Serum Ferritin and Hemoglobin Are Independently Associated With Wider Retinal Venular Caliber: The Tromsø Study 2001–2008

Abstract

Purpose.: To investigate if markers of iron body stores and hemoglobin are associated with retinal vascular caliber.

Methods.: This was a population-based study including 2993 participants of the fifth and sixth surveys of the Tromsø Study in Norway, conducted in 2001–2002 and 2007–2008. The association between levels of s-ferritin, transferrin saturation, and hemoglobin in the fifth survey and retinal vascular caliber in the sixth survey was assessed by multivariable linear regression models.

Results.: Men had higher levels of hemoglobin and s-ferritin than women. Hemoglobin was associated with wider retinal venules in both men and women (mean difference between highest compared to lowest sex-specific quartile: men 5.99 μm, P = 0.001; women 7.28 μm, P < 0.001). S-ferritin was associated with wider retinal venules in men but not women, independent of traditional risk factors (mean difference between highest compared to lowest sex-specific quartile: men 4.21 μm, P = 0.013; women −0.21 μm, P = 0.89). The association in men was attenuated, but still significant, with adjustment for hemoglobin. S-ferritin and level of hemoglobin were not associated with arteriolar caliber in either sex. Transferrin saturation was not associated with retinal vascular caliber.

Conclusions.: Level of hemoglobin was associated with wider retinal venules in both sexes while s-ferritin was associated with wider retinal venular caliber only in men. Men have higher levels of stored iron, and this may underlie the observed sex difference in the association between s-ferritin and retinal vascular caliber.

Introduction

In 1981, Sullivan ¹ launched the hypothesis that stored body iron is involved in the pathogenesis of atherosclerosis and that chronic iron depletion in menstruating women could contribute to the reduced risk of cardiovascular disease in premenopausal women. S-ferritin concentration is correlated to amount of stored iron. ² Iron is closely bound to ferritin in an inactive form while it may pathologically be liberated in an active noxious form, inducing oxidative damage and oxidation of low-density lipoprotein (LDL) cholesterol, which is a proatherogenic factor. ^3–6 Animal studies have shown that iron overload is associated with oxidative stress and atherosclerosis. ^7,8 In humans, s-ferritin level has been demonstrated to be directly correlated to oxidant damage of atherosclerotic plaques. ⁹ However, epidemiologic findings have been conflicting on whether iron status is correlated to incident cardiovascular disease. ^5,10–18  

The retinal vasculature is part of the microcirculation and is accessible through noninvasive high-quality imaging and software-assisted grading methods with high reproducibility. ^19–21 Wider retinal venular caliber has been linked to endothelial dysfunction and inflammation and has been demonstrated to independently predict ischemic heart disease and stroke. ^22–25  

In a cross-sectional study, level of hemoglobin was associated with wider retinal arterioles and venules; and proposed mechanisms were blood viscosity and systemic factors related to oxygenation, for example, lung disease and smoking. ²⁶ Iron body store is correlated to level of hemoglobin and may possibly be an additional pathway underlying the correlation between retinal vessel caliber and level of hemoglobin. The relationship between markers of iron body stores and retinal vascular caliber has not been explored in other cohorts, and we wanted to investigate whether s-ferritin, transferrin saturation, and hemoglobin were associated with retinal vascular caliber measured 6 years later. 

Methods

Study Sample

The Tromsø Study is a population-based multipurpose longitudinal study from the municipality of Tromsø, Norway, started in 1974 and based on the Population Registry of Norway. Six surveys have been carried out so far and are referred to as Tromsø 1 through 6. The present study is based on data from Tromsø 5, conducted in 2001–2002, and Tromsø 6, conducted in 2007–2008, with an attendance rate of 78.5% and 65.7%, respectively. Both surveys consisted of two visits. The study design and methods are described in detail elsewhere. ^19,27  

Eligible for the present study were all who participated in Tromsø 5 and who were examined with retinal photography in Tromsø 6. Of the 8130 participants in Tromsø 5, a total of 3309 had measurements of retinal vascular caliber in Tromsø 6. We excluded those with images of poor quality (n = 313) and those who were pregnant in either Tromsø 5 or 6 (n = 3), leaving 1819 women and 1174 men to be included in this study. The study adhered to the tenets of the Declaration of Helsinki and was approved by the Data Inspectorate of Norway and Regional Committee for Medical and Health Research Ethics. 

Assessment of Body Iron Stores, Hemoglobin, and Cardiovascular Risk Factors

Data were collected from questionnaires, laboratory testing, and physical examinations. Nonfasting blood samples were obtained by venipuncture and were analyzed at the Department of Clinical Chemistry, University Hospital of North Norway in Tromsø. Blood samples in Tromsø 5 were drawn at the first visit, except for samples for hemoglobin and high-sensitivity C-reactive protein (hsCRP) drawn at the second visit a few weeks later (Fig. 1). In Tromsø 6, all blood samples were drawn at the first visit. S-ferritin and transferrin saturation were measured in Tromsø 5 only, while hemoglobin was measured in both surveys. Serum for analyses on ferritin, iron, and transferrin was stored at −70°C for a maximum of 12 months before analyses on a Hitachi 917 analyzer (Boehringer, Ingelheim, Germany). Iron was measured by the ferrozine method, and s-ferritin and transferrin were measured by a turbidimetric assay. In an effort to harmonize s-ferritin levels within Norway at the time, the laboratory used a factor of 0.82 for s-ferritin analysis. ²⁸ S-transferrin was reported in grams per liter, and s-TIBC (total iron-binding capacity) was calculated as s-TIBC μmol/L = 25.1 × s-transferrin. Transferrin saturation (%) was calculated as 100 × (s-iron/s-TIBC). Hemoglobin was measured with an automated blood cell counter (Coulter Counter; Beckman Coulter, Inc., Brea, CA) within 24 hours. ²⁹ High-sensitivity CRP was measured by a particle-enhanced immunoturbidimetric assay from Roche (Mannheim, Germany). Serum total cholesterol and high-density lipoprotein (HDL) cholesterol were measured with an enzymatic colorimetric test (Modular PTM; Roche). Glycosylated hemoglobin (HbA1c) was analyzed with Bayer DCA 2000 (Bayer AG, Leverkusen, Germany) in Tromsø 5 and with Variant II (Bio-Rad Laboratories, Inc., Hercules, CA) in Tromsø 6. Blood pressure was measured three times at 1-minute intervals on one arm after 2 minutes of seated resting with the use of an automatic device (Dinamap Vital Signs Monitor; Critikon, Tampa, FL), and the mean of the two last measurements was used in the present analyses. Weight and height were measured with the participant wearing light clothes and without shoes, and body mass index was calculated as kilograms per square meter. Smoking status was dichotomized as current daily smoking versus not. Hypertension was defined as systolic blood pressure >140 mm Hg and/or diastolic blood pressure >90 mm Hg and/or use of antihypertensive drugs. Diabetes was defined as self-reported diabetes and/or use of diabetes medication and/or HbA1c ≥6.5%. Cardiovascular disease was defined as self-reported myocardial infarction and/or stroke. Anemia was defined as hemoglobin <12.0 g/dL in women and <13.0 g/dL in men. ³⁰ Women were classified as premenopausal if they were <40 years, did not use hormone replacement therapy (HRT), and did not state that they were postmenopausal. Women aged 40 to 55 years, not using HRT, and still menstruating were classified as premenopausal. Women who were 55 years or older and women aged 40 to 55 years stating that they were postmenopausal or using HRT were classified as postmenopausal. 

Figure 1. 

 

Timeline showing when the different examinations were performed. CVD risk factors include blood pressure, smoking, body mass index, glycosylated hemoglobin, total and high-density lipoprotein cholesterol, age. Transf.sat, transferrin saturation.

Measurement of Retinal Vascular Caliber and Refraction

Eye measurements were performed in Tromsø 6 (Fig. 1). Refraction was measured by Nidek AR 660A autorefractor (Nidek Co., Ltd., Gamagori, Japan). Retinal photography was performed with a Visucam PRONM (Carl Zeiss Meditec, Jena, Germany) digital retinal camera. Five-field 45° retinal color photographs with resolution 2196 × 1958 pixels were taken on both eyes using the camera preset internal fixation, with the disc-centered image taken as the second image on each eye. Retinal vessel calibers were measured using computer-assisted IVAN software, the updated version of Retinal Analysis (University of Wisconsin, Madison, WI), and details are described elsewhere. ¹⁹ The optic disc–centered image of one eye, preferably the right eye, was selected for the measurements, while measurement of the left eye was done only if that of the right eye was not usable. The diameters of all vessels coursing through the area of one-half to one disc diameter from the optic disc margin were measured, and the six widest of each type were summarized as the central retinal artery equivalent (CRAE) and the central retinal vein equivalent (CRVE). ³¹ Image quality was marked as poor when the grader considered that the image quality was likely to affect the measurements. Reliability testing (temporal drift, intra- and intergrader) gave intraclass correlation coefficients (ICC) above 0.93 for all of the tests. ¹⁹  

Statistical Methods

Data for s-ferritin were missing in 5%, transferrin saturation in 4%, and hemoglobin in 16% of the participants (Supplementary Material). To increase power, missing data were imputed using multiple chained imputation in STATA/MP 12.0 (StataCorp LP, College Station, TX) separately for women and men (Supplementary Material). Statistical tests were performed by mi estimate in STATA. 

Possible interaction between sex and markers of body iron stores and hemoglobin was imputed with the JAV approach (“just another variable”). ³² Due to significant interaction between sex and s-ferritin (P = 0.035) and substantially lower levels of s-ferritin and hemoglobin in women than in men, all analyses were stratified by sex. 

The associations between the markers of iron status and hemoglobin with retinal vascular calibers were analyzed using linear regression models, with retinal vascular caliber measured in Tromsø 6 as the dependent variable and levels of s-ferritin, transferrin saturation, and hemoglobin measured in Tromsø 5 as independent variables in separate analyses. The independent variables were analyzed both as continuous variables and quartiles entered as dummy variables, treating the lowest quartile as the reference category. S-ferritin was log-transformed to approximate normal distribution. Nonlinear associations between log-transformed s-ferritin, hemoglobin, transferrin saturation, and retinal vascular calibers were analyzed in age-adjusted first- and second-degree fractional polynomial models with fracpoly in STATA. These analyses included complete cases only, as nonlinear analyses are not supported by mi estimate in STATA. As there was evidence of a nonlinear relationship between the untransformed s-ferritin levels and CRVE in men, we further compared CRVE in the two highest to the two lowest quartiles of s-ferritin. We examined three multivariable adjusted models. Model 1 included adjustment for traditional cardiovascular risk factors (age, systolic and diastolic blood pressure, smoking, body mass index, HbA1c, and total and HDL cholesterol). Model 2 included adjustment for hemoglobin or s-ferritin, as appropriate, in addition to all the variables in model 1. Model 3 included adjustment for hsCRP in addition to the variables in model 1. As hemoglobin was measured in both surveys, additional analyses were made with levels from Tromsø 6 and the mean of the levels from Tromsø 5 and 6 as independent variables. In these models, we used covariates assessed in the same survey as hemoglobin (Tromsø 5, 6 or mean of Tromsø 5 and 6). Refraction and axial length impose a magnification effect on images, and we performed supplementary analysis with additional adjustment for the spherical equivalent refraction. We also performed supplementary analyses of possible interaction with prevalent hypertension, diabetes, and cardiovascular disease in Tromsø 5, respectively, using linear passive approach. ³² Supplemental analyses were performed with model 1. P values < 0.05 were considered significant. 

Results

Participant characteristics are shown in Table 1. Estimated means including imputed data (not shown) gave only slight differences from data on the complete cases. 

Table 1. 

 

Participant Characteristics at Baseline (2001–2002) and Retinal Vascular Caliber at Follow-up (2007–2008), by Sex*: The Tromsø Study

	Women	Men
Age, y (SD)	60.7 (9.4)	61.5 (9.2)
Systolic blood pressure, mm Hg (SD)	138.2 (21.5)	139.6 (19.6)
Diastolic blood pressure, mm Hg (SD)	79.6 (11.9)	82.2 (11.4)
Body mass index, kg/m2 (SD)	26.9 (4.4)	27.2 (3.4)
Total cholesterol, mmol/L (SD)	6.4 (1.2)	6.2 (1.1)
HDL cholesterol, mmol/L (SD)	1.6 (0.4)	1.3 (0.4)
HbA1c, % (SD)	5.4 (0.6)	5.5 (0.7)
Current daily smoking % (n)	25.9 (468)	23.1 (271)
C-reactive protein, mg/L (SD)	3.1 (6.2)	3.0 (6.3)
Ferritin, μg/L†	65 (62)	104 (108)
Transferrin saturation, % (SD)	25.7 (9.3)	27.3 (9.1)
Hemoglobin, g/dL (SD)	13.5 (0.9)	14.6 (1.0)
Anemia % (n)	3.5 (63)	3.1 (36)
Hypertension % (n)	50.3 (905)	53.0 (617)
Cardiovascular disease % (n)	3.8 (66)	11.2 (131)
Diabetes % (n)	4.5 (80)	5.8 (68)
CRAE, μm (SD)	139.9 (14.4)	137.0 (13.9)
CRVE, μm (SD)	209.8 (21.1)	210.0 (20.7)

*  Based on complete cases.

†  Median (interquartile range).

Log-transformed s-ferritin in Tromsø 5 was linearly associated with increased CRVE in Tromsø 6 in men (Fig. 2, Table 2). Adjustment for hemoglobin attenuated the estimates, while they were unchanged with adjustment for hsCRP. Figure 3 shows the unadjusted means of CRAE and CRVE for each quartile of untransformed s-ferritin in women and men. In men, level of s-ferritin in the highest quartile was significantly associated with wider CRVE compared to the lowest quartile, independent of cardiovascular risk factors (Table 2). The estimated difference in CRVE when the two highest were compared with the two lowest quartiles of s-ferritin was 3.02 μm (95% confidence interval [CI]: 0.62–5.41 μm, P = 0.014), adjusted for traditional cardiovascular risk factors and hemoglobin (model 2). In women, level of s-ferritin was not associated with CRVE in either analysis (Table 2). Eighty-nine percent of the women were postmenopausal in Tromsø 5. In analyses stratified by menopausal status, there was no significant association between s-ferritin and retinal vascular caliber in either group (estimated difference in CRVE pr. 1 unit change in log-transformed s-ferritin [95% CI]: 1.54 [−1.54 to 4.63] μm in premenopausal women, −0.04 [−1.53 to 1.44] in postmenopausal women). In analysis comparing CRVE in women with level of s-ferritin above 167 μg/L (equals fourth quartile in men) (n = 109) with women with s-ferritin either ≤167 or ≤104 μg/L gave nonsignificant thinner venular caliber in the high-level s-ferritin group (results not shown). Level of s-ferritin was not associated with CRAE in either women nor men (Table 3). Additional adjustment for spherical equivalent refraction tended to slightly attenuate the estimates and narrow the standard errors, though it did not essentially change the results (results not shown). 

Figure 2. 

 

Scatter plots of s-ferritin (log-transformed) versus CRVE and CRAE, by sex.

Figure 3. 

 

Means of CRVE and CRAE in sex-specific quartiles of s-ferritin, by sex. Spikes indicate the standard errors.

Table 2. 

 

Association Between Baseline Levels of s-Ferritin, Hemoglobin, and Transferrin Saturation With Retinal Venular Caliber (CRVE) at Follow-up in Men and Women*: The Tromsø Study 2001–2008

Men	Quartiles of s-Ferritin, μg/L				s-Ferritin, μm/L, Log-Transformed
	≤59	60–104	105–167	≥168	Per 1 Unit Change	P
Model 1	0.00, reference	−1.32 (−4.62 to 1.99)	1.65 (−1.66 to 4.95)	4.21 (0.87 to 7.54)†	1.77 (0.25 to 3.29)	0.022
Model 2	0.00, reference	−1.88 (−5.19 to 1.44)	0.99 (−2.33 to 4.31)	3.19 (−0.19 to 6.56)	1.22 (−0.33 to 2.78)	0.12
Model 3	0.00, reference	−1.32 (−4.63 to 1.98)	1.61 (−1.70 to 4.92)	4.16 (0.82 to 7.50)†	1.74 (0.22 to 3.26)	0.025
Men	Quartiles of Hemoglobin, g/dL				Hemoglobin, g/dL
	≤14.0	14.1–14.7	14.8–15.3	≥15.4	Per 1 Unit Change	P
Model 1	0.00, reference	−0.09 (−3.63 to 3.45)	5.38 (1.93 to 8.82)‡	5.99 (2.32 to 9.67)‡	2.50 (1.17 to 3.83)	<0.001
Model 2	0.00, reference	−0.25 (−3.80 to 3.29)	5.15 (1.69 to 8.61)‡	5.55 (1.83 to 9.28)‡	2.34 (0.99 to 3.69)	0.001
Model 3	0.00, reference	0.04 (−3.51 to 3.59)	5.58 (2.12 to 9.05)‡	6.22 (2.53 to 9.90)‡	2.63 (1.29 to 3.97)	<0.001
Men	Quartiles of Transferrin Saturation, %				Transferrin Saturation, %
	≤21	22–27	28–33	≥34	Per 10 Units Change	P
Model 1	0.00, reference	−2.24 (−5.44 to 0.95)	−0.58 (−3.85 to 2.69)	−1.05 (−4.42 to 2.32)	−0.39 (−1.68 to 0.90)	0.56
Model 3	0.00, reference	−2.17 (−5.37 to 1.03)	−0.47 (−3.76 to 2.81)	−0.94 (−4.32 to 2.44)	−0.34 (−1.64 to 0.96)	0.61
Women	Quartiles of s-Ferritin, μg/L				s-Ferritin, μm/L, Log-Transformed
	≤38	39–65	66–100	≥101	Per 1 Unit Change	P
Model 1	0.00, reference	0.97 (−1.81 to 3.75)	1.28 (−1.55 to 4.10)	−0.21 (−3.05 to 2.63)	0.41 (−0.90 to 1.72)	0.54
Model 2	0.00, reference	0.25 (−2.53 to 3.03)	0.21 (−2.63 to 3.06)	−1.39 (−4.26 to 1.49)	−0.33 (−1.67 to 1.02)	0.64
Model 3	0.00, reference	0.94 (−1.84 to 3.72)	1.29 (−1.54 to 4.11)	−0.21 (−3.05 to 2.63)	0.41 (−0.90 to 1.72)	0.54
Women	Quartiles of Hemoglobin, g/dL				Hemoglobin, g/dL
	≤12.9	13.0–13.5	13.6–14.1	≥14.2	Per 1 Unit Change	P
Model 1	0.00, reference	3.53 (0.67 to 6.39)†	4.21 (1.31 to 7.10)‡	7.28 (4.06 to 10.50)‡	3.08 (1.82 to 4.34)	<0.001
Model 2	0.00, reference	3.54 (0.68 to 6.41)†	4.22 (1.33 to 7.12)‡	7.32 (4.09 to 10.56)‡	3.11 (1.84 to 4.37)	<0.001
Model 3	0.00, reference	3.64 (0.77 to 6.50)†	4.31 (1.42 to 7.21)‡	7.42 (4.19 to 10.64)‡	3.14 (1.88 to 4.40)	<0.001
Women	Quartiles of Transferrin Saturation, %				Transferrin Saturation, %
	≤19	20–25	26–31	≥32	Per 10 Units Change	P
Model 1	0.00, reference	−0.45 (−3.11 to 2.22)	−0.30 (−3.01 to 2.42)	−1.62 (−4.44 to 1.21)	−0.48 (−1.53 to 0.57)	0.37
Model 3	0.00, reference	−0.32 (−2.99 to 2.35)	−0.17 (−2.89 to 2.55)	−1.51 (−4.34 to 1.32)	−0.45 (−1.50 to 0.60)	0.40

 Model 1: adjusted for age, systolic and diastolic blood pressure, smoking, body mass index, HbA1c, total and HDL cholesterol; model 2: model 1 + hemoglobin or ferritin as appropriate; model 3: model 1 + high-sensitivity C-reactive protein.

*  Linear regression with CRVE as dependent variable; numbers give difference in caliber in micrometers (95% confidence interval) by quartiles or per defined change in continuous variable.

†   P < 0.02.

‡   P < 0.005.

Table 3. 

 

Association Between Baseline Levels of s-Ferritin, Hemoglobin, and Transferrin Saturation With Retinal Arteriolar Caliber (CRAE) at Follow-up in Men and Women*: The Tromsø Study 2001–2008

Men	Quartiles of s-Ferritin, μg/L				s-Ferritin, μm/L, Log-Transformed
	≤59	60–104	105–167	≥168	Per 1 Unit Change	P
Model 1	0.00, reference	−0.94 (−3.18 to 1.30)	0.67 (−1.57 to 2.92)	1.51 (−0.74 to 3.76)	0.76 (−0.27 to 1.78)	0.15
Model 2	0.00, reference	−1.11 (−3.36 to 1.14)	0.47 (−1.79 to 2.73)	1.20 (−1.09 to 3.49)	0.59 (−0.45 to 1.64)	0.27
Model 3	0.00, reference	−0.94 (−3.18 to 1.30)	0.69 (−1.56 to 2.94)	1.53 (−0.72 to 3.79)	0.77 (−0.25 to 1.80)	0.14
Men	Quartiles of Hemoglobin, g/dL				Hemoglobin, g/dL
	≤14.0	14.1–14.7	14.8–15.3	≥15.4	Per 1 Unit Change	P
Model 1	0.00, reference	0.17 (−2.24 to 2.58)	0.48 (−1.87 to 2.84)	2.39 (−0.13 to 4.90)	0.78 (−0.13 to 1.69)	0.09
Model 2	0.00, reference	0.12 (−2.30 to 2.53)	0.40 (−1.97 to 2.77)	2.24 (−0.33 to 4.80)	0.72 (−0.21 to 1.65)	0.13
Model 3	0.00, reference	0.15 (−2.27 to 2.57)	0.44 (−1.93 to 2.81)	2.35 (−0.19 to 4.88)	0.76 (−0.16 to 1.68)	0.11
Men	Quartiles of Transferrin Saturation, %				Transferrin Saturation, %
	≤21	22–27	28–33	≥34	Per 10 Units Change	P
Model 1	0.00, reference	−0.31 (−2.47 to 1.84)	1.22 (−0.97 to 3.40)	0.63 (−1.64 to 2.90)	0.31 (−0.56 to 1.18)	0.49
Model 3	0.00, reference	−0.34 (−2.50 to 1.82)	1.18 (−1.01 to 3.37)	0.59 (−1.69 to 2.87)	0.29 (−0.59 to 1.16)	0.52
Women	Quartiles of s-Ferritin, μg/L				s-Ferritin, μm/L, Log-Transformed
	≤38	39–65	66–100	≥101	Per 1 Unit Change	P
Model 1	0.00, reference	0.28 (−1.59 to 2.15)	2.32 (0.41 to 4.22)†	0.44 (−1.49 to 2.36)	0.48 (−0.40 to 1.37)	0.28
Model 2	0.00, reference	0.16 (−1.72 to 2.04)	2.14 (0.21 to 4.08)†	0.24 (−1.71 to 2.20)	0.36 (−0.55 to 1.27)	0.44
Model 3	0.00, reference	0.26 (−1.61 to 2.13)	2.32 (0.42 to 4.23)†	0.44 (−1.49 to 2.36)	0.48 (−0.40 to 1.37)	0.28
Women	Quartiles of Hemoglobin, g/dL				Hemoglobin, g/dL
	≤12.9	13.0–13.5	13.6–14.1	≥14.2	Per 1 Unit Change	P
Model 1	0.00, reference	1.05 (−0.91 to 3.01)	0.90 (−1.11 to 2.90)	1.06 (−1.14 to 3.27)	0.60 (−0.27 to 1.47)	0.18
Model 2	0.00, reference	1.04 (−0.92 to 3.00)	0.89 (−1.12 to 2.89)	1.04 (−1.17 to 3.26)	0.59 (−0.28 to 1.47)	0.18
Model 3	0.00, reference	1.10 (−0.87 to 3.06)	0.95 (−1.06 to 2.95)	1.13 (−1.08 to 3.34)	0.63 (−0.24 to 1.50)	0.16
Women	Quartiles of Transferrin Saturation, %				Transferrin Saturation, %
	≤19	20–25	26–31	≥32	Per 10 Units Change	P
Model 1	0.00, reference	0.00 (−1.79 to 1.79)	−0.07 (−1.91 to 1.77)	−0.01 (−1.92 to 1.90)	−0.32 (−1.02 to 0.39)	0.38
Model 3	0.00, reference	0.07 (−1.72 to 1.87)	0.00 (−1.85 to 1.85)	0.05 (−1.86 to 1.96)	−0.31 (−1.01 to 0.40)	0.40

*  Linear regression with CRAE as dependent variable; numbers give difference in caliber in micrometers (95% confidence interval) by quartiles or per defined change in continuous variable.

†   P < 0.05.

Transferrin saturation in Tromsø 5 was not associated with CRVE (Table 2) or CRAE (Table 3) in Tromsø 6. 

Hemoglobin in Tromsø 5 was independently linearly associated with increased CRVE in Tromsø 6 in both men and women (Fig. 4, Table 2), while there was no significant association with CRAE in either sex (Fig. 4, Table 3). Analyses using the hemoglobin level measured in Tromsø 6 or mean value of hemoglobin measured in Tromsø 5 and 6 gave essentially the same results (Supplementary Material, Supplementary Table S1). There was no interaction between sex and hemoglobin in the association with retinal vascular caliber. Anemia was not significantly associated with retinal vascular calibers (results not shown). 

Figure 4. 

 

Scatter plots of hemoglobin versus CRVE and CRAE, by sex.

In men, the effect of hemoglobin on vascular caliber was stronger in participants without hypertension. This was evident in both venular caliber (5.70 [95% CI: 3.44–7.95] μm per 1 unit change in hemoglobin, P < 0.001, P for interaction = 0.003) and arteriolar caliber (2.24 [95% CI: 0.74–3.73] μm per 1 unit change in hemoglobin, P = 0.003, P for interaction = 0.041). There was no significant interaction between prevalent hypertension and markers of body iron stores or between diabetes and cardiovascular disease and markers of body iron stores and hemoglobin, respectively. 

Discussion

Our main finding was that in men, a higher level of s-ferritin was independently associated with wider retinal venular caliber measured 6 years later. In women the level of s-ferritin was not associated with retinal vascular caliber. Further, we found that higher level of hemoglobin was independently associated with wider retinal venular caliber in both sexes. Transferrin saturation was not associated with retinal vascular caliber. To the best of our knowledge, this is the first time that an association between s-ferritin and retinal venular caliber has been shown in a population-based study. 

Since the first publication in 1992 of a positive association between s-ferritin and risk of incident myocardial infarction, ¹⁸ the relationship between iron stores and incident cardiovascular disease and mortality has been investigated in several studies, with conflicting results. ^5,10–18 Wider retinal venular caliber has been demonstrated to independently predict ischemic heart disease and stroke and has been linked to endothelial dysfunction and inflammation. ^22–25 A possible mechanism for the association between s-ferritin and venular caliber in men is through endothelial dysfunction. This is supported by a study on blood donors, in which high-frequency donors had decreased body iron stores and enhanced flow-mediated vasodilatation, a marker of endothelial function, compared to low-frequency donors. ³³  

Iron stores remain stable throughout adulthood in men, except for a decline in the oldest age groups, while iron stores are lower in premenopausal women than in men but increase after the onset of menopause. ^34,35 Apparently s-ferritin levels ≥ 105 μg/L (the two highest quartiles) drive the association in men. Women with s-ferritin levels > 100 μg/L (fourth quartile) had slightly thinner venular caliber than those in the lower quartiles, though not significantly. Rather than hormonal influences, we hypothesize that this sex difference is related to higher lifetime iron stores in men than in women. 

Level of s-ferritin changes with age, infections, states of inflammation, blood loss (bleeding and blood donation), and intake of food and supplements. The age-related changes imply a need for age adjustment in the analyses. ^34–36 In recent years there has been evolving evidence of the importance of chronic low-grade inflammation in the pathogenesis of cardiovascular disease. ^37,38 In vitro studies have demonstrated that iron may be involved in oxidation of LDL cholesterol, ^39,40 which acts as a proinflammatory factor in atherosclerosis. ^3,4,6 In a randomized clinical trial it was demonstrated that reduction in body iron stores by venesection increased the oxidation resistance of serum VLDL/LDL. ⁴¹ Markers of inflammation, in particular CRP, are associated with retinal venular widening independent of other risk factors. ^42,43 We found that adjustment for hsCRP did not change the estimates for the association between s-ferritin and venular caliber, suggesting that the association is not solely reflecting a possible relationship between low-grade inflammation and microvascular changes. 

Level of s-ferritin was associated with retinal venular caliber even with adjustment for level of hemoglobin, though it was attenuated. This indicates that part of the effect of s-ferritin on retinal venular caliber is independent of the effect of hemoglobin. 

In a cross-sectional study, hemoglobin level was positively associated with retinal venular caliber, and to a lesser degree, arteriolar caliber. The proposed mechanisms were blood viscosity and factors related to oxygenation. ²⁶ We equally found that hemoglobin was independently associated with venular widening measured 6 years later, in both men and women, but not with arteriolar caliber. The association between hemoglobin and venular caliber was only slightly attenuated with adjustment of s-ferritin. 

The effect of hemoglobin on retinal vascular caliber was stronger in men without hypertension. Hypertension is associated with narrower vessels and increased vascular resistance, while part of the microvascular effect of hemoglobin may be related to the effect of viscosity. Possibly the observed increased effect of hemoglobin in nonhypertensive men may be related to viscosity with increased effect in vessels having less vascular resistance due to arteriosclerosis. 

A limitation of this study is the lack of retinal vascular measurements in Tromsø 5, which prohibits inferences about temporal and causal relations. The population-based design with a large number of participants and sufficient power for sex-specific analyses, the high attendance rates, and the use of computer-assisted methods with high reproducibility are important strengths of this study. 

In conclusion, we found that level of s-ferritin was associated with wider retinal vessel caliber 6 years later in men, independent of traditional cardiovascular risk factors, level of hemoglobin, and markers of inflammation. A possible mechanism is endothelial dysfunction, but the underlying pathological process is unknown. Retinal vascular caliber is a subclinical measure and may possibly contribute to refined risk evaluation and targeted therapies. Future studies should investigate how s-ferritin and hemoglobin are related to change over time in retinal vascular caliber, as well as how change over time in s-ferritin and hemoglobin is related to retinal vascular caliber. The clinical relevance of the demonstrated associations remains to be investigated. 

Supplementary Materials

pdf S1 

Acknowledgments

Supported by The North Norway Regional Health Authority and Norwegian Ophthalmological Society. The authors alone are responsible for the content and writing of the paper. 

Disclosure: T. von Hanno, None; G. Bertelsen, None; A.R. Broderstad, None; T. Wilsgaard, None; E.B. Mathiesen, None