Are Inflammatory Factors Related to Retinal Vessel Caliber?The Beaver Dam Eye Study

Objective  To examine the relationship of systemic markers of inflammation, endothelial dysfunction, and serum folate level to retinal vessel diameter.

Methods  Cross-sectional analyses were completed for data from a random sample of 396 persons aged 50 to 86 years who underwent a baseline examination from 1988 to 1990. Standardized protocols for blood collection and measurement of markers were used. Diameters of arterioles and venules were measured from digitized photographs. Standard univariate and multivariate analyses were performed.

Results  While controlling for age, smoking status, diabetes status, serum high-density lipoprotein cholesterol, and hematocrit, wider retinal venular diameters were associated with higher serum high-sensitivity C-reactive protein, interleukin 6, and amyloid A levels. While controlling for age, systolic and diastolic blood pressure, smoking, serum high-density lipoprotein cholesterol level, and gout, smaller arteriolar diameters were associated with higher serum amyloid A and lower serum albumin and folate levels but not high-sensitivity C-reactive protein or interleukin 6 levels. Levels of serum soluble intercellular adhesion molecule-1 and serum soluble E-selectin, markers of endothelial dysfunction, were not associated with retinal arteriolar or venular diameters.

Conclusions  These data show an association of inflammatory markers with larger retinal venular diameter, suggesting that retinal venular caliber may be a marker of systemic inflammation.

The retinal blood vessels, accessible to noninvasive visualization, allow one to investigate the structure and pathologic features of the microcirculation and its relationship to systemic vascular diseases. Retinal arteriolar narrowing, for example, was shown to be a sign of chronic systemic hypertension.^1-3 Whether inflammation and other processes are related to changes in retinal vessel caliber is unclear. Data from the Atherosclerosis Risk in Communities Study (ARIC) showed that independent of hypertension and smoking status, nonspecific inflammatory markers (white blood cell [WBC] count, fibrinogen levels, and reduced albumin levels) were associated with a smaller arteriolar-to-venular ratio (AVR), a measure of generalized retinal arteriolar narrowing.⁴ In contrast, data from the Rotterdam study showed WBC count to be associated with larger venules and, to a lesser extent, dilation of arterioles.⁵ These observations are important because of the predictive value of retinal vascular caliber, independent of hypertension and other risk factors, for increased risk of incident systemic diseases including coronary heart disease, diabetes, and hypertension.^6-8 In this article, we describe the relationships of nonspecific (WBC count and serum albumin levels) and specific (high-sensitivity C-reactive protein [hsCRP], interleukin 6 [IL-6], tumor necrosis factor α [TNF-α], and serum amyloid A [SAA] levels) inflammatory markers, IgG antibodies to Chlamydia pneumoniae, and serum soluble intercellular adhesion molecule-1 (s-ICAM-1) and serum soluble E-selectin (s-E-selectin) levels, markers of endothelial dysfunction, to retinal vascular caliber in data from a sample of persons who participated in the Beaver Dam Eye Study.^9,10

We conducted a cross-sectional analysis using data from a random sample of subjects in the Beaver Dam Eye Study cohort. The population and recruitment methods for the full cohort have been described in previous reports.^9,10 In brief, at baseline, a private census of the population of Beaver Dam, Wis, was performed from September 15, 1987, to May 4, 1988, to identify all residents in the city or township of Beaver Dam who were aged 43 to 84 years. The tenets of the Declaration of Helsinki were followed, institutional human experimentation committee approval was granted, and informed consent was obtained from each subject. Of the 5924 eligible individuals, 4926 participated in the baseline examination between March 1, 1988, and September 14, 1990.⁹ Ninety-nine percent of the population was white. Comparisons between participants and nonparticipants at baseline have been presented elsewhere.⁹

A random sample of 400 participants aged 50 to 84 years with gradable lens photographs at the baseline examination was chosen from 2152 Beaver Dam Eye Study participants with dietary information and who had been selected for a different study. A total of 396 of these persons had serum available. Of these persons, 383 had fundus photographs gradable for retinal vessel diameters.

Procedures used at the baseline examination have been described in detail elsewhere.^9,10 The examination included measuring weight, height, and blood pressures (using a random-zero sphygmomanometer and following the Hypertension Detection and Follow-up Program protocol¹¹). A standardized questionnaire was administered, which included specific questions regarding a history of gout, emphysema, arthritis, cigarette smoking, alcohol consumption, and use of insulin and oral hypoglycemic agents, aspirin, other nonsteroidal anti-inflammatory drugs, and multivitamins. Casual blood specimens were obtained at the time of the baseline examination. An aliquot of serum was used immediately for determination of serum albumin, WBC count, hematocrit, and serum total and high-density lipoprotein (HDL) cholesterol levels. Remaining serum was stored without preservative at −80°C in cryogenic vials with O-rings for up to 4.5 years until the vials were shipped on dry ice for a study unrelated to this one. The serum remaining after the analyses was refrozen, shipped back on dry ice, and then stored for an additional 12 years at −80°C until it was shipped on dry ice to the University of Minnesota laboratory for the analyses reported herein.

White blood cell count and hematocrit levels were determined using a Coulter counter method, and serum albumin levels were determined by Technicon, Inc (Tarrytown, NY), at the time of the baseline examination. Serum hsCRP was measured in EDTA plasma using a latex particle–enhanced immunoturbidimetric assay kit (Kamiya Biomedical Company, Seattle, Wash) on the Hitachi 911 Automatic Analyzer (Boehringer Mannheim, Mannheim, Germany). Serum amyloid A was measured using a solid-phase sandwich enzyme-linked immunosorbent assay (ELISA) kit from BioSource International, Inc (Camarillo, Calif). Serum IL-6 was measured using the quantitative sandwich enzyme technique of the ELISA QuantiKine High Sensitivity kit from R&D Systems (Minneapolis, Minn). Serum s-ICAM-1 and s-E-selectin levels were measured using the quantitative sandwich enzyme technique of the ELISA Parameter kit from R&D Systems. The intensity of the color of SAA, IL-6, s-ICAM-1, and serum s-E-selectin was measured on a SpectraMax spectrophotometer (Molecular Devices, Sunnyvale, Calif). Serum TNF-α level was measured using the quantitative sandwich enzyme technique of the ELISA QuantiGlo kit from R&D Systems. The intensity of the color was measured on a Packard luminometer (Packard Instrument Company, Meriden, Conn, incorporated by PerkinElmer, Boston, Mass). Serum folate level was measured on the Hitachi 911 using the CEDIA7 folate enzyme immunoassay (Boehringer Mannheim). The IgG antibodies to C pneumoniae were detected in serum using a microimmunofluorescent antibody assay using a 2-stage sandwich procedure for the qualitative and semiquantitative detection of IgG antibodies to C pneumoniae (Focus Technologies, Cypress, Calif). Positive reactions appear as bright apple-green fluorescent elementary bodies with a background matrix of yolk sac. Fluorescence is graded as follows: 2 to 4+, moderate to intense apple-green fluorescence; 1+, definite but dim fluorescence; negative, no fluorescence. A positive test result is reported when fluorescence is 1+ or greater; a negative result is reported when no fluorescence is seen.

In some assays the range of values was very wide, some being below the lowest standard and some being above the highest standard for the specific test. Because of the limited amount of remaining serum, we truncated the values as follows: hsCRP, <0.01 to >1.4 mg/dL; SAA, <9400 to >300 000 ng/mL; TNF-α, <0.44 pg/mL to no upper limit; IL-6, no lower limit to >10 pg/mL; s-ICAM-1, no lower limit to >878.6 ng/mL; s-E-selectin, no lower limit to >213.4 ng/mL; and folate, <0.6 to >18 ng/mL.

The reference range for serum hsCRP was 0 to 5.5 mg/L, for SAA it was <10 ug/mL (<10 000 000 pg/mL), for TNF-α it was 0 to 4.7 pg/mL, for IL-6 it was 0.45 to 10.0 pg/mL, for s-ICAM-1 it was 115 to 306 ng/mL, for serum s-E-selectin it was 11.9 to 80.7 ng/mL, and for serum folate it was 2.7 to 16.1 ng/mL. The coefficient of variability was 5.5% and 4.5% for serum hsCRP in control samples with mean levels of 0.43 and 1.56 mg/L, respectively; for SAA it was 14.4% for mean values of 13 075 ng/mL in control samples; for IL-6 it was 8.8% and 4.5% with mean values of 1.48 and 2.76 pg/mL in control samples, respectively; for serum TNF-α it was 13.8% and 8.0% for mean values of 2.28 and 89.36 pg/mL; for s-ICAM-1 it was 7.8% for a mean value of 210.2 ng/mL in control samples; for s-E-selectin it was 8.5% for a mean value of 35.9 ng/mL; and for serum folate it was 6.7% for a mean value of 8.6 ng/mL in the sample.

Stereoscopic 30° color fundus photographs centered on the optic disc (Diabetic Retinopathy Study¹² standard field 1) and macula (standard field 2) and a nonstereoscopic color fundus photograph temporal to but including the fovea of each eye were taken.¹⁰ Diameters of retinal vessels were measured after converting the photographs to digital images. All retinal arterioles and venules were measured in the area between 0.5 and 1 disc diameter from the optic disc margin. Measurements of individual arterioles and venules were combined according to formulas developed by Parr and Spears¹³ and Hubbard et al¹⁴ and recently revised by Knudtson et al¹⁵ to provide central equivalents in that eye equal to average arteriolar and venular diameters. The range of the arteriolar diameter equivalent was 105 to 222 μm, and for venular diameter the range was 172 to 309 μm. Reproducibility of ocular gradings has been presented elsewhere.¹⁴ In brief, reliability coefficients were 0.84 for intragrader and 0.79 for intergrader repeated measurements.

Cigarette smoking status at the time of the baseline examination was determined as follows. Subjects were classified as nonsmokers if they had smoked fewer than 100 cigarettes in their lifetime, as ex-smokers if they had smoked more than this number of cigarettes in their lifetime but had stopped smoking before the examination, and as current smokers if they had not stopped smoking. The medical history questionnaire contained questions regarding alcohol consumption. Subjects were asked about past periods of drinking, including whether or not they ever consumed 4 or more drinks daily. From these data, a current heavy drinker was defined as a person consuming 4 or more servings of alcoholic beverages daily, a former heavy drinker had consumed 4 or more servings daily in the past but not in the previous year, and a nonheavy drinker had never consumed 4 or more servings daily on a regular basis. Hypertension was defined as a mean systolic blood pressure of 140 mm Hg or higher, a mean diastolic blood pressure of 90 mm Hg or higher, and/or a history of hypertension with use of antihypertensive medication at the time of examination. Body mass index (BMI) was calculated by dividing weight in kilograms by the square of height in meters. Mean arterial blood pressure was computed as two thirds of the diastolic plus one third of the systolic value. Persons were defined as having diabetes if they reported having this disease and were treated with insulin, oral hypoglycemic agents, and/or diet or were newly discovered to have diabetes at baseline based on age-specific glycosylated hemoglobin values.

Because some of the laboratory data were censored for low and high values, we categorized the laboratory data by quartiles. The quartile variables were then analyzed both as categorical and as ordered factors. We also created a binary variable for each laboratory marker by using the highest (or lowest) 10% of the distribution to define an “abnormal” value. To test for an interaction between endothelial dysfunction and the inflammatory markers, a 3-level variable was created for endothelial dysfunction according to the following scheme: persons with values of both serum s-E-selectin and s-ICAM-1 in the lowest 2 quartiles were considered to have low levels of endothelial dysfunction, persons with values of either serum s-E-selectin or s-ICAM-1 in the highest 2 quartiles were considered to have medium levels of endothelial dysfunction, and persons with values of both serum s-E-selectin and s-ICAM-1 in the highest 2 quartiles were considered to have high levels of endothelial dysfunction.

The distributions for both arteriolar and venular diameters were continuous and relatively normally distributed in the population. Therefore, linear regression procedures were used for both age-adjusted and multivariate-adjusted models. Adjustment variables were determined in 2 ways. First, a stepwise regression procedure was used without the laboratory marker data to identify any potential confounders. Second, any variable known from other studies (eg, diabetes status) to have an influence on both the vessel measurements and the laboratory data was included in the multivariate models. We used SAS statistical software version 9.1 (SAS Institute Inc, Cary, NC) for all analyses.

Persons selected in the sample were generally similar to the whole Beaver Dam cohort except that they were slightly older by design (Table 1). Twenty percent were current smokers, 57% had hypertension, 9% had a history of gout, 2% had a history of emphysema, and 35% used nonsteroidal anti-inflammatory drugs.

Current smoking was associated with higher mean arteriolar and venular diameters, whereas older age and higher serum HDL cholesterol level were associated with lower mean arteriolar and venular diameters (Table 2). A history of hypertension and gout and higher systolic and diastolic blood pressures were associated with smaller arteriolar diameters, whereas higher hematocrit levels and male sex were associated with larger venular diameters (Table 2). No relationships were found between a history of diabetes, use of vitamin supplements, BMI, a history of cardiovascular disease, serum total cholesterol level, a history of thyroid disease, a history of emphysema, and a history of aspirin or other nonsteroidal anti-inflammatory drug use and arteriolar and venular diameters (R.K., unpublished data, 2005).

In the multivariable models, while controlling for other characteristics, higher SAA level was associated with smaller arteriolar diameters (Table 3). Otherwise, none of the other markers of inflammation or C pneumoniae were associated with arteriolar diameters. When folate and serum albumin levels were dichotomized by above and below the 10% cutoff, persons with the highest serum folate (mean arteriolar diameters, 172.0 μm vs 166.1 μm) and the lowest albumin levels (mean arteriolar diameters, 177.8 μm vs 170.9 μm) had the highest adjusted mean arteriolar diameters. Higher serum hsCRP, IL-6, and SAA levels were associated with larger venular diameters (Table 3). These associations remained with or without smoking or mean arterial blood pressure in the model. We reran the multivariate models excluding the one third of the cohort taking anti-inflammatory medications in one analysis and nondiabetic subjects in another and found similar associations, as reported for the full sample (data not shown).

We examined adjusted mean venular diameters by quartile of inflammatory markers and level of endothelial dysfunction markers (Figure). Larger venular diameters were significantly associated with increasing levels of inflammatory markers in persons with high levels of endothelial dysfunction markers but not those with low or intermediate levels.

We found significant relationships between some specific inflammatory markers and larger retinal venular diameter in our sample, independent of other factors. These findings are consistent with those from the Rotterdam Study, which showed a relationship of nonspecific markers of inflammation such as WBC count and lower serum HDL cholesterol level with larger retinal venular diameters.⁵ They are also consistent with a growing body of experimental data showing venular dilation in the presence of active inflammation.^16,17 For example, lipid hydroperoxide administration in the vitreous of rats has been shown to result in an increase in the number of leukocytes in the retinal microvasculature and in an increase in the diameter of retinal venules but not arterioles.¹⁶ Dilation of the retinal vessels has been attributed to increased production of nitric oxide as a result of up-regulated nitric oxide synthase messenger RNA secondary to release of cytokines such as IL-1 and TNF-α.¹⁸ In one study, administration of low doses of an Escherichia coli endotoxin, lipopolysaccharide, in humans was associated with an increase in WBC count and a 9% increase in retinal venous diameter.¹⁷ Larger retinal venules have also been found in other conditions associated with inflammation, such as metabolic syndrome,¹⁹ obesity (J. J. Wang, MMed, PhD, unpublished data, 2005), and type 1 diabetes mellitus.²⁰ In persons with larger BMI, increased leptin levels and increased blood volume may contribute to a larger venular diameter.^21-23 In our study, however, the association of inflammatory markers with venular diameter was independent of diabetic status, HDL cholesterol levels, and BMI.

While controlling for other factors, we found that 2 endothelial dysfunction markers, s-ICAM-1 and s-E-selectin, were not associated with retinal venular diameter. However, we found that the largest venular diameters were found in those persons with the highest levels of both inflammatory and endothelial dysfunction markers. This is consistent with the hypothesis that damage to the vascular endothelial surface by activated leukocytes plays a role in the increase in venular diameter.⁵ Because our data are cross-sectional, we cannot determine the temporal sequence of inflammation, endothelial dysfunction, and venular diameter.

To our knowledge, there have been no previous reports of associations between markers of endothelial dysfunction and retinal venular diameter except for the data from the ARIC study, where we first reported that higher von Willebrand factor and factor VIII levels, 2 markers of endothelial cell dysfunction, were associated with higher levels of AVR.⁴ At that time, we did not report the specific relationship of arteriolar and venular diameters with those markers based on the initial reports that smaller AVR reflected smaller arteriolar diameters without a change in venular diameters.^13,24,25 On re-examination of these data, we found that higher levels of von Willebrand factor but not factor VIII were statistically significantly associated with larger and not smaller arteriolar diameters. Moreover, higher WBC count was associated with larger venular but not arteriolar diameters in the ARIC study (T.Y.W. and R.K., unpublished data, 2005). In the Beaver Dam Eye Study, among the inflammatory-specific markers, only higher SAA was associated with smaller retinal arteriolar diameters. These findings add further weight to our current study findings and show the importance of the evaluation of retinal arteriolar and venular diameters separately rather than using the AVR.

We had hypothesized that high serum folate levels, by protecting against potentially harmful effects of homocysteine on the vascular endothelium, would be associated with larger retinal arteriolar diameters.^26,27 We did not measure serum homocysteine levels in the cohort but did find that high serum folate levels were associated with higher mean arteriolar diameters. We are aware of no other comparable finding.

There are several strengths to our study, including the objective determination of retinal vessel caliber using standardized protocols for photography and grading. However, caution must be observed in interpreting the findings described herein. First, the analyses were cross-sectional and do not provide any insights regarding antecedent-consequent associations. Thus, the temporal sequence of endothelial dysfunction, inflammation, and venular diameter remains to be determined. We collected no information on whether subjects had acute infections or flare-ups in arthritis or other conditions that might have affected the levels of systemic markers of inflammation. Nonetheless, we were able to control for arthritis, emphysema, gout, and urinary tract infection, as suggested by presence of urine nitrites in separate models, and found that these did not affect our findings. Another possible limitation involves the large proportion of subjects (about one third) who were taking anti-inflammatory medications. This might reduce the levels of markers of inflammation, reducing our ability to find an association. However, we included a history of use of these anti-inflammatory agents at baseline in our models, and this did not change the associations found. Although the samples were collected nearly 17 years prior to testing and had been thawed once for other measurements, they were rapidly separated and at −80°C in cryogenic vials with O-ring seals, thus minimizing any degradation of the samples. Therefore, we do not believe that this had a major impact on our findings. The tests chosen were of proteins that are relatively stable in the frozen condition and not affected by thawing.^28-30 Serum hsCRP has been shown in one study to be stable when stored at −70°C for more than 20 years.³¹ Furthermore, 4 cycles of freeze-thaw had no effect on serum IL-6 and TNF-α, and 5 cycles of freeze-thaw did not affect the ability to assay C Pneumoniae (R. Tracy, PhD, and M.Y.T., unpublished data, 2004).

In summary, retinal vascular caliber changes have now been shown to predict a range of systemic vascular diseases such as coronary heart disease, diabetes, hypertension, renal disease, and stroke in different population-based studies.^6-8,32-34 Our data suggest that inflammatory processes may play a role in these previous associations and that a quantitative assessment of retinal vascular caliber may be useful in clinical research studies of systemic and ocular diseases linked with inflammation.

Correspondence: Ronald Klein, MD, MPH, Department of Ophthalmology and Visual Sciences, University of Wisconsin–Madison, 610 N Walnut St, 450 WARF, Madison, WI 53726-2336 (kleinr@epi.ophth.wisc.edu).

Submitted for Publication: April 4, 2005; final revision received June 20, 2005; accepted July 27, 2005.

Reprints: Not available.

Financial Disclosure: None.

Funding/Support: This study was supported by grants EY06594 (R. Klein and B. E. K. Klein) and HL59259 (R. Klein and B. E. K. Klein) from the National Institutes of Health, Bethesda, Md, and in part by Research to Prevent Blindness, New York, NY (R. Klein, Senior Scientific Investigator Award).