Hemostatic Factors as Risk Markers for Intracerebral Hemorrhage
A Prospective Incident Case-Referent Study
Background and Purpose— Abnormalities in the hemostatic system may cause hemorrhagic complications. The aim of the present study was to examine whether total concentrations of tissue plasminogen activator (tPA), plasminogen inhibitor-1 (PAI-1), tPA/PAI-1 complex, von Willebrand factor (VWF), and soluble thrombomodulin were associated with a first-ever intracerebral hemorrhage (ICH).
Methods— This prospective study was an incident case-referent study nested within the Västerbotten Intervention Program and the Northern Sweden Monitoring Trends and Determinants in Cardiovascular Disease (MONICA) cohorts. By 2000, ≈74 000 subjects had been screened, and 39 ICH cases were defined according to the World Health Organization MONICA criteria. A total of 78 matched controls were selected from the same cohort.
Results— The average time from screening to the ICH event was 5.1 years. tPA/PAI-1 complex, systolic and diastolic blood pressures, and hypertension were associated with ICH in the univariate analysis. In the multivariate model, only hypertension (odds ratio [OR], 3.96; 95% confidence interval [CI], 1.27 to 12.36) and the tertile with the highest level of VWF compared with the lowest tertile (OR, 0.27; 95% CI, 0.08 to 0.90) were independently associated with ICH. The OR for the combined exposure to hypertension and low levels of VWF was 8.95, indicating a possible synergistic interaction. No associations were observed for smoking, cholesterol, body mass index, PAI-1, tPA, and soluble thrombomodulin.
Conclusions— This study showed that hypertension and low concentrations of VWF were independently associated with ICH. Furthermore, we observed a possible synergistic interaction between low levels of VWF and hypertension, suggesting 2 different pathways in the development of ICH.
Intracerebral hemorrhage (ICH) accounts for ≈10% to 15% of all stroke cases,1 and hypertension is considered the most important modifiable risk factor for ICH.2 Other potential risk factors such as low levels of total cholesterol, high alcohol consumption, smoking, and low body mass index (BMI) have been investigated.3 In our previous prospective studies on first-ever stroke, we found increased levels of tissue plasminogen activator/plasminogen inhibitor-1 complex (tPA/PAI-1 complex) and low levels of soluble thrombomodulin (sTM) in the ICH subgroup compared with all controls. In the multivariate analysis, we found a positive independent association between tPA/PAI-1 complex and ICH and an inverse independent association between sTM and ICH.4,5
The primary aim of the present prospective study was to test the hypothesis that concentrations of tPA/PAI-1 complex, tPA, PAI-1, sTM, and von Willebrand factor (VWF) were associated with hemorrhagic stroke in a study population without previous stroke, myocardial infarction, or cancer. Furthermore, a secondary objective was to investigate the relationship and possible interactions between tPA/PAI-1 complex, tPA, PAI-1, sTM, VWF, and cardiovascular risk factors and the risk for ICH in this study population.
Materials and Methods
We used a prospective case-control study design nested within the Västerbotten Intervention program (VIP) and the World Health Organization (WHO) Monitoring Trends and Determinants in Cardiovascular Disease (MONICA) study in northern Sweden. Details about the VIP and MONICA projects have been given elsewhere.6,7 By September 20, 2000, ≈74 000 subjects had been screened in the VIP health survey (age groups, 40, 50, and 60 years) and the MONICA survey (age groups, 25 to 74 years). Participants were requested to donate blood samples and were asked answer questionnaire items on social background, smoking habits, medical history, and drug intake. All participants gave informed consent.
Cases with definite first-ever stroke were defined by the Northern Sweden MONICA incidence registry8 during the period of January 1, 1985, to September 20, 2000. Cases and controls were excluded if they had been registered for a previous acute myocardial infarction (AMI) or stroke according to the MONICA registry or for cancer according to the Swedish National Cancer Registry. In an earlier study on this study base between January 1, 1985, and August 31, 1996, we identified 108 first-ever stroke cases, including 18 ICH cases. The results have previously been published, and these cases were not included here.4,5 In the present study, we identified 228 first-ever stroke cases who fulfilled the inclusion criteria and had donated blood samples that were adequate for analysis. Two controls for each case were randomly selected from the same population-based health surveys. Controls were matched for sex, age (±2 years), date of health survey (±1 year), type of survey (VIP or MONICA), and geographic region. Controls were excluded if they had died, had moved out of the MONICA region before the event date of the index case, or had been registered for a previous AMI or stroke according to the MONICA registry. A questionnaire was sent to each control to confirm the absence of prior stroke and/or AMI.
The study was approved by the Research Ethics Committee of Umeå University. Data handling procedures were approved by the National Computer Data Inspection Board.
Clinical Variables and End Points
The diagnosis of stroke was made according to WHO MONICA criteria.7 The WHO criteria excluded all transient ischemic attacks, subdural hemorrhages, and acute strokes with concomitant brain tumor or severe blood disease. Stroke subtypes were divided into intracerebral hemorrhage (International Classification of Diseases, ninth revision [ICD-9] code 431), cerebral infarction (ICD-9 code 434), and unspecified stroke (ICD-9 code 436). Hemorrhagic stroke was diagnosed through a positive finding on CT scan and/or autopsy, and in cerebral infarction, no sign of hemorrhage on CT scan or autopsy was allowed. Unspecified stroke had neither CT scan nor autopsy performed. Cases diagnosed as subarachnoid hemorrhage were excluded. In the present study, the 39 first-ever stroke cases with ICH and 78 matched controls were investigated.
Hypertension was defined as systolic blood pressure (SBP) ≥160 mm Hg and/or diastolic blood pressure (DBP) ≥90 mm Hg or reported antihypertensive medication use during a 14-day period before the health survey. Smokers were divided into daily smokers and nonsmokers. Previous smokers and occasional smokers were classified as nonsmokers. BMI was calculated as weight divided by height squared (kg/m2). Statement on presence of diabetes was obtained from the questionnaire.
Blood Sampling and Laboratory Procedures
Venous blood samples for hemostatic assays were drawn without stasis into evacuated glass tubes (Venoject) containing 1/100 volume of 0.5 mmol/L EDTA. Plasma was obtained by centrifugation at 1500g for 15 minutes, placed in aliquots, and stored frozen at −80°C until analysis. Blood specimens from cases and controls were analyzed in triplets of 1 case and 2 controls; the position was varied randomly within each triplet to avoid systemic bias and interassay variability. The investigators and laboratory staff had no knowledge of case and control status. The mass concentrations (antigen) of tPA, PAI-1, and tPA/PAI-1 complex were determined with enzyme-linked immunosorbent assay (ELISA).9 The reagent kits used—Imulyse t-PA (for tPA mass concentration), Imulyse PAI-1 (for PAI-1 mass concentration), and TintElize tPA/PAI-1 (for tPA/PAI-1 complex mass concentration)—were purchased from Biopool. For tPA, the coefficient of variation (CV) was 10%; for PAI-1, 9%; and for tPA/PAI-1, 6.6%, all according to the manufacturer. sTM was measured with an ELISA method from STAGO, France. For sTM, it is reported that the intra-assay CVs (n=15) ranged from 4.1% to 7.4%, and interassay CVs (n=12) were between 4.8% and 8.6%.10 Calibration for sTM was made against the manufacturer’s standard. VWF was measured with an ELISA method11 using reagents from DAKO, Denmark. Values are expressed as percent of normal subjects. Calibration for VWF was made against a local standard. For VWF, the CV at a level of 126% was 4.1% in our hands (n=81). Serum samples for lipid measurement were obtained after a minimum of 4 hours of fasting. Total cholesterol was measured by enzymatic methods using Reflotron bench-top analyzers (Boehringer Mannheim GmbH Diagnostica).
Mean and SD values for baseline cardiovascular risk factors were calculated for cases and controls. Data were analyzed by conditional logistic regression using the maximum likelihood routine for matched analysis in the STATA 5.0 statistical package. To test the relation between high and low levels and risk of ICH, variables were categorized into tertiles or defined levels by the distribution of the control values. Univariate logistic regression was performed on each variable to estimate the odds ratios (ORs) and 95% confidence intervals (CIs). Multivariate logistic regression was performed to estimate the effects on different risk factors when controlling for other factors. Factors were excluded from the model if the univariate logistic regression resulted in a clearly nonsignificant OR (P>0.2). Spearman rank-order correlation coefficients were used to examine the associations of the studied factors, and Wilcoxon signed-rank statistics were used to compare the mean values of the studied factors between normotensive and hypertensive subjects and between nonsmokers and smokers. A value of P<0.05 (2 tailed) was considered statistically significant. Interaction between 2 risk factors was defined as departure from the additivity of effect of each risk factor.12 Synergy index with 95% CI was calculated, which gives the OR of the combined effects to the sum of the effects of 2 risk factors.13 With 39 ICH cases and 78 controls and a statistical power of 80%, ORs >3.1 were calculated to be significant at the 5% level when exposure prevalence was 30%.
The numbers of individuals with missing values per variable were as follows: diabetes, 2; smoking, 7; and hypertension, 3. Cholesterol, BMI, DBP, and SBP all had 4 missing values; PAI-1, tPA, tPA/PAI-1 complex, sTM, and VWF all had 5 missing values. In the multivariate logistic regression analysis, missing values for continuous variables were replaced by the mean value for controls to ensure a conservative estimate. For categorical variables, missing values were treated as a separate category and were omitted from the tables.
A first-ever ICH occurred in 39 cases (13 women, 26 men) after the baseline examination. The mean age at screening was 54.7 years (range, 39.4 to 61.1 years). The time from baseline examination to the ICH event was on average 5.1 years (range, 0.1 to 9.3 years). Case fatality within 3 days after the ICH onset was 17.9% (n=7) and within 28 days was 25.6% (n=10). Baseline characteristics for cases and controls are given in Table 1.
The univariate conditional logistic regression analysis with matched controls showed significant positive associations between SBP, DBP, and tPA/PAI-1 complex as continuous variables with ICH as outcome. No association was found for cholesterol, BMI, PAI-1, tPA, sTM, or VWF. Hypertension was significantly associated with ICH in the univariate analysis (OR, 4.00; 95% CI, 1.55 to 10.30). Each variable was categorized into tertiles, and ORs for the highest tertile (tertile 3) compared with the lowest tertile (tertile 1) as reference were calculated. However nonsignificant, the OR for BMI (tertile 3 compared with 1) was 2.45 (95% CI, 0.98 to 6.15), the OR for tPA/PAI-1 complex was 1.98 (95% CI, 0.70 to 5.63), and that for VWF was 0.49 (95% CI, 0.19 to 1.27) (see Table 2). In a post hoc analysis, tPA/PAI-1 complex concentrations were categorized into specified cutoff levels according to distribution of controls (below 70th, 70th to 80th, 80th to 90th, and >90th percentile), and a stepwise increase in ORs was observed for increasing concentrations compared with subjects below the 70th percentile as reference (OR, 1.0 versus 1.4 versus 2.2 versus 3.1). We found no association between tertiles of cholesterol and ICH or when cholesterol levels below the 15th percentile (<5.0 mmol/L) were compared with those above (data not shown). No associations were found for diabetes or smoking.
In the conditional multivariate logistic regression analysis, with controls matched for age and sex, we included variables with a value of P<0.2. After adjustment for hypertension, BMI, tPA/PAI-1 complex, and VWF, only hypertension and VWF remained independently associated with ICH (Table 2).
Possible interactions between hypertension and tPA/PAI-1 complex and VWF were investigated. The OR for the combined exposure for hypertension and VWF concentrations below the median was 8.95 (95% CI, 1.70 to 47.16) compared with exposure to hypertension and high levels of VWF (OR, 4.04; 95% CI, 0.81 to 20.23) and normotension and low levels of VWF (OR, 1.51; 95% CI, 0.25 to 9.05), yielding a calculated interaction effect of 4.4, thus indicating synergy. However, the 95% CI of synergy index was nonsignificant. No interaction between tPA/PAI-1 complex and hypertension was observed.
PAI-1, tPA, and tPA/PAI-1 complex were correlated to BMI, SBP, DBP, VWF, and each other, with the strongest correlation between tPA and tPA/PAI-1 complex (Spearman’s r=0.72). VWF, on the other hand, was correlated only with the fibrinolytic variables (see Table 3). Subjects with hypertension had significantly higher BMI and tPA/PAI-1 complex compared with normotensive subjects, and smokers were significantly younger than nonsmokers.
The main objective of this prospective study was to investigate the association between tPA, PAI-1, tPA/PAI-1 complex, sTM, VWF, and future risk of ICH. We observed a significant association with tPA/PAI-1 complex as a continuous variable and ICH as outcome. For tertiles of tPA/PAI-1 complex, a nonsignificant increased OR was observed when the highest tertile was compared with the lowest tertile (OR, 1.98). When tPA/PAI-1 complex concentrations were categorized into specified cutoff levels, we observed a stepwise increase in ORs for increasing concentrations (OR, 1.0 versus 1.4 versus 2.2 versus 3.1). This finding is in concordance with our earlier published results on the relation between tPA/PAI-1 complex and ICH.4 However, in our previous study, we found an independent association with tPA/PAI-1 complex in the subgroup of hemorrhagic stroke (n=18) when ICH cases were compared with all controls (n=216). This result was not confirmed in the present study in which we used a conditional multivariate model with 2 matched controls for each case, which may have influenced the result by decreasing the statistical power. For PAI-1 and tPA, we found no associations with ICH, as in our previous study.
There was no association between VWF and ICH in the univariate analysis. However, a nonsignificant trend toward an inverse relation between tertiles of VWF and ICH was observed. In the multivariate analysis, an inverse independent association between VWF and ICH was found. For tertile 3 compared with tertile 1, the observed OR was 0.27 (95% CI, 0.08 to 0.90). In our previous study on first-ever stroke, we found no significant association between VWF and ICH, although the mean level of VWF was lower for ICH cases (111.5%) compared with controls (129.8%).5 Furthermore, our present results suggest a possible interaction between VWF and hypertension, revealing an OR of 8.9 for the combined exposure, an OR of 4.0 for single exposure to hypertension, and an OR of 1.5 for single exposure to low levels of VWF.
VWF is an essential component in the aggregation of platelets and formation of thrombi, a mechanism that is central to the acute bleeding event. It is well known that abnormalities of VWF result in defective platelet adhesion with propensity for bleeding complications, as seen in von Willebrand disease. The biological mechanism for the present finding could be explained by such disturbances in the primary hemostasis caused by decreased levels of VWF and consequently a higher risk for bleeding complications such as ICH. VWF levels are raised in subjects with atherosclerotic disease and hypertension and are considered a marker for endothelial dysfunction in a variety of vascular disorders.14 In the present study, subjects with low levels of VWF and hypertension were found to have an additional risk for ICH. This observed synergistic effect supports the hypothesis of 2 causal pathways in the development of ICH, 1 pathway mediated through the effect of hypertension and 1 through hemostatic dysfunction.
There was no association of ICH with sTM as either a continuous or a tertilized variable, a finding that contradicts our earlier result showing an inverse association between sTM and ICH.5 Our previous study consisted of 18 ICH cases, and apparently, this result may have been due to chance. The present study included about twice as many ICH cases but still lacked power to detect possible associations at low risk levels, and a relationship between ICH and sTM cannot be excluded.
Hypertension, as expected, was the strongest risk factor for ICH, giving an OR of 4.0 in the univariate analysis. Hypertension remained independently associated with ICH in the multivariate model with an unchanged OR after adjustments for other determinants.
Other studies have found a relationship between BMI and ICH.15 In our study, we observed a nonsignificant trend toward increased risk of ICH with increasing BMI. This trend was clearly weakened after adjustments for other variables. No association was observed between cholesterol level and ICH in this work. The overall concentrations of cholesterol were high in our study population—few subjects had serum cholesterol levels <5.0 mmol/L—and an increased risk in the interval below this level cannot be ruled out.
ICH has a high early mortality. In a traditional case-control study, cases with severe symptoms or fatal outcome are difficult to include. This can result in a bias toward a selection of mild to moderate ICH cases among the study population. In the present study, all cases were defined by the MONICA registry, and the prospective design allowed cases with severe ICH and early mortality to be included. The observed case fatality within 3 days in our study was as high as 18%.
The main limitation of the present study was the relatively small number of ICH cases. Hypertension is by far the strongest risk factor for ICH, and other independent determinants may be difficult to find. Consequently, larger and more powerful studies are needed to identify possible risk determinants with ORs <3.
This study suggests that VWF and tPA/PAI-1 complex carry information about the risk for ICH. The study also supports the hypothesis of a possible synergistic interaction between VWF and hypertension. Today, regular blood pressure measuring, together with a general risk assessment, guides the prevention of hemorrhagic and ischemic stroke. Biochemical markers that identify high-risk patients would be valuable as indicators for intensified treatment of known modifiable risk factors for ICH, mainly antihypertensive treatment.
This research was supported by grants from Norrbotten and Västerbotten counties and by the Joint Committee of Northern Sweden Health Care Region, Swedish Public Health Institute, Swedish Medical Research Council, Heart and Chest Foundation, Stroke Fund, King Gustaf V’s and Queen Victoria’s Foundation, Foundation for Strategic Research, and Foundation of Medical Research in Skellefteå. We also thank the County Council of Västerbotten for its persistence in maintaining the community promotion program care and thereby creating and guaranteeing the study base.
- Received July 6, 2003.
- Revision received November 27, 2003.
- Accepted December 10, 2003.
Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project: 1981–86, 2: incidence, case fatality rates and overall outcome at one year of cerebral infarction, primary intracerebral and subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry. 1990; 53: 16–22.
Thrift AG, McNeil JJ, Forbes A, Donnan GA. Three important subgroups of hypertensive persons at greater risk of intracerebral hemorrhage: Melbourne Risk Factor Study Group. Hypertension. 1998; 31: 1223–1229.
Ariesen MJ, Claus SP, Rinkel GJ, Algra A. Risk factors for intracerebral hemorrhage in the general population: a systematic review. Stroke. 2003; 34: 2060–2065.
Johansson L, Jansson JH, Boman K, Nilsson TK, Stegmayr B, Hallmans G. Tissue plasminogen activator, plasminogen activator inhibitor-1, and tissue plasminogen activator/plasminogen activator inhibitor-1 complex as risk factors for the development of a first stroke. Stroke. 2000; 31: 26–32.
Weinehall L, Johnson O, Jansson JH, Boman K, Huhtasaari F, Hallmans G, Dahlen GH, Wall S. Perceived health modifies the effect of biomedical risk factors in the prediction of acute myocardial infarction: an incident case-control study from northern Sweden. J Intern Med. 1998; 243: 99–107.
Stegmayr B, Asplund K, Wester PO. Trends in incidence, case-fatality rate, and severity of stroke in northern Sweden, 1985–1991. Stroke. 1994; 25: 1738–1745.
Cejka J. Enzyme immunoassay for factor VIII-related antigen. Clin Chem. 1982; 28: 1356–1358.
Rothman KJ, Greenland S, Walker AM. Concepts of interaction. Am J Epidemiol. 1980; 112: 467–470.
Smith FB, Lee AJ, Fowkes FG, Price JF, Rumley A, Lowe GD. Hemostatic factors as predictors of ischemic heart disease and stroke in the Edinburgh Artery Study. Arterioscler Thromb Vasc Biol. 1997; 17: 3321–3325.
Thrift AG, McNeil JJ, Forbes A, Donnan GA. Risk factors for cerebral hemorrhage in the era of well-controlled hypertension: Melbourne Risk Factor Study (MERFS) Group. Stroke. 1996; 27: 2020–2025.