software program (SAS version 9.1; SAS Institute Inc, Cary, North Carolina). Rate ratios (RRs) with 95% CIs and P values were calculated. Covariates were added to the model to adjust for the presence of known risk factors for MI or CVA during the 183 days of continuous eligibility preceding the index date. This approximately 6-month period allowed for the detection of medical office visits for chronic conditions, such as hypertension and hyperlipidemia. Covariates included history of MI or CVA (inpatient only); congestive heart failure (inpatient only); cardiac arrhythmia (inpatient only); other cerebrovascular disease, including transient ischemic attack, diabetes mellitus, hypertension, hyperlipidemia, heart disease (coronary artery disease), and angina; and previous use of prescription anticoagulant or antiplatelet drugs.
In addition, the Charlson score was used as a measure of health status in the 183-day period before the index date. The Charlson Comorbidity Index (CCI) was developed as a method for estimating the risk of death from comorbid disease. The CCI is weighted by the number and severity of comorbid diseases, including but not limited to MI, cerebrovascular disease, chronic pulmonary disease, dementia, diabetes, and various malignancies.15 A modified version of the CCI is commonly used for adjustment of health status in claims analyses.15- 18 In this study, ICD-9-CM codes were used to calculate CCI scores for each patient. The CCI was modified to exclude codes for MI and CVA because they are the outcomes of the analysis. The CCI score was included as a continuous variable in the regression model such that each 1-point increase in CCI score could be assessed.
Patients with RVO (n = 4500) and controls (n = 13 500) were included in this analysis (Figure 1). Among patients with RVO, 62.9% had claims for BRVO (n = 2830) and 37.1% had claims for CRVO (n = 1670) (Table 1). Patients with RVO and controls were matched in terms of age (mean age, 64 years) and sex (50.2% female) (Table 1). Mean duration of follow-up was 550 days for patients with RVO and 506 days for controls. Modified CCI scores were generally higher in the RVO cohort than in the control cohort; a larger proportion of patients with RVO had scores of 2 or greater (22.2% [RVO] vs 12.9% [control]), reflecting their greater disease burden (Table 1). Patients with RVO had a significantly higher percentage of claims for comorbidities, including angina, cardiac arrhythmia, diabetes, heart disease, ATEs, hyperlipidemia, hypertension, and cerebrovascular events other than stroke, than did controls (Table 1).
In the RVO cohort, 58 patients had claims for MI, yielding a rate of 0.87 (95% CI, 0.67-1.12) MI events per 100 PYs, and in the control cohort, 125 patients had MIs, yielding a rate of 0.67 (0.56-0.80) MI events per 100 PYs (Figure 2). Although men and patients younger than 65 years with RVO had a 1.6- and 1.9-fold higher risk of MI, respectively, compared with controls, there were no statistically significant differences in MI rates between patients with RVO and controls when they were stratified by sex or age (<65 or ≥65 years) (Figure 2).
After adjustment for risk factors and comorbidity scores, the difference in rates of MI events was not significant between the RVO and control cohorts (adjusted RR, 1.03; 95% CI, 0.75-1.42; P = .85) (Table 2 and Figure 3). Positive predictors of MI among all patients with RVO, using the Cox regression model, included higher CCI score (P < .001), congestive heart failure (P = .02), diabetes mellitus (P = .01), and heart disease (P = .001) (Table 2). The RRs for patients with BRVO (adjusted RR, 1.07; 95% CI, 0.73-1.57; P = .73) and CRVO (0.97; 0.55-1.72; P = .92) were similar to those for all patients with RVO (Table 2 and Figure 3). Positive predictors of MI among patients with BRVO included angina (P = .04), higher CCI score (P < .001), and heart disease (P = .003); positive predictors of MI among patients with CRVO included higher CCI score (P < .001), congestive heart failure (P = .01), and diabetes mellitus (P = .04) (Table 2).
In the RVO cohort, 78 patients had claims for CVA, yielding a rate of 1.16 (95% CI, 0.93-1.45) CVA events per 100 PYs, and in the control cohort, 96 patients had claims for CVA, yielding a rate of 0.52 (0.42-0.63) CVA events per 100 PYs (Figure 4). Rates of CVA were significantly higher in patients with RVO than in controls regardless of sex or age (<65 or ≥65 years) (Figure 4).
After adjustment for risk factors, the rate of CVA events in the RVO cohort was almost double that in the control cohort (adjusted RR, 1.72; 95% CI, 1.27-2.34; P = .001) (Table 3 and Figure 5). Positive predictors of CVA among all patients with RVO, using the Cox regression model, included angina (P = .003), higher CCI score (P < .001), hypertension (P = .01), and other cerebrovascular diseases, including transient ischemic attack (P = .004) (Table 3). The RRs were similar when patients with RVO were stratified by BRVO (adjusted RR, 1.79; 95% CI, 1.18-2.72; P = .01) and CRVO (1.57; 0.99-2.49; P = .06) (Figure 5). Positive predictors of CVA in patients with BRVO included higher CCI score (P < .001), hyperlipidemia (P = .003), and hypertension (P = .05); positive predictors of CVA in patients with CRVO included angina (P = .03), and higher CCI score and other cerebrovascular diseases (P < .001 for each) (Table 3).
In this retrospective cohort study, patients with RVO had an almost 2-fold higher incidence of CVA than that of age- and sex-matched controls. This significantly higher incidence of CVA in patients with RVO held true when adjusting for cardiovascular comorbidities. Similarly, a significantly higher incidence of CVA was observed when analyzing only patients with BRVO; the incidence of CVA approached statistical significance when analyzing only patients with CRVO. In contrast, the incidence of MI was not statistically significantly different between patients with RVO and age- and sex-matched controls. The numerical increase in MI incidence rates for patients with RVO compared with controls was attenuated with adjustment for cardiovascular comorbidities as indicated by RRs closer to 1 when analyzing patients with BRVO and CRVO separately. However, certain RVO subgroups, for example, men and patients younger than 65 years, seemed to have an increased risk of MI, but results were not statistically significant, and the clinical significance of these findings is unknown.
Retinal vascular occlusive disease is characterized by numerous systemic risk factors. Thus, in this large retrospective cohort study representing a broad demographic of the US population, it was not surprising to find that patients with RVO carried a significant disease burden. Patients with RVO had a significantly higher percentage of claims for angina, cardiac arrhythmia, diabetes, heart disease, ATEs, hyperlipidemia, hypertension, and cerebrovascular events other than stroke compared with age- and sex-matched controls.
Predictive comorbidity factors for patients with RVO who went on to develop MI included higher CCI score, congestive heart failure, diabetes mellitus, and heart disease. Predictive comorbidity factors for patients with RVO who went on to develop CVA included higher Charlson score, angina, hypertension, and other cerebrovascular diseases, such as transient ischemic attack. Although patients with RVO and a history of ATEs were approximately 2 times more likely than controls to have an MI or CVA, this increased risk was not statistically significant. These findings are of particular interest in light of a recent analysis of MI and CVA incidence in patients with age-related macular degeneration. In this retrospective analysis of the 5% Medicare Database from 2001 to 2003, patients with age-related macular degeneration and previous ATEs were at higher risk for subsequent events: 7.4% for MI and 35.1% for CVA.19 Patients with RVO tend to be younger than those with age-related macular degeneration, but both retinal conditions have a strong association with systemic cardiovascular diseases.
This study found a higher prevalence of BRVO claims (62.9%) than CRVO claims (37.1%) during the study period. The 15-year cumulative incidences of RVO in the Beaver Dam Eye Study20 were 1.8% for BRVO and 0.5% for CRVO. Thus, BRVO was more than 3-fold as common as CRVO in a large population-based study in which fundus photographs were used to make the diagnosis.20 The present results may differ from those of the Beaver Dam Eye Study because only symptomatic patients seeking health care would be detected with the “paid claims” method used in this study. Patients with BRVO may be asymptomatic depending on the location of the BRVO and whether macular edema or vitreous hemorrhage develops; therefore, patients with BRVO may be underreported in this study.
The relative risk of MI was similar for patients with BRVO and CRVO compared with the entire RVO cohort. The relative risk of CVA was significantly higher in the BRVO group only. Although the risk of CVA was not statistically significant in patients with CRVO (P = .06), this could be attributed to the smaller sample size of patients with CRVO compared with that of patients with BRVO and, consequently, to the larger CIs.
A well-established index of health status, the CCI score, was a positive predictive factor for patients with BRVO and CRVO experiencing MI or CVA; however, other predictive factors differed for these 2 patient groups, including other cerebrovascular diseases, such as transient ischemic attack, which were approximately 6-fold more predictive for CVA in CRVO than in BRVO, potentially relating to differences in systemic vascular health between patients with BRVO and CRVO. This suggests that BRVO and CRVO, although characterized as venous occlusive diseases, may in fact be different clinical manifestations with separate demographics and different predictive factors. Indeed, clinical experience shows that patients with BRVO are far more likely to have recurrent BRVO than to develop CRVO. Similarly, patients with CRVO are far more likely to have recurrent CRVO or bilateral CRVO than to develop BRVO. It is uncommon for a single patient to have CRVO and BRVO.
In contrast to the findings of this study, previous studies have suggested that patients with RVO are not at increased risk for CVA. In general, most of these studies5- 7,12 were small retrospective case series in which a cohort of patients with known RVO was followed up for morbidity or mortality related to CVA. However, a recent study by Ho and colleagues13 made use of a large national database similar to the one used in this study. Data were collected from the Taiwan National Health Insurance Research Database, and an increased risk of CVA was observed only in patients with RVO aged 60 to 69 years.13 In the present study, the average age of patients with RVO was 64 years. Thus, these 2 studies may be expected to have consistent findings. Of note, however, the work by Ho and colleagues13 compared a smaller number of patients (350 patients and 2100 controls). It is possible that the number of CVA events was too small in the younger and older age groups for statistical significance to be determined. Another possible explanation for the different findings between the 2 studies is ethnicity: the Taiwanese study examined a Chinese population, whereas this study examined a large US population. Although exact ethnic demographic data were not available, the database patient population has a demographic distribution approximately equivalent to that of the US population, suggesting a more diverse patient population. The risk of CVA may be different for ethnically diverse populations.
To our knowledge, ours is the first study to report an increased incidence of CVA in patients with RVO in a US population. The strengths of this study include the large number of patients included in the database, the geographically diverse sampling of the entire United States, and the reliable and validated method by which the data collected represent medical care received for each outcome (CVA and MI).16,21 The weaknesses of this study include possible underreporting of asymptomatic BRVO or CRVO; the possibility of misclassification of diagnoses and procedures; the paid claims method, which may result in the underreporting of deaths; and the inability to collect other important health-related information, such as smoking status and weight. Although population-based health care claims databases allow for robust analyses of large patient populations, we must acknowledge that subgroup analyses based on rare risk factors (eg, history of ATEs) may be underpowered for identifying statistically significant results. Nevertheless, these data suggest that physicians and patients should be aware of the possible increased risk of CVA but not of MI in patients with RVO. Additional studies of large diverse populations are necessary to confirm these findings.
Correspondence: Winifred Werther, PhD, Department of Global Patient Safety, Vertex Pharmaceuticals, 130 Waverly St, Cambridge, MA 02139 (firstname.lastname@example.org).
Submitted for Publication: January 6, 2010; final revision received April 22, 2010; accepted July 6, 2010.
Financial Disclosure: Drs Werther and Rubio and Ms Chu are employees of Genentech Inc.
Funding/Support: This study was funded by Genentech Inc.
Role of the Sponsor: Genentech Inc participated in the analysis, writing, and review of the manuscript and provided support for third-party writing assistance.