While the prevalence of atherosclerosis is similar in women and men, women enjoy a 5-10 year lag in onset of cardiovascular events compared to men [1]. After menopause, a state associated with marked reduction in ovarian sex steroid production, the relative risk of events such as heart attack and stroke caused by atherosclerotic plaque rises up to threefold regardless of age range [1]. This has prompted numerous investigations of hormone therapy (HT) to lower cardiovascular risk to premenopausal levels. However, initiating HT in large, randomized trials has not realized a cardiovascular benefit [2-4].

Another change that occurs following menopause that may contribute to atherosclerosis risk is cessation of monthly iron loss [5]. The remarkable lack of atherosclerosis in patients with iron overloaded states like hereditary hemochromatosis (HH) [6] has been used to argue against the 'iron hypothesis'. Recent advances in understanding iron homeostasis have identified an important hepatically-synthesized protein called hepcidin-a crucial regulator of macrophage iron uptake; its near absence in HH may explain these patients' relative protection from atherosclerosis. Together, increased available iron and hepcidin may yield redox-active iron able to oxidize lipid and fuel plaque progression; thus, more informed studies of iron and perimenopausal atherosclerosis risk that address both are warranted.

Iron in tissues affects local magnetic properties that can be quantified using the magnetic resonance imaging (MRI) parameter T2*, which decreases with increased tissue iron content. We sought to design a study testing the hypothesis that changes in iron homeostasis fuel the development of atherosclerosis, and that increased vascular iron content in women explains the postmenopausal woman's loss of relative protection against atherosclerotic events.

Study objectives

The primary objectives of the present study are to 1) measure changes in arterial wall iron content in women entering menopause and 2) identify changes in the iron uptake signal hepcidin through perimenopause. The hypotheses driving this longitudinal study in a cohort of women entering menopause are the following: arterial wall iron increases in the transition from perimenopausal to postmenopausal status, carotid wall volume and dysfunction increases with increased arterial wall iron, and the macrophage iron uptake signal hepcidin increases after menopause in women at high risk of developing atherosclerosis.

Study design

This is an observational, prospective single-center cohort study approved by the Ohio State University Institutional Review Board, and all participants provide written, informed consent to participate.

Study population

Perimenopausal women, defined as women age 40 years or older with at least 1 but no more than 6 menstrual cycles in the 12 months preceding enrollment, with two or more of the following cardiovascular disease CVD risk factors are eligible: total cholesterol = 240 mg/dL, diastolic blood pressure =90 mmHg, systolic blood pressure =140 mmHg, current smoker, or history of diabetes. In addition, 15 perimenopausal women with zero or one CVD risk factors are enrolled as controls. To avoid confounding effects on iron homeostasis, patients with a history of renal dysfunction (GFR =40 ml/min/m².) or those unwilling to donate blood during the study will be excluded from enrollment. Excluded will be subjects with clinically-apparent atherosclerotic heart disease or peripheral arterial disease defined as prior acute coronary syndrome, exertional angina or equivalent, prior coronary intervention or bypass surgery, claudication or prior peripheral vascular intervention will be excluded. Enrollment criteria are summarized in Table 1.

Table 1: Inclusion and exclusion criteria.

Measurements

Recruited eligible subjects will undergo baseline, one year and two years follow-up laboratory evaluations that include clinical assessment (e.g. body mass index, blood pressure and waist/hip circumference ratio), serologies and noncontrast MRI. Blood pressure is measured in the right or left arm while the patient is resting on the scanner table post-MRI. In addition to recording medications, intake of supplemental iron or multivitamin containing iron will be quantified in mg/month derived from a standardized food frequency questionnaire [7] completed at the second MRI visit. Follow-up phone interviews at 6-month intervals will document interval occurrence of atherosclerotic endpoints defined as any of the following: coronary heart disease events (coronary death, myocardial infarction, and angina); cerebrovascular events (including stroke and transient ischemic attack); peripheral artery disease; or heart failure due to ischemic heart disease. Any reported events will be further documented by review of the medical records by a blinded physician adjudicator.

Serologies including creatinine, lipid levels, iron, total ironbinding capacity, ferritin and high-sensitivity C-reactive protein will be measured at each MRI visit. Bioavailable estradiol will be measured using estradiol RIA (Esoterix, Inc.) and sex hormone binding globulin assays [8] at baseline, 1-year follow-up and 2-year follow-up. Plasma hepcidin is measured at each MRI visit using hepcidin-25 EIA (Bachem Americas, Inc.).

MRI

All subjects are imaged on a 3 Tesla whole body scanner (MAGNETOM Verio, Siemens Healthcare; Erlangen, Germany) with maximum gradient amplitude of 45mT/m and a slew rate of 200 mT/m/ms. The system's integrated body coil will be used for radiofrequency signal transmission. Signal reception will be assured by a flexible 6-element phase array body coil in combination with the table embedded spine coil for liver imaging and by a custom built 8-element (4 left and 4 right) carotid coil for carotid imaging (Massachussetts General Hospital). All subjects undergo the same imaging protocol that includes the elements outlined in Table 2.

Table 2: MRI Protocol.

The multi-echo gradient echo images (Figure 1) are processed offline by drawing a region of interest (ROI) encompassing a nonvascular region of the liver for liver T2* (a well-validated noninvasive estimate of tissue iron content [9]) using semi-automated software developed in-house for carotid artery wall T2* measurement [10]. The in-plane velocity-encoded cine images (Figure 1) are processed offline using inhouse software that automatically detects the time (ms after R wave) of the foot of aortic velocity upslope at each pixel in the descending aorta, then computes pulse wave velocity from a linear fit of distance vs. time-to-peak [11].

Figure 1: Left panels: Images obtained at serial echo times (top: 1.9 ms, middle: 5 ms, bottom: 8.1 ms) from a tranverse plane abdominal multiecho gradient echo acquisition show the images used for liver T2* computation, a wellvalidated noninvasive technique to estimate liver iron content. In the absence of secondary causes of hepatic iron overload, liver iron content may be considered a suitable surrogate for total body iron. Middle panel: magnitude image showing a longitudinal section of the descending aorta. Right panel: velocity-encoded image showing in-plane velocities in the descending aorta, from which pulse wave velocity is calculated.

To insure reproducibility of vessel wall measurement location across serial MRI scans, each subject's T1-W TSE acquisition images are co-registered using 3D postprocessing software (FUSION Siemens Healthcare; Princeton, New Jersey).Multiplanar (MPR) images perpendicular to the arteries are generated at the same location on the co-registered data sets by one observer. Five consecutive 1 mm thickness slices are generated from each CCA and ICA, with the most cranial slice for CCA volumes located just below the carotid bulb and the most caudal slice for ICA volumes located just above the carotid bifurcation (Figure 2). Using the MPR images, two experienced observers measure lumen area and lumen plus wall areas by delineating the inner and outer border of the vessel wall. Vessel wall area is calculated by subtracting lumen area from the outer border enclosed area. Volume carotid wall for each of the four locations (right CCA and ICA, left CCA and ICA) will be calculated by multiplying the sum of the five slices wall areas with the slice thickness. A total carotid volume for a subject at a time point will be generated by summing the four wall volumes.

Figure 2: 3D noncontrast T1W-TSE MRI shows right (A) and left (B) carotid arteries along their lengths. Multiplanar reformatted sections for vessel wall volume are generated from the 3D acquisition over the planes shown in green; panels 1 through 5 below show representative sections for the right common carotid artery.

To calculate carotid distensibility, end systolic (As) and end diastolic (Ad) areas of the carotid lumen are measured off-line three times (Medviso AB; Lund, Sweden). All measurements for each artery are averaged and distensibility is calculated using the following formula: distensibility= [As-Ad)/As] ÷ (systolic BP - diastolic BP) in kPa^-1.

Sample size calculation

The primary outcome on which sample size is based is arterial wall T2* measurement. Using the preliminary data from four women with CVD risk factors who underwent baseline and 1-year followup arterial T2* measurements, we expect that the arterial wall T2* in perimenopausal women have a mean ranging from 25 ms to 53 ms and a standard deviation of 7 ms. Assuming a conservative Bonferroni adjustment to account for the two possible pairwise comparisons (baseline vs. year 1 and year 1 vs. year 2), we used a Type I error rate of .025 (.05/2=.025). The sample size was calculated based on a paired sample t-test (one-sided) with a power of 80%, a correlation of 0.6 and 2.0 ms as the smallest detectable difference in arterial wall T2*. When the sample size is adjusted to account for 9% of patients unable to undergo MRI due to claustrophobia or obesity and an additional 12% of patients not completing follow-up due to various reasons (dropout, death, etc.), the required sample size is 100 for perimenopausal women with CVD risk factors. Assuming a two-sided test at alpha = 0.05 and a coefficient of variation of 88% [12], a sample size of 100 cases and 15 controls will provide 91% power to detect a two-fold difference in serum hepcidin levels at two year follow-up.

Statistical analysis

Demographic, medical history and other baseline variables will be summarized as appropriate to the type of data. Summary statistics will include the number of observations available, mean, standard deviation, median, minimum, and maximum for continuous variables such as vessel wall T2*, vessel wall volume, and carotid artery distensibility. Descriptive statistics will also be calculated for binary variables such as smoker vs. nonsmoker, diabetes and statin drug use. Cardiovascular disease (CVD) event risk score is expressed as a number of points between -2 and 21 (with values below -2 assigned to -2, and values above 21 assigned to 21), which is computed based on the following clinical variables: age, gender, systolic blood pressure, diabetes Y/N, tobacco Y/N, total cholesterol, and HDL cholesterol level [13].

To examine the decrease of arterial wall T2* in women transitioning from perimenopausal to postmenopausal status, marginal models (GEE) will be used. GEE is a flexible approach to produce robust estimates of covariate effects after adjusting for the within-person correlation in outcomes between the three clinical assessments. It is well suited to this balanced longitudinal design with three repeated clinical assessments. Analyses will be conducted in SAS by fitting linear spline models with a common knot at the time of the second clinical assessment (year 1), which is helpful to assess the potential non-linear trends in the mean arterial wall T2* over time. The regression coefficients with respect to the assessment time will be used to test whether the T2* decreases over the transition period (from baseline to year 1 and from year 1 to year 2) as hypothesized due to carotid iron increase. In the marginal models (GEE), covariates will be included as potential confounders, such as serum estradiol level, statin drug use, BMI, CVD event risk score, etc. When sample sizes permit, some interaction terms will be explored in the modeling procedure.

To assess the association between carotid artery wall T2* and wall volume or distensibility, similar marginal models (GEE) will be used with a common knot at the time of the second clinical assessment (year 1). In this linear spline model, the outcome measure carotid artery wall T2* will be regressed onto wall volume and distensibility by adjusting for the assessment time with a knot at year 1 and potential confounders including serum estradiol level, hsCRP, statin drug use, BMI, dietary iron intake, and CVD risk score.

Analyses will be conducted in SAS by fitting linear spline models with a common knot at the time of the second clinical assessment (year 1), which is helpful to assess the potential non-linear trends in the mean plasma hepcidin level over time. The regression coefficients with respect to the assessment time will be used to test whether the outcomes increase over the transition period (from baseline to year 1 and from year 1 to year 2) as hypothesized. In the marginal models, covariates will be included as potential confounders, such as serum estradiol level, statin drug use, BMI, CVD event risk score, etc. When sample sizes permit, some interaction terms will be explored in the modeling procedure. For the marginal models, the normality assumption for the plasma hepcidin level will be assessed to determine whether a certain transformation of the outcome is needed. Our power analysis for detecting yearly differences in plasma hepcidin (from baseline to year 1 and from year 1 to year 2) indicates that we have a power of more than 93.9% to detect a true difference of 0.4 nM in plasma hepcidin using the above sample size.

We will also extend our marginal models for serum hepcidin level to include the control data. The extended models will include effects for group (case or control) and group-by-time interactions to allow the linear splines to differ by case/control status. Our primary analysis will be a comparison of serum hepcidin level between cases and controls at two year follow-up. As a secondary analysis, we will test for differences in the linear splines between cases and controls to determine if differences between the groups change over time (i.e., a test of time-by-group interaction). A similar approach will be used to compare vessel wall T2* measurements in cases and controls.

Ethical considerations

The study is conducted according to the principles of the Declaration of Helsinki. The study design, all research aims and the specific measurements in this trial have been approved by the Institutional Review Board, and only participants who provide written informed consent after sufficient time to review study information undergo study procedures.

Data management and privacy protection

Data are manually entered into an electronic database by study investigators and research staff from written case report forms and manually-processed with periodic audits by research staff to insure quality of manual data entry. Study data are analyzed using unique subject identifiers that do not enable identification of individual subjects.

While the primary focus of this work is to provide the design and rationale of our prospective clinical study, we have made considerable progress in enrollment to date. The characteristics of study participants are summarized in Table 3. Claustrophobia precluded completion of baseline evalution in 5 subjects. Data analysis is ongoing as the study hypotheses are based on changes over 2 years in serologies and imaging parameters. Description of baseline serologic studies in the cohort enrolled to date is provided in Table 4.

Table 3: Characteristics of Enrolled Participants to Date (N=100).

Table 4: Baseline Serologies of Enrolled Participants to Date (N=100).

The present study for the first time evaluates longitudinal changes in the arterial wall across the menopause transition that may be related to the cessation of monthly iron loss and may help explain the increased risk of atherosclerotic events seen in post- vs. pre-menopausal women. Having validated vessel wall T2* measurement as a reproducible, noninvasive map of arterial wall iron allows us to generate this evidence using serial evaluation of perimenopausal women at high risk of developing atherosclerosis. Arterial wall iron quantification with vessel wall volume and physiologic measures including carotid artery distensibility and aortic pulse wave velocity may reveal vascular abnormalities that precede clinically-evident atherosclerotic plaque.

Araujo et al. [14] showed that iron overload augments atheroma formation in hypercholesterolemic rabbits. Iron aggregates were seen in the arterial intima of the iron-overloaded rabbits suggesting that iron overload augments the formation of atherosclerotic lesions. In autopsy-derived coronary arteries and thoracic aortas, Yuan [15] found pronounced ferritin accumulation, intra-macrophage low-molecularweight iron and macrophage-derived foam cells in atherosclerotic lesions; normal arterial tissues lacked CD68 (macrophage identifier) and ferritin positivity. Yet for all of the mechanistic appeal of iron as a contributor to atherosclerosis, many studies of epidemiology, genetics and therapeutics to lower serum iron levels have been largely negative. Danesh et al. Published Date: a meta-analysis in 1999 looking at serum levels of ferritin, tranferrin saturation, total iron binding capacity and iron and found no predictive value of coronary heart disease risk [16]. Studies in patients with hemochromatosis have notably found less atherosclerosis in patients with genetically-induced increases in total body iron [6,17]. Studies using phlebotomy or blood donation to lower cardiovascular events have found that these methods lowered serum iron but did not affect cardiovascular events [18,19].

Our study design includes assessment of hepcidin, a circulating peptide and central regulator of iron metabolism that controls iron uptake and release by macrophages. Thus, studying changes in hepcidin levels in parallel with vessel wall measurements allows direct investigation of atherosclerosis substrate after monthly iron loss ceases. High hepcidin levels have been reported in patients with inflammation and certain anemias who have increased macrophage iron [26,27]. This may explain why epidemiologic studies ignoring the role of hepcidin have provided inconsistent results [16].

Thus, hepcidin represents a potentially important yet unexplored link between inflammation, a well-established risk factor of cardiovascular disease, and iron homeostasis. Missing from prior studies of iron and atherosclerosis risk to date have been (1) direct iron measurements in arterial wall in humans to corroborate the positive results of animal studies and (2) any consideration of hepcidin, without which iron cannot enter macrophages. By incorporating these 2 crucial components, our work may overcome the pitfalls that have plagued prior studies that have been unable to provide conclusive evidence relating iron and atherosclerosis.

The current trial is the first longitudinal study of perimenopausal changes with suitable clinical, serologic and imaging endpoints to measure the extent to which vascular changes that precede atherosclerosis development may be attributable to cessation of monthly iron loss. By using in vivo vessel wall iron measurement, we may identify novel mechanisms by which the gap in incident CVD between women and men rapidly closes after menopause. If arterial wall iron is shown to increase in women after menopause, we will have identified a truly novel target for prevention where current approaches directed solely at hormonal changes have fallen short in delivering meaningful cardiovascular risk reduction.

This study was funded by NIH R01 HL095563.