eTable 1. Medication Use and Effect on CRAE
eTable 2. Medication Use and Effect on CRVE
Howard KP, Klein BEK, Dreyer JO, Danforth LG, Klein R. Cross-sectional Associations of Medication and Supplement Use With Retinal Vascular Diameter in the Beaver Dam Eye Study. JAMA Ophthalmol. 2014;132(1):23-31. doi:10.1001/jamaophthalmol.2013.6326
Retinal vessel diameters are being measured to examine their relationship with ocular and systemic disease and, in some studies, to calculate the risk of disease. Important factors that directly affect retinal vessel diameters, such as medication use, should be considered when estimating these associations.
To quantify the association between selected medications and supplements and retinal vessel diameters.
Design, Setting, and Participants
In a prospective cohort investigation, 4926 participants (aged 43-86 years at baseline) in the Beaver Dam Eye Study were evaluated every 5 years during 20 years of follow-up from 1988 to 2010.
Main Outcomes and Measures
Central retinal arteriolar equivalent and central retinal venular equivalent measured from the Early Treatment Diabetic Retinopathy Study fundus photograph field 1.
After Bonferroni correction, the use of any blood pressure medication (β = 0.75; P = .04), specifically calcium channel blockers (β = −1.02; P < .001), was significantly associated with wider central retinal arteriolar equivalent adjusting for refraction, photograph focus, age, systolic blood pressure, height, examination phase, educational level, smoking and drinking histories, and presence of diabetes mellitus and emphysema. Use of prostaglandin analogues was marginally associated with narrower central retinal arteriolar equivalent (β = −2.04; P = .09); β-blockers (β = −1.02; P = .10) and oral corticosteroids (β = 2.13; P = .07) were marginally associated with changes in the central retinal venular equivalent.
Conclusions and Relevance
Several medications are associated with central retinal arteriolar and venular equivalents. Prostaglandin analogues, calcium channel blockers, and oral corticosteroids have the largest relative effects. After Bonferroni correction was applied, the use of calcium channel blockers was most strongly associated with change in the central retinal arteriolar equivalent. The presence of factors that are associated with retinal vessel diameters should be considered when using retinal vessel diameter as an outcome or when using these measures to assess the risk of systemic or ocular disease.
Retinal vessel diameters (RVDs) are associated with cardiovascular diseases such as hypertension, myocardial infarction, and cardiovascular mortality.1- 3 Many factors affect RVDs. Men tend to have larger RVDs than women, and older individuals tend to have narrower RVDs than younger individuals.4,5 Other systemic factors, such as blood pressure, glycemia, lipid levels, obesity, smoking, renal function, and chronic inflammation, also affect RVDs.6- 9
Retinal vessel diameters are also associated with ocular characteristics and diseases.10- 14 Wider retinal venules are associated with severity of diabetic retinopathy10 and open-angle glaucoma.12 One study15 reported evidence of an association between RVDs and age-related macular degeneration.
A class of exposures that is often ignored in examining associations between cardiovascular end points and RVDs is use of medications and supplements. Many different classes of drugs have the potential to alter RVDs. Medication associations with RVDs may include not only cardiovascular medications classified as vasodilators but also other blood pressure medications, systemic medications for other conditions, and herbal supplements.16- 18 Of particular interest are relatively new medications in classes such as phosphodiesterase 5 inhibitors, angiotensin receptor–blocking agents, and angiotensin-converting enzyme inhibitors as well as agents that act as antioxidants and others that have anti-inflammatory actions.19,20 These medications have the potential to alter RVDs as well as the associations of RVDs with cardiovascular disease. In addition, garlic extract or powder, taken as a nonvitamin, nonmineral dietary supplement, may have direct effects on hypertension21 that may be secondarily associated with smaller retinal arteriolar diameters. Garlic also may decrease oxidative stress that could affect vessel diameters.17
Because of the potential for RVDs to be associated with diseases not only in the eye but in other organs, exposures that are related to these diameters may affect health outcomes. The following is a report of the findings regarding the use of medications and their effects on concurrent retinal venular and arteriolar diameters from the Beaver Dam Eye Study, a population cohort study with 20 years of follow-up.
There were 4926 persons aged 43 to 86 years who participated in the baseline examination of the Beaver Dam Eye Study in 1988-1990. Ninety-nine percent of the population was of European ancestry. Written informed consent was obtained from participants before each examination, and all examinations followed a similar protocol that was approved by the institutional human subjects committee of the University of Wisconsin and conformed to the tenets of the Declaration of Helsinki. The participants did not receive financial compensation.
Data regarding anthropometric measurements, health history, educational level, and lifestyle factors were collected at each examination. Blood samples were obtained. Diabetes mellitus was defined as self-reported use of insulin or diet to control blood glucose levels and confirmed by a glycated hemoglobin level above 6.5% or by absence of self-report with a glycated hemoglobin level above 7.0%. Cardiovascular disease was defined as self-report of stroke, myocardial infarction, or angina. Smoking status was defined as never, past, (smoked >100 cigarettes in lifetime but not currently smoking), or current. Ever-heavy drinking was defined as self-report of ever drinking more than 4 alcoholic beverages daily on a regular basis. Refraction was measured according to a modification of the Early Treatment Diabetic Retinopathy Study protocol22 to obtain the best-corrected visual acuity when the automated refraction yielded visual acuity of 20/40 or worse. Serum creatinine was measured using frozen samples from a subset of participants at baseline, as well as at the 5-, 10-, and 15-year examinations; it was not measured at the 20-year examination. Estimated glomerular filtration rate was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation23 and dichotomized into categories of 60 mL/min/1.73 m2 or more or less than 60 mL/min/1.73 m2.
Participants were asked to bring all current medications to the examination. These, as well as if a participant was not taking any medications, were recorded. Examiners followed up with participants by telephone to verify the name of any medication that was reported but not brought to the examination. All medication names were recorded by trained examiners, and coding classification was adapted from the American Hospital Formulary Service. Vasodilators were identified by American Hospital Formulary Service code 2412 or as being a direct vasodilator, such as minoxidil or hydralazine.24 Medication dosage form was recorded; however, frequency of use, dose, time since initiation, and, for ocular medication, the eye in which medication was used were not specified or recorded.
Stereoscopic 30° color retinal photographs of the optic disc were converted to digital images by a high-resolution scanner (Nikon LS2000; Nikon Inc) using standard settings for all photographs. Images were displayed on 21-inch monitors set to a screen resolution of 1600 × 1200 pixels. The diameters of all arterioles and venules coursing through a standard area 0.5 to 1.0 disc diameters from the optic disc margin (zone B) were measured using a custom computer program.25 Photograph focus was subjectively determined by a trained grader to be good, fair, or poor. For these analyses, vessel measurements from the right eye were used, and a participant was required to have data for both central retinal arteriolar equivalent (CRAE) and central retinal venular equivalent (CRVE).
To be included in the analyses, participants must have been evaluated at 1 or more of the 5 examination phases of the Beaver Dam Eye Study. Of the 15 898 total person-visits, 1282 visits were excluded because of missing or ungradable photographs. An additional 1534 person-visits were excluded because of cataract surgery in the right eye. In addition to having complete data for vessel measurement in the right eye and medication use, participants included in the analyses were required to have complete data for age, sex, height, systolic blood pressure, smoking history, presence of diabetes, history of emphysema, right eye refraction, and a graded measure of retinal photograph focus. An additional 1104 person-visits were excluded because of missing information for 1 or more covariate. After exclusions, 11 978 person-visits remained for analyses. Medications and supplements included in primary analyses were restricted to those used by at least 40 persons at a minimum of 1 visit (ie, ranging from <1% at the baseline visit in 1988-1990 to 2.5% at the most recent visit in 2008-2010).
To analyze the immediate effect of vasodilating medications on RVDs, we explored the relationship between vessel diameters and concurrent medication use cross-sectionally at each examination using a mixed-modeling approach to account for the correlation between the repeated measurements for a person seen at multiple visits during 20 years of follow-up. We report P values for categorical variables for each category as compared with the reference category. The overall P value for a categorical variable is a type 3 P value that tests for any difference between any of the categories; the test of trend models categorical variables as continuous to identify a linear trend with RVDs for the ordered categories. The β estimate reported for the effect of medications on CRAE and CRVE represents the change in RVDs associated with taking a medication compared with not taking the given medication. We tested for significant multiplicative interactions between medications and examination phase and between medications and inferred indication for their use. Any interaction that was significant at the α = .05 level in the minimally adjusted model is reported. Bonferroni corrections26 for multiple comparisons are reported with P values calculated without such corrections. For the major categories of medications (vasodilator, prostaglandin analogue, any nonsteroidal anti-inflammatory drug, any blood pressure medication, β-blocker, any diuretic, corticosteroid, herbal vasodilator, and garlic supplement), a correction was made for 9 comparison groups. For all subcategories, a correction was made for 29 comparison groups. Statistical analysis was performed using SAS, version 9.3 (SAS Institute, Inc).
Age, sex, height, systolic blood pressure, refraction, and photograph focus affect vessel measurement from photographic images; therefore, these covariates were included in the minimally adjusted model. We also adjusted for examination phase because there may be time trends associated with medication use (eg, changing indications) that may in turn affect RVDs. The maximally adjusted model included all factors in the minimally adjusted model as well as educational level, smoking, emphysema, ever-heavy drinking, and diabetes.
Although a specific reason for medication use was not reported, medical history including emphysema, diabetes, and cardiovascular disease was obtained and analyses for confounding by inferred indication were performed. Stratified analyses were used to identify relationships that may have been the result of illness rather than the given medication. Results were confirmed by testing for interactions between medication use and illness status to identify significant differences in the effect of the given medication on RVD between individuals with and those without the specified illness.
A subset of analyses was performed to examine the effect of renal function (estimated glomerular filtration rate, ≥60 vs <60 mL/min/1.73 m2) on the minimally and maximally adjusted models. Approximately 13% of the person-visits included in our primary analysis were missing a measure of serum creatinine.
Participants whose data contributed to these analyses, as reflected in person-visits over all examination phases, tended to be younger and to have a higher body mass index, higher diastolic blood pressure, and fewer comorbidities, such as diabetes, history of cardiovascular disease, and emphysema, than did those whose data were excluded (Table 1).
Characteristics other than medication status associated with RVD are summarized in Table 2. Mean CRAE was narrower in the presence of increased age, greater body mass index, more years of education, and higher systolic and diastolic blood pressures. Mean CRAE was wider in the presence of a history of cardiovascular disease, current smoking, emphysema, diabetes, poor focus of the fundus photograph, and more hyperopic refractive error. The mean CRVE was narrower in the presence of increased age, more years of education, and higher systolic blood pressure. Mean CRVE was wider in the presence of greater height, male sex, history of cardiovascular disease, current smoking, heavy drinking, diabetes, poorer focus of the fundus photograph, and more hyperopic refraction.
Medication use varied by study visit (Table 3). Vasodilators were being taken at 624 person-visits. Medications that are hypothesized to have vasodilating effects, other than those so classified by the American Hospital Formulary Service, were relatively commonly used at some visits (eg, antihypertensives, diuretics, nonsteroidal anti-inflammatory drugs, and prostaglandin analogues). Within medication categories, the patterns of use varied over time, which is attributable in part to the availability of some (usually newer) medications.
Unadjusted mean RVDs by medication use, the effects of medication use on concurrent RVDs in the minimally and maximally adjusted models, and results of Bonferroni correction are presented in the Supplement (eTables 1 and 2) for CRAE and CRVE, respectively. Use of prostaglandin analogues was associated with CRAE in the minimally and maximally adjusted models; however, after correction for multiple comparisons, the use of these agents was only marginally associated with CRAE (β = −2.04; corrected, P = .09). There was a significant interaction between the use of any blood pressure medication and examination phase and between the use of calcium channel blockers and examination phase. Taking each of these medications was associated with increased CRAE at each examination phase, and the impact of taking the medication was stronger at later examination phases. Therefore, the adjusted main effects of taking any blood pressure medication and calcium channel blocker are presented in the tables. After correction for multiple comparisons, the effect of use of any blood pressure medication (β = 0.73; corrected, P = .04), specifically calcium channel blockers, was associated with a wider CRAE (β = 1.52; corrected, P < .001).
β-Blockers (β = −1.02; P = .01) were associated with CRVE in the minimally and maximally adjusted models but were only marginally associated (β = −1.02; corrected, P = .10) after correction for multiple comparisons. Use of oral corticosteroids, although relatively uncommon, was marginally associated with CRVE after Bonferroni correction (β = 2.13; corrected, P = .07).
Stratified analyses were performed to examine confounding by presumed indication (data not shown). The effect of vasodilators (P = .82), oral nitroglycerin (P = .25), blood pressure medications (P = .22), calcium channel blockers (P = .34), and oral β-blockers (P = .10) on CRAE were not significantly different between individuals with and those without self-reported cardiovascular disease. The effects of any β-blockers (P = .20), oral β-blockers (P = .06), and diuretics (P = .60) on CRVE were also not significantly different in persons with and those without self-reported cardiovascular disease. The effect of oral corticosteroids on CRVE was not significantly different between persons with and those without diabetes (P = .95) or between those with and those without emphysema (P = .86).
A subset of analyses was performed to examine the effect of renal function (estimated glomerular filtration rate, ≥60 vs <60 mL/min/1.73 m2) on the relationships between medications and RVDs in the maximally adjusted model. The addition of this variable did not significantly alter the effects of medications on RVDs compared with the results of analyses for the full population (data not shown).
Analyses of associations were undertaken to examine relationships of medication use and RVD in a large representative American community during 5-year intervals. In multivariable adjusted models during 20 years in a population-based study, there were several significant (P < .05) and marginally significant (P < .10) associations after Bonferroni correction for multiple comparisons.
Calcium channel blockers appear to have relatively large associations with increased CRAE and should be considered when examining the effects of other exposures on CRAE. These medications are widely used, especially in older adults, and therefore may be more important on a population basis. A possible mechanism for the widening effect of calcium channel blockers on CRAE is their reported relaxation effect on the microcirculation.27,28
The changing effect of any blood pressure medication, specifically calcium channel blockers, during the 20-year follow-up period may be associated, in part, with the evolving definition of clinical hypertension. Lower systolic and diastolic blood pressures have been shown29- 31 to be significantly associated with wider CRAE, and because persons are currently prescribed antihypertensive medication at lower blood pressure measures than in the past, the association of the medication appears to become stronger (wider CRAE) over the course of the examination.
The relationship of prostaglandin analogues to CRAE is in the opposite direction, although the significant relationship was attenuated when adjusting for multiple comparisons. This class of medications is used relatively often in persons with suspected and confirmed glaucoma to reduce intraocular pressure. This association with CRAE was found after adjustment for age and blood pressure. A researcher might attribute a smaller CRAE solely to blood pressure and/or age when part of the effect may be due to use of this category of medications.
Use of oral corticosteroids was marginally associated with a wider CRVE in models that included other associated characteristics, such as diabetes and smoking. Although relatively few persons were taking these medications, the effect was relatively large and suggests the need to stratify by such use in analyses that are aimed at CRVE as an outcome. The opposite effects were found for individuals using β-blockers. These exposures should be considered in studies examining RVDs.
The associations of RVDs with phosphodiesterase 5 inhibitors were not addressed in the primary analyses because there were few users in this older population (n = 25). These medications were associated with wider retinal arterioles (β = 3.39; P = .05) and were suggestive of an association with retinal venules (β = 3.88; P = .11) in the maximally adjusted models without adjusting for multiple comparisons. These medications may have strong vasodilating effects; those effects should be evaluated in populations in which their use is more widespread than has been reported in the Beaver Dam Study during the intervals under examination.
There are many limitations to our investigation. First, we focused on medications for which a vasodilating effect had been reported in the literature and to medications and supplements for which such effects were hypothesized. Thus, we may have excluded other agents possibly associated with RVDs. In addition, our analyses may have underestimated the use and effects of unreported medications on RVDs. However, underreporting of the medications included in this analysis was minimized because the examination contained questions specifically regarding use of glaucoma drops, products containing aspirin, and antihypertensive and diuretic agents. Second, we examined the associations between RVD and given classes of medications rather than analyzing the effects of specific agents within a class (eg, β-blockers, such as propranolol and carvedilol, may have different effects on RVDs). Third, we examined relationships of each class of medication independently of others that may be taken in combination by the same individual. To the extent that there are synergistic or antagonistic effects on the outcome, we may have underestimated or overestimated associations of any given class of medications or supplements. Fourth, the dose, frequency, and history of medication use were not recorded. For participants who reported use of eyedrops, the eye in which medication was administered was not recorded. To this end, we cannot evaluate time trends or dose response from our data. Also, the effect of eye-specific medications may be overestimated or underestimated.
The Bonferroni corrections made for multiple comparisons are somewhat conservative and may have underestimated the relationships between some medications and RVDs. This may lead to discarding observed findings that may be meaningful32; thus, we reported both P values so that readers might evaluate the results of our analyses from the observed data. In addition, our investigation was observational and prospective. As such, we report the temporal finding of exposures to an end point. Although an association in this context may be causal, we cannot be certain. Confirmation of these findings needs corroboration or replication and, if possible, further support from bench research to be more certain of causality. Finally, our analysis excluded persons with cataract severe enough to affect the ability to obtain and/or grade the retinal photographs, as well as persons who had undergone cataract surgery.
In summary, the role of RVDs as risk factors or as important confounders of the relationships of other exposures to disease is still uncertain. If the associations of specific medications to RVDs are confirmed, consideration of effect and possible adjustment for such use will be important in evaluating the relationship of the RVDs to the medical conditions in question.
Submitted for Publication: April 17, 2013; final revision received June 21, 2013; accepted July 24, 2013.
Corresponding Author: Kerri P. Howard, MS, Department of Ophthalmology and Visual Sciences, University of Wisconsin–Madison, 610 N Walnut St, 405 WARF, Madison, WI 53726 (firstname.lastname@example.org).
Published Online: December 19, 2013. doi:10.1001/jamaophthalmol.2013.6326.
Author Contributions: Ms Howard had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: B. E. K. Klein, R. Klein.
Acquisition of data: B. E. K. Klein, Dreyer, Danforth, R. Klein.
Analysis and interpretation of data: Howard, B. E. K. Klein, Dreyer, Danforth.
Drafting of the manuscript: Howard, B. E. K. Klein, Dreyer.
Critical revision of the manuscript for important intellectual content: Howard, Danforth, R. Klein.
Statistical analysis: Howard.
Obtained funding: B. E. K. Klein, R. Klein.
Administrative, technical, and material support: B. E. K. Klein, Dreyer, Danforth, R. Klein.
Study supervision: B. E. K. Klein.
Conflict of Interest Disclosures: Dr B. E. K. Klein has served as a member of a Scientific Advisory Committee for Pfizer. No other disclosures were reported.
Funding/Support: This research was supported by grant EY06594 from the National Institutes of Health (Drs R. Klein and B. E. K. Klein), the Retina Research Foundation, and an unrestricted grant from Research to Prevent Blindness.
Role of the Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily reflect the official views of the National Eye Institute or the National Institutes of Health.