High-molecular-weight kininogen and the risk of a myocardial infarction and ischemic stroke in young women: the RATIO case–control study



    1. Department of Clinical Epidemiology, Leiden University Medical Center, Leiden
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden
    Search for more papers by this author

    1. Department of Clinical Epidemiology, Leiden University Medical Center, Leiden
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden
    3. Department of Thrombosis and Haemostasis, Leiden University Medical Center, Leiden
    Search for more papers by this author
  • A. ALGRA

    1. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center, Leiden
    2. UMC Utrecht Stroke Center, Department of Neurology and Neurosurgery, University Medical Center, Utrecht
    3. Julius Center, University Medical Center, Utrecht, the Netherlands
    Search for more papers by this author

Ale Algra, Department of Clinical Epidemiology, Leiden University Medical Center/C7-P, PO Box 9600, RC Leiden, the Netherlands.
Tel.: +31 71 526 4037; fax: +31 71 526 6994.
E-mail: a.algra@umcutrecht.nl

An emerging body of evidence suggests that intrinsic coagulation proteins are involved in the etiology of cardiovascular disease. Murine and other laboratory studies indicate that these proteins are indeed not involved in regular hemostasis, however, they are involved in pathological processes leading to thrombosis [1–5]. Previous clinical studies also suggest that the intrinsic coagulation proteins are involved in the development of human disease [6–13].

The actions of an activated intrinsic coagulation system are two-fold: (i) coagulation activation through factor (F)XI activation by FXIIa and (ii) formation of bradykinin through high-molecular-weight kininogen (HMWK) splicing by kallikrein, which is formed from its precursor prekallikrein by FXIIa [14,15]. Both actions require a negatively charged surface to facilitate protein–protein interactions; the area of the surface might be pivotal in determining the result of these interactions [14,16–19]. HMWK is a non-enzymatic cofactor of FXI and prekallikreine that facilitate this binding; both enzymes are non-covalently bound to HMWK in the circulation through one of the four apple domains present in these structurally homologous proteins [14,15,20,21]. Additionally, HWMK may interact with surfaces such as endothelium, platelets and neutrophils, which confirms the pleiotropic role of HWMK in biological systems [17,22].

HMWK is part of this intrinsic coagulation system, but is also involved in other biological systems related to cardiovascular diseases such as the innate inflammatory response, blood pressure control, brain edema, the renin–angiotensin system and endothelial cell differentiation [23–27]. HMWK has been sparsely studied in a clinical setting for both a myocardial infarction (MI) and ischaemic stroke. We therefore set out to determine whether HMWK is a risk factor for a MI and ischemic stroke within the RATIO study, a population-based case–control study [28–31]. This study includes women under the age of 50 years who where diagnosed using standard criteria with a MI or ischemic stroke in one of the 16 participating centers who were eligible to participate. Healthy controls, frequency matched on age (5 years categories), area of residence and index date, were approached via random digit dialing. Participants were recruited in two phases and asked to donate blood or buccal swabs for DNA analyzes. Ultimately, blood samples were available from 205 MI cases, 175 ischemic stroke cases and 638 healthy controls.

HMWK levels were measured with a sandwich ELISA-based assay using a polyclonal, commercially available antibody kit optimized to reduce the signal-to-noise ratio (CEDARLANCE Inc., Burlington, ON, Canada). Data on medical history, oral contraceptive (OC) use and other patient characteristics were obtained by questionnaire and reflect the year prior to the event (for the cases) or the frequency matched index date (controls) unless otherwise stated. The study was approved by institutional review committees of the participating hospitals and all subjects gave informed consent.

Odds ratios (ORs) as measures of relative risk and corresponding 95% confidence intervals (CIs) were calculated by means of logistic regression; all models included age (continuous), area of residence and index date (categorical) to account for the frequency matching procedure in this study. Traditional risk factors were included in the models as putative confounders. To determine the risk associated with extreme levels of HWMK, the 10th and 90th percentile of the control group were used as predefined cutoff values for low and high levels (i.e. ≤ p10 and ≥ p90). Quartile analyzes, with categories based on the 25th, 50th and 75th percentile of controls were predefined to investigate possible dose-response associations.

The mean HMWK levels in the two case group were similar to the control group (mean difference for MI 0%, 95% CI −4% to 3% and for ischemic stroke 4%, 95% CI 0% to 8%, see Supporting information). HMWK levels were increased in those suffering from diabetes mellitus (mean difference 14%, 95% CI 1% to 27%) and hypercholesterolemia (mean difference 8%, 95% CI −1% to 18%), whereas hypertension, previous OC use and previous smoking habits were not associated with substantial changes in HMWK levels (see Supporting information).

The risk of a MI was not increased in women with high levels (≥ p90) of HWMK after adjustment for confounders (ORadjusted1.05, 95% CI 0.57–1.91, see Table 1). The ischemic stroke risk, however, was increased about 80% (ORadjusted 1.82, 95% CI 1.00 to 3.29). Low levels of HMWK increased the risk of a MI by approximately 40% and the risk for ischemic stroke by approximately 50%, both after adjustment for confounders. Table S4 shows that there was no clear relationship between categories of increasing levels of HWMK and MI risk. The risk of ischemic stroke, however, was reflected by a U-curve (see Supporting information). Interaction analyzes for extreme levels of HMWK suggested that the effect of extreme levels of HMWK might be restricted to OC users, although broad confidence intervals prevent strong conclusions (see Supporting information).

Table 1.  Extreme levels of high-molecular-weight kininogen (HWMK) and the risk of a myocardial infarction (MI) and ischemic stroke
 ControlMIIschemic stroke
N % N %OR1(95% CI)OR2(95% CI) N %OR1(95% CI)OR2(95% CI)
  1. OR, odds ratio; 95% CI, 95% confidence interval; ref, reference group; p10, 10th percentile of control group; p90, 90th percentile of control group; OR1, odds ratios adjusted for stratification factors (i.e. age, area of residence and index year); OR2, odds ratios additionally adjusted for potential confounders (i.e. hypertension, diabetes mellitus and hypercholesterolemia) percentages might not add owing to rounding.

 ≥ p90640.10250.131.26(0.75–2.10)1.05(0.57–1.91)290.171.69(0.99–2.89)1.82(1.00–3.29)
 ≤ p10610.10240.121.69(0.98–2.89)1.39(0.74–2.61)200.121.62(0.88–2.99)1.49(0.77–2.89)

These results are not in line with previous studies which focused on MI [32,33]. However, as the blood samples in these previous studies were collected in the acute phase, their results are subject to reverse causation and should therefore not be interpreted causally. We did not identify prior studies which focused on the relation of plasma levels of HMWK and ischemic stroke. However, some studies on tissue kallikrein levels as well as the bradykinin receptor indicate that bradykinin is involved in stroke etiology through the exacerbation of stroke-related brain edema [27,34,35]. Although the biological mechanisms that were investigated as well as the outcome in these studies are not directly comparable to our study, they do suggest that HMWK and especially activation of bradykinin plays a role in the pathophysiology of ischemic stroke. Unfortunately, our current observational epidemiological study cannot discriminate between the possible mechanisms by which HWMK exerts its risk increasing action and further study is needed to elucidate the underlying mechanisms.

Our case–control study has some limitations. A major problem in case–control studies in particular is the possibility of reverse causation where cause and effect are mistaken for each other. It is unclear whether HMWK increases or decreases after a major event and could therefore be considered an acute phase reactant which could lead to reverse causation. Our blood samples were drawn a minimum of 23 months after the event, i.e. after the acute phase, which renders reverse causation unlikely. Treatment of patients could also change HMWK levels affecting the comparison with the controls. However, as treatment of a MI and ischemic stroke largely overlap, treatment cannot easily explain the differential effect of high levels of HMWK on MI and ischemic stroke. In this study design, blood samples could only be drawn after the event, limiting the inclusion to ‘survivors’. This selection would have affected our results only if the etiologic role of HMWK is different for fatal and non-fatal disease. We believe that HMWK levels are not likely to have a large effect on the case fatality rate, minimizing this effect in these analyzes.

Another problem could be confounding; a mechanism in which an observed effect is not caused by the exposure of interest but by a third factor related to the exposure of interest. We included putative confounders in our fully adjusted models, which yielded similar results as the non-adjusted models. Although we cannot rule out that our analyses are still confounded to some extent, we expect that the observed effects cannot be completely explained by confounding mechanisms. Although this study is of a reasonable size, especially considering the low incidence of vascular disease in young women, it has limited statistical power which is reflected in the CIs. In the interaction analyzes, these CIs become wide and too strong conclusions without confirmation are not warranted.

The results from the interaction analyzes suggest that the observed effect on ischemic stroke was independent of OC use. All other analyzes (i.e. low HMWK levels and the risk of ischemic stroke, low HMWK levels and the risk of MI, high HMWK levels and the risk of MI) suggested, if anything, an increase in risk only for women who were exposed to both risk factors. Further research, both epidemiological and from the laboratory, is needed to determine the role of HMWK in cardiovascular disease. Topics of interest are the differences between MI and ischemic stroke, the observed U-shaped dose-response curve and the role of OC use.

In conclusion, this study shows that extreme levels of HMWK are associated with an increased risk of ischemic stroke in young women. HWMK does not seem to have a major impact on the risk of a MI.


B. Siegerink: hypothesis formulation, measurements, data analysis, data interpretation and manuscript drafting. A. Algra: study design of RATIO, data interpretation and critical revision of manuscript no conflict of interest. F. R. Rosendaal: study design of RATIO, data interpretation and critical revision of manuscript. The final version of this manuscript was approved by all authors.


The authors wish to thank the participants and earlier contributors of the RATIO study. This study was supported by grants from the Netherlands Heart Foundation, The Netherlands (grant 1997.063, 2001.069 and 2005B060) and the Prevention Fund, The Netherlands (no. 28-2879).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.