Menopausal hormone therapy, blood thrombogenicity, and development of white matter hyperintensities in women of the Kronos Early Estrogen Prevention Study

Supplemental Digital Content is available in the text

W hite matter hyperintensities (WMH) on T2weighted magnetic resonance imaging (MRI) are associated with ischemic small vessel disease 1,2 and may precede exhibition of mild cognitive impairment. [3][4][5] Conventional cardiovascular risk factors (ie, age, hypertension, smoking, hyperlipidemia) are proposed to be risk factors for development of WMH, especially in elderly persons. 5,6 Other factors not traditionally considered as cardiovascular risk factors, specifically, thrombogenic microvesicles (MVs), have, however, been implicated in microvascular disease contributing to the formation of WMH. [7][8][9] Thrombogenic microvesticles are blood-borne, cell membrane-derived vesicles that carry surface markers of the cell of origin, as well as phospholipids and proteins, associated with coagulation and inflammation. 10 In recently menopausal women participating in the Kronos Early Estrogen Prevention Study (KEEPS) who were at low risk for cardio-and cerebrovascular disease as defined by a rigorous set of exclusion and inclusion criteria, 11 WMH increased over the course of the 4 years of the study. 12 An exploration of nontraditional risk factors that might affect cerebral blood flow and perhaps cerebral microvascular permeability suggested that the thrombogenicity and proinflammatory state of the blood, defined by activated circulating platelets and platelet-derived MVs at the time women enrolled in the study, that is before randomization to treatment, associated with development of WMH over the study course. 12 KEEPS participants were randomized to either oral conjugated equine estrogen (oCEE), transdermal 17b-estradiol (tE2) both with pulsed progesterone, or placebo pills and patch (PBO) over the 4 years of the study. 11 These formulations of menopausal hormone treatments (MHTs) were shown to affect platelet secretory products, reactivity, and aggregation. [13][14][15] Therefore, the aim of this analysis was to determine whether the type of MHT modified the association of WMH with thrombogenicity of the blood defined by a set of markers of platelet function and reactivity, intravascular cells, cell-derived MVs, endothelial activation, and inflammation.

Participants
Women enrolled in an ancillary MRI study of the KEEPS (NCT000154180) at Mayo Clinic were eligible for this study. KEEPS was a double-blind, placebo-controlled study to determine the effects of two different hormonal treatments on progression of atherosclerosis defined by increases in carotid intima-medial thickness in recently menopausal women. 11 In brief, women were between 42 and 59 years old and within 6 months to 3 years past their last menses at the time of enrollment. Women were excluded if they had a coronary artery calcium score of more than 50 Agatston Units, smoked over 10 cigarettes per day, had body mass index more than 35 kg/m 2 , had a history of cardiovascular disease, or had low-density lipoprotein (LDL) cholesterol higher than 190 mg/dL, triglycerides higher than 400 mg/ dL, diagnosis of diabetes, uncontrolled hypertension (systolic blood pressure >150 mm Hg and/or diastolic blood pressure >95 mm Hg) or current or recent (6 months) use of cholesterol lowering medications (statins, fibrate, or >500 mg/day niacin). Women were randomized to: oCEE (Premarin, 0.45 mg/ day); transdermal tE2 (Climera, 50 mg/day); or PBO pills and patch. Micronized progesterone was given orally (Prometrium; 200 mg/day) for 12 consecutive days each month to both active MHT groups. 11 The study was approved by the Mayo Clinic Institutional Review Board and all participants gave written informed consent.

Brain imaging
Women underwent MRI at baseline before randomization, and at 18, 36, and 48 months for measurement of WMH volumes on fluid-attenuated inversion recovery sequences as previously described. 16,17 Blood collection and analysis Women were asked to refrain from aspirin 2 weeks before blood collection. Fasting venous blood was collected into a syringe through a 19 gauge butterfly needle and dispensed into plastic tubes containing anticoagulants needed for each assay and maintained at 338C until processed within 30 minutes. 18 Platelet activation using this technique and as measured by surface expression of P-selectin and fibrinogen receptors is less than 5%. 19 Collections occurred at baseline before randomization to study treatments and at each study time point. 18 Total cholesterol, LDL cholesterol, and highdensity lipoprotein cholesterol, triglycerides, blood glucose, and 17b-estradiol were measured by Kronos Science Laboratories (Phoenix, AZ) and the Mayo Clinic Department of Laboratory Medicine and Pathology (Rochester, MN).
Platelet count was determined by Coulter counter, as previously described. 20 Expression of activated platelet membrane P-selectin and glycoprotein IIb/IIIa complex binding to PAC-1 antibody (an indirect measure of the expression of membrane fibrinogen receptor) was measured by validated flow cytometry techniques. [18][19][20][21] Total numbers of thrombogenic (phosphatidylserine-positive defined by annexin V binding) MV and other MV staining positive for selected cell-specific markers were measured using fluorophore conjugated recombinant proteins and antibodies by flow cytometry. 18,21,22 Statistical analysis Data reduction methods were used to achieve some degree of parsimony in these analyses so that the complexity could be reasonably supported by the sample size. Analysis of longitudinal WMH data was performed with a two-stage approach to first characterize the time-response profile and then assess differences in profiles across treatment. The goal of the first stage is to adequately summarize the repeat measurements into single WMH responses per person. For this, slope coefficients were computed using least squares regression in fitting each woman's time-response data with a linear equation. To minimize the impact of skewed values, we took the logarithm of WMH, with a constant of 1 added to the raw value, before modeling. One-sample Wilcoxon signed-rank tests for zero slopes were used to test for a significant change in WMH over 48 months separately in each group. The slopes were then assessed for treatment difference using the proportional odds ordinal logistic model, which provides a generalization of the Kruskal-Wallis test for pairwise testing. The results of the two-stage approach were compared with generalized least squares in which all the serial log-transformed data were analyzed jointly in a single model, and the correlation of the repeat measurements taken into account.
For each of the 14 cellular activation markers, repeat measurements on the same person were averaged across visits, and the resulting mean values were converted to normal scores based on their ranks. Scored dimensions of platelet reactivity and MVs were derived using principal component (PC) analysis on the 14 transformed variables. The multiple PCs were carried forward and analyzed for association with three-level treatment in multinomial logistic regression model (an extension of the binary logistic model for >2 unordered outcome categories), and for association with WMH response in a proportional odd ordinal logistic regression model while adjusting for randomized treatment group. P values are computed from the likelihood ratio x 2 statistic for the model that is due to the individual or multiple variables of interest, with the exception of pairwise treatment comparisons which are based on approximate Wald x 2 statistics.

RESULTS
Baseline characteristics of the 95 participants, a subset of the 118 KEEPS participants at the Mayo Clinic site for whom WMH data were available 12 did not differ across treatment group assignments except for smoking status (Table 1). WMH increased in all three groups over the 48 months of treatment (P < 0.001 each; Table 2). The extent to which WMH increased, as summarized by within-person slopes, differed across treatments (P ¼ 0.044), with pairwise comparisons indicating greater increases in oCEE than in PBO (P ¼ 0.011). Results from repeated measures modeling (model-predicted WMH at 48 months are shown by treatment group in Supplemental Table 1, http://links.lww.com/MENO/A510) were in reasonable agreement with the summary measure analysis, supporting the validity of the simpler two-stage approach,

HORMONES, THROMBOGENICITY, AND WMH
and with differences between oCEE and PBO reported in the original analysis (Tables 1 and 3 of reference 17). The average of each of the 14 intravascular cellular activation markers (6 platelet reactivity variables and 8 MV variables) over treatment follow-up is shown by treatment group in Table 3. Because of the multiplicity of variables, and to differences in directional change (increases or decreases compared to placebo) for each of the variables, we used PC analysis to reduce these dimensions to their most important components. The first five PCs were retained as they could explain most (62%) of the variability in the 14 standardized variables (Supplemental Table 2, http://links.lww. com/MENO/A511).
A global test of association with treatment for these five PCs approached significance (P ¼ 0.059), with partial tests revealing an overall group difference for PC 3 (P ¼ 0.006). PC 3 represents a contrast of platelet microaggregates, adenosine triphosphate secretion, basal expression of P-selectin, and fibrinogen receptor complex (PAC-1 binding) versus total platelet count and numbers of leukocyte-derived MV. The composite scores in oCEE were marginally to significantly higher compared to other groups (P ¼ 0.003 for oCEE vs tE2, and P ¼ 0.063 for oCEE vs PBO; Supplemental Table 3, http://links.lww.com/MENO/A512).
Using multivariable regression to test the joint influence of the PCs and treatment on the slope measure for WMH, the global contribution of all five PCs did not reach statistical significance (P ¼ 0.104). However, of the individual components PC 1 reflecting MV positive for expression of tissue factor, intracellular adhesion molecule 1 and vascular cell adhesion protein 1, and MVs derived from leukocytes and monocytes showed the most prominent effect (P ¼ 0.003). This finding indicates that, after controlling for treatment, the higher the composite score for PC 1 the greater the rate of increase in WMH. Also from this model, the association between treatment group and WMH increase persisted after adjustment for PC variables (P ¼ 0.009). Based on the global test of interaction on 10 degrees of freedom, there was no evidence (P ¼ 0.204) that the overall association between PCs and WMH increase differed by MHT (Table 4).

DISCUSSION
The results of this study support previous observations that blood thrombogenicity and proinflammatory status associate with WMH, 7-9,12 and extend those observations that this association may be influenced by factors other than the type and dose of menopausal hormones used for the treatment in the Kronos Early Estrogen Prevention Study.

JAYACHANDRAN ET AL
Factors influencing generalized inflammation and, indirectly, blood thrombogenicity, include conventional cardiovascular risk factors such as age, blood pressure, hyperlipidemia, insulin resistance, and life-style choices such as diet, activity, and smoking. In KEEPS, the conventional cardiovascular risk factors such as body mass index, blood pressure, triglycerides, HDL and LDL, glucose, and smoking status, however, did not associate with WMH at 48 months, which is consistent with other studies. 9,12,23,24 Other potential sources for inflammation in KEEPS participants are unclear. Only 5% of participants were current smokers, 12 but other behavior factors such as diet and activity were not analyzed, nor were potential sources of commensal or low-grade infectious or inflammatory conditions such as periodontal disease, asthma, or prior histories of hypertensive pregnancy disorders. 25,26 Each of these conventional and nonconventional risk factors may individually be insufficient to initiate an inflammatory response of the cells within the vascular compartment. Their collective effects of the endothelium, platelets, and monocytes may, however, reach a threshold to alter changes in the macrovasculature (carotid artery intima-media thickness) 18,27 and cerebral microvasculature affecting development of WMH.
In the present study, the overall association between treatment and the five PCs describing a number of cellular activity measures did not reach statistical significance at the P < 0.05 level. The PC 3 that represented a contrast of platelet reactivity measures, however, differed significantly in the oCEE compared to the tE2 or PBO groups. This result is consistent with previous findings of significant differences in platelet functions between tE2 and oCEE groups. 13,14,15 Effects of various genetic variants on the responses to treatment might also have masked potential treatment effects as genetic variants associated with metabolism and uptake of estrogen, with innate immunity, and with APOE e4 was observed for MHT effects measured by differences in chronological age for onset of menopause, in carotid artery intima-medial thickness, and deposition of b-amyloid in the brain. 9,28-32 In spite of these effects, after controlling for treatment, the overall association of the five PCs with increase in WMH reflected the strong positive correlation between PC 1 score and WMH increase. Taken together these results suggest that both MHT and the composite of the MV measurements explaining PC 1 show an independent effect on development of WMH.
There are several limitations of this study that should be considered. First, the results may not be applicable to the general population as the KEEPS enrolled recently menopausal women within a relatively narrow age range. In addition, these women were predominantly white, healthy, educated, and most were nonsmokers. The advantage of this homogenous population is that the findings may, however, reflect general physiological processes that are not confounded by manageable cardiovascular risk factors. Second, the influences of the MHT used in KEEPS on development of the WMH may not apply to other doses or formulations of MHT used in other studies. Third, the overall association between PCs and WMH increase did not apparently differ by MHT. However, the relatively small sample in our study may have limited the power to detect such a difference.

CONCLUSIONS
The findings of the present study are consistent with those of other investigations that implicate thrombogenicity of the blood and inflammation as contributors to development of WMH. 9,12,33 Activation of blood platelets, endothelium, and monocytes associated with development of WMH are most likely multifactorial including synergistic effects of conventional risk factors such as age, blood pressure, and components of metabolic syndrome. In addition, other potential sources of platelet and cellular activation such as effects of natural menopausal aging processes, adverse pregnancy histories, commensal infections, and comorbid inflammatory conditions and behaviors could have additive effects. Specific mechanisms by which these activated cells and MV affect cerebral microvascular function leading to formation of WMH remain to be determined.