Biomarkers and the Risk of Stroke in the Women's Health Initiative

“…biomarkers of exposure in such important areas as dietary and physical activity epidemiology have the potential to usher in a new cycle of association analyses of enhanced reliability…”

The Editors of this special focus issue on stroke in women have asked me to provide a perspective on the role and potential of biomarkers in stroke risk assessment and prevention. This invitation undoubtedly arises from our studies of stroke in relation to postmenopausal hormone therapy (HT), among other exposures, in the context of the NIH-sponsored Women's Health Initiative (WHI), for which I have been a principal investigator of the Clinical Coordinating Center since the inception of the program in 1992.

“Biomarkers may also help to understand the biological processes that mediate a treatment or intervention effect on stroke risk.”

Biomarkers have several possible uses in stroke epidemiology and prevention research. For example, serum or plasma biomarkers may help to predict stroke risk, leading to risk monitoring and intervention activities. Plasma cholesterol fractions provide an important example. Such biomarkers may also be used in observational studies in an attempt to control confounding, and thus to enhance disease-association study reliability. Biomarkers may also help to understand the biological processes that mediate a treatment or intervention effect on stroke risk. Other biomarkers may be able to provide an objective assessment of potential stroke risk factors that are difficult to measure; for example, dietary and physical activity exposures that traditionally have been self-reported. Each of these possible uses of biomarkers will be expanded upon below, in the context of WHI studies, and the need and opportunity for further biomarker development will be discussed.

Postmenopausal HT & stroke incidence in WHI

Stroke incidence was among the major clinical outcomes used in the monitoring and reporting of the WHI randomized, placebo-controlled HT trials. A total of 10,639 women who were post-hysterectomy were randomized to 0.625 mg/day conjugated equine estrogens (Premarin) or placebo, and 16,608 women with a uterus were randomized to this same estrogen regimen plus 0.625 mg/day medroxyprogesterone acetate (Prempro) or placebo. These preparations were used by approximately 8 million and 6 million women, respectively, in the USA when the WHI trials began in 1993. All women were postmenopausal and in the age range of 50–79 years at randomization. Both HT trials were stopped prematurely; the estrogen plus progestin (E + P) trial in 2002 when it was judged that health risks exceeded benefits over a 5.6 year average follow-up period [1], and the estrogen-alone (E-alone) trial in 2004 after an average 7.1 years of follow-up, largely because of an elevation in stroke incidence of similar magnitude to that seen for E + P [2]. Specifically, the estimated hazard ratio (95% CI) for stroke was 1.31 (95% CI: 1.02–1.68) for women assigned to E + P compared with placebo [3], and 1.37 (95% CI: 1.09–1.73) for women assigned to E-alone versus placebo [4]. The elevated risk appeared to be confined to ischemic strokes, which constituted approximately 80% of incident strokes, with no evidence of hemorrhagic stroke risk elevation. The excess stroke risk appeared to be present in all subsets of women considered, including younger and more recently postmenopausal women.

The WHI also included a prospective cohort study among 93,676 postmenopausal women in the same age range, drawn from the same catchment populations as the clinical trials, with much commonality in data collection and outcome ascertainment. Specifically, data collection included baseline HT history ascertainment by trained interviewers assisted by a structured questionnaire and chart displaying colored photographs of various hormone preparations, followed by annual updates. These data provided an excellent opportunity to compare HT hazard ratios from a high-quality observational study to those from a gold-standard randomized controlled trial. The usual observational study analysis approach compares disease incidence between HT users and nonusers, while controlling for a list of potential confounding factors, which typically focus on available established risk factors for the disease in question. Data analyses may also calculate hazard ratio estimates according to important timing issues, which in this context, included time from menopause to first use of HT, and duration of HT use. When analyses of this type were carried out for E + P and E-alone in relation to such important clinical outcomes as coronary heart disease and breast cancer, there was good agreement between WHI cohort study and clinical trial results [5]. Not so, however, for stroke. For example, among adherent women, the stroke hazard ratios from the cohort study were lower by 54% (95% CI: 16–75) for E-alone, and by 67% (95% CI: 22–86) for E + P, when compared with the corresponding clinical trial hazard ratio estimates [5]. This suggests some serious residual confounding in the observational study analyses, which controlled for age, ethnicity, BMI, education, smoking history, physical functioning assessment, duration of prior use of postmenopausal HT (estrogen and E + P separately), history of treated diabetes, history of hypertension and history of high cholesterol requiring medication.

Although confounding factors for other treatments or exposures may differ from those for postmenopausal HT, these analyses suggest that there are important stroke risk factors beyond those included in the analyses just described, which may be needed for reliable observational study analyses. Knowledge of such risk factors could also provide novel insight into disease pathogenesis, as well as targets for stroke prevention efforts.

Biomarkers related to HT use & stroke risk

A case–control study within the HT trials was carried out to study coronary heart disease, stroke and venous thromboembolism in relation to blood biomarkers that had been reported to be associated with cardiovascular disease risk. The study also aimed to examine the extent to which post-treatment changes in these biomarkers could provide an explanation for observed HT effects on these diseases. The candidate biomarkers were identical for stroke and coronary heart disease and included: C-reactive protein (CRP), E-selectin, IL-6 and matrix metalloproteinase-9 as markers of inflammation; high-density lipoprotein cholesterol (HDL-C) and its subfractions HDL-2 and HDL-3, low-density lipoprotein cholesterol, lipoprotein(a), triglyceride and total cholesterol, as markers of blood lipids and lipoproteins; and a number of factors related to thrombosis, coagulation or fibrinolysis, including D-dimer, fibrinogen, (clotting) factor VIII, plasminogen activator inhibitor-1 antigen, plasmin–antiplasmin complex, thrombin activatable fibrinolysis inhibitor, von Willebrand factor and prothrombin F1–2. In addition, leukocyte count, platelet count, hematocrit and hemocysteine were assessed. Certain related genetic polymorphisms were also determined.

Many of these markers were found to relate to stroke incidence in the E + P and E-alone trial cohorts [6]. These include: all four of the inflammation markers; HDL-C and HDL-3, low-density lipoprotein cholesterol and triglyceride; D-dimer, factor VIII, thrombin activatable fibrinolysis inhibitor, von Willebrand factor and leukocyte count. Hence, there is quite a long list of blood-based markers that are relevant to the risk of stroke among postmenopausal women. Furthermore, several of these biomarkers changed significantly in the active HT compared with the placebo group, including inflammatory markers (i.e., CRP, E-selectin and matrix metalloproteinase-9), all the measured lipids, and some of the thrombosis markers (D-dimer, plasminogen activator inhibitor-1 antigen and plasmin–antiplasmin complex). However, when these biomarker changes were added to HT hazard ratio analyses, none of them appeared to mediate the E + P or E-alone effect on ischemic stroke risk [6]. It seems, at least superficially, that this rather detailed study provided little insight into the biological processes leading to the observed elevation in stroke risk among HT users.

Many of the markers of inflammation and hemostasis, along with a few others, have also been studied in relation to ischemic stroke risk in the WHI Observational Study. Rather similar results were obtained, with CRP emerging as strongly associated with disease risk [7]. This effort also included a focus on biomarkers that interact with or mediate HT effects, and lipoprotein-associated phospholipase A₂ was identified as a possible interactive factor [8].

We recently conducted a plasma proteomic discovery effort in an attempt to identify novel stroke risk factors [9]. This work employed an in-depth profiling strategy that entailed matching stroke cases arising in the WHI Observational Study to one-to-one matched controls, isotopic labeling of cases and corresponding control specimens, and extensive fractionation prior to tandem mass spectrometry. Owing to throughput limitations, plasma pools were formed from sets of 100 cases and from their corresponding matched controls, with eight such pool pairs. This work was able to quantify relative concentrations between stroke cases and controls for 366 proteins, including 47 that had concentration ratios that differed between cases and controls at a nominal p = 0.05 level of significance, compared with 18.3, which would be expected by chance. Three of these have estimated false discovery rates <0.05, while an additional 14 have estimated false discovery rates <0.20, providing a substantial set of candidates for further evaluation as stroke risk markers. To date, only one of these, IGF-binding protein (IGFBP)4 has been subjected to a validation study. Baseline IGFBP4 concentrations were determined by a standard ELISA for individual stroke cases and matched controls in the WHI HT trials. IGFBP4 concentrations were found to be 16.6% (p = 0.005) higher in cases versus controls, and yielded results in good agreement with the proteomic discovery data [9]. Corresponding validation exercises are planned for other proteins that were highly ranked in this discovery work. The list of candidates includes several plausible but unrecognized stroke risk markers, including, for example, IGFBP2, IGFBP6 and IGF2; suggesting an important role for the insulin-like growth factor pathway, prothrombin and antithrombin-related proteins and β-2 microglobulin, which was also confirmed to be a marker of coronary heart disease risk in these studies.

The same proteomic platform was also used to identify circulating proteins for which the concentration changed between baseline and 1 year following randomization among women assigned to active E + P [10] (50 women) or active E-alone [11] (50 women) in the WHI HT trials, this time using pools formed by baseline or 1-year specimens from ten women. A remarkable feature to emerge from this research effort was that there was evidence of change for nearly half (44.7%) of the 378 proteins quantified, many of which were recognized elements of coagulation, inflammation, immune response, metabolism, cell adhesion, growth factor or osteogenesis pathways. In fact, a majority of proteins for which there was evidence of a stroke association appeared to be affected by both E + P and E-alone. Interestingly, there were few proteins for which there was evidence of differential change from baseline to 1 year, between the two HT regimens [10].

The use of either E + P or E-alone was associated with an increase of approximately 20% in IGFBP4. The ELISA analysis previously mentioned in the HT trials suggests that a 20% increase in IGFBP4 is associated with an increase of approximately 40% in stroke risk (95% CI: 6–85). However, when we included the baseline and 1-year IGFBP4 concentration along with E + P or E-alone randomization indicator in data analysis, we found that there was little evidence that the E + P or E-alone effects on stroke were mediated by HT effects on this risk marker.

This experience, and that previously described for established inflammatory, lipid and thrombotic markers of stroke risk, is causing us to look carefully at the methodology used to assess treatment effect mediation. It is premature to go into further detail regarding this effort here, but suffice it to say that variation in the biomarker assessments for specific study subjects, including both technical measurement error and temporal variations over the lengthy time period that may be relevant to stroke risk, could be obscuring the ability of available biomarkers to explain the observed effects of HT on clinical outcomes. This is an important biomarker issue that may require biomarker measurements over a sustained period of time for individual study subjects, for an adequate resolution.

Biomarkers of dietary consumption

Biomarkers have a different, but equally important, role in efforts to study the association between dietary habits and the risk of stroke or other clinical outcomes; namely that of providing objective assessments of nutrients or other dietary components. Briefly, after many years of relating self-reported diet to the risk of major chronic diseases, there are few associations that are regarded as convincing or even probable [12,13], in spite of established relationships between obesity and the risk of such diseases, and in spite of factors such as blood cholesterol fractions that are diet related and strong risk factors for stroke and other vascular diseases. This situation may arise because of systematic errors in dietary self-report, whether based on the ubiquitous food frequency questionnaire (FFQ) assessment approach or other approaches, such as food records or recalls.

There is an opportunity for more reliable diet and disease risk data through the use of biomarkers for an objective assessment of consumption. There are established biomarkers, such as a doubly labeled water assessment of total energy consumption [14], and a urinary nitrogen assessment of protein consumption [15], that, when applied to a subsample of a study cohort, can be used to calibrate (i.e., correct) the self-report data on an entire study cohort. We applied this biomarker strategy in a study of 544 women enrolled in the WHI dietary modification randomized trial of a low-fat eating pattern (48,835 women in total). For example, according to the doubly labeled water assessment, the FFQ assessments underestimated energy consumption overall by approximately 30%, with much greater underestimation among obese compared with slim women, and among younger compared with older postmenopausal women, among other systematic bias [16]. When biomarker-calibrated estimates of energy and protein consumption were associated with the risk of major cancers [17] and cardiovascular diseases [18], practically important associations were indicated, none of which were evident using the FFQ data without biomarker calibration.

These analyses pointed to an inverse association between energy consumption and ischemic stroke risk that did not appear to be mediated by body fat deposition over time. This inverse association did not seem to be dependent on the percentage of energy that derived from protein. Although speculative at present, it may be that greater physical activity by higher energy consumers provides an explanation for this inverse association, also highlighting the need for objective measures of physical activity and its components, perhaps through the use of accelerometers, to advance knowledge concerning the relationship of diet, physical activity, and energy balance with the risk of stroke. There is a need for a major focus on the development of objective measures for additional components of diet and physical activity in upcoming years to realize this research potential. Our WHI investigator group is currently mounting a sizeable human feeding study for this purpose, to investigate potential biomarkers for various nutrients and dietary factors.

Summary

In summary, biomarkers have several important roles to play in the future research agenda for the prevention and control of stroke. These include the discovery of novel stroke risk markers for a better understanding of disease risk and disease pathogenesis, and for more thorough confounding control in observational studies of stroke risk. Our proteomic discovery work suggests that there may be many such biomarkers yet to be identified, and as such we encourage the application of additional ‘omics’ technologies for this purpose. Biomarkers may also be crucial to understanding the important intermediate variables and pathways whereby a treatment or exposure affects stroke risk, although improved methods may be needed to strengthen mediation analyses of this type. Finally, biomarkers of exposure in such important areas as dietary and physical activity epidemiology have the potential to usher in a new cycle of association analyses of enhanced reliability, depending on a much-needed effort in exposure biomarker development.

This article has not touched on the use of biomarkers in the treatment and management of stroke patients, which may, for example, include many additional markers of tissue damage and repair, and biomarkers derived from imaging technologies.