Which Hemostatic Markers Add to the Predictive Value of Conventional Risk Factors for Coronary Heart Disease and Ischemic Stroke
http://www.100md.com
《循环学杂志》
the Department of Epidemiology and Public Health (A.S., C.P., J.Y.), Queen’s University, Belfast, Northern Ireland
the Cardiovascular and Medical Division (A.R., G.L.), University of Glasgow, Glasgow, Scotland
the Department of Social Medicine (Y.B.-S.), University of Bristol, Bristol, United Kingdom.
Abstract
Background— Few studies have examined whether hemostatic markers contribute to risk of coronary disease and ischemic stroke independently of conventional risk factors. This study examines 11 hemostatic markers that reflect different aspects of the coagulation process to determine which have prognostic value after accounting for conventional risk factors.
Methods and Results— A total of 2398 men aged 49 to 65 years were examined in 1984 to 1988, and the majority gave a fasting blood sample for assay of lipids and hemostatic markers. Men were followed up for a median of 13 years, and cardiovascular disease (CVD) events were recorded. There were 486 CVD events in total, 353 with prospective coronary disease and 133 with prospective ischemic stroke. On univariable analysis, fibrinogen, low activated protein C ratio, D-dimer, tissue plasminogen activator (tPA), and plasminogen activator inhibitor-1 (PAI-1) were associated significantly with risk of CVD. On multivariable analyses with conventional risk factors forced into the proportional hazards model, fibrinogen, D-dimer, and PAI-1 were significantly associated with risk of CVD, whereas factor VIIc showed an inverse association (P=0.001). In a model that contained the conventional risk factors, the hazard ratio for subsequent CVD in the top third of the distribution of predicted risk relative to the bottom third was 2.7 for subjects without preexisting CVD. This ratio increased to 3.7 for the model that also contained the 4 hemostatic factors.
Conclusions— Fibrinogen, D-dimer, PAI-1 activity, and factor VIIc each has potential to increase the prediction of coronary disease/ischemic stroke in middle-aged men, in addition to conventional risk factors.
Key Words: coagulation fibrinolysis fibrinogen coronary disease stroke
Introduction
There is increasing evidence that elevations in plasma markers of activated coagulation and fibrinolysis are associated with increased risk of coronary heart disease (CHD), but there are limited data for ischemic stroke.1,2 Meta-analyses of prospective cohort studies have shown significant associations of CHD risk with plasma levels of fibrinogen,3 hematocrit, viscosity and erythrocyte sedimentation rate,4 fibrin D-dimer5 (a measure of thrombin and plasmin activity that generates turnover of cross-linked fibrin), and the endothelial markers von Willebrand factor6 (vWF, which plays an important role in platelet adhesion and aggregation) and tissue plasminogen activator antigen7 (tPA, which plays a major role in endogenous fibrinolysis). At present, there are limited data from prospective studies of other hemostatic variables, including coagulation factors VII and VIII; markers of thrombin formation (eg, prothrombin fragments [F1+2], thrombin-antithrombin [TAT] complexes); plasminogen activator inhibitor (PAI-1); or activated protein C (APC) resistance.1,2,7 The latter phenotype is associated with genetic mutations including factor V Leiden, which is strongly associated with risk of venous thrombosis but only weakly with risk of CHD or stroke.1,8 However, the APC resistance phenotype has been associated in observational studies with risk factors for arterial thrombosis,9 coronary artery disease,10,11 and stroke.12
A major aim of the Caerphilly Prospective Study is to examine the associations of hemostatic variables with risk of CHD and stroke in middle-aged men. In previous reports from this cohort, we have reported significant associations of risk of CHD with plasma fibrinogen and viscosity,13,14 factor VIII and vWF,15 tPA (but not PAI-1),16 and fibrin D-dimer (but not factor VII, F1+2, TAT, or APC resistance).16,17 In these reports, lack of significant associations may have reflected the limited number of CHD events and hence a lack of statistical power. The aims of the present study were therefore (1) to report an updated analysis from further follow-up of the Caerphilly Prospective Study of the association of hemostatic variables with cardiovascular disease (CVD) events occurring after the baseline examination (ie, prospective events) and for its 2 components, CHD and ischemic stroke, and (2) to determine which hemostatic variables add to the prediction of CVD after accounting for conventional risk factors.
Methods
The general design and methods of the Caerphilly Prospective Study have been described elsewhere.13–17 Briefly, at each examination, the men were invited to attend an afternoon or evening clinic at which a detailed medical and lifestyle history was obtained, the London School of Hygiene and Tropical Medicine (LSHTM) chest pain questionnaire was administered, a full 12-lead ECG (ECG) was recorded, and weight and blood pressure were measured. The men were invited to return, fasting, to an early morning clinic where a blood sample was taken.
Study Population
Men from the general population were seen between 1984 and 1988, when they were aged 49 to 65 years. A total of 2398 men attended the evening clinic, and a fasting blood sample was obtained from 2223 (93%).
Blood Collection, Storage, and Analysis
Blood was taken between 7 and 10 AM for 91% of the men. It was taken before 7 AM for 7% and between 10 and 11 AM for the remaining 2%. The blood was collected without venous stasis into evacuated containers with a 19-gauge butterfly needle and Sarstedt monovette adaptors. Centrifugation was performed within 1 hour. Citrated plasma stored at –70°C was used for all samples assayed in the Department of Medicine, University of Glasgow, and fresh plasma samples anticoagulated with edetic acid (EDTA) were used for the measurement of nephelometric fibrinogen (Department of Hematology, Southmead Hospital, Bristol).
Assays for fibrinogen (Clauss and heat-nephelometry),14 factor VIII activity (VIIIc) and vWF activity and antigen,15 tPA antigen and PAI activity,16 fibrin D-dimer by the "original" ELISA assay16 and the more specific (Gold) assay17 (both from AGEN), F1+2, TAT, total factor VII activity (VIIc), activated partial thromboplastin time (APTT), and APC ratio17 were performed as described previously. Not all assays could be performed on all subjects because of the decreasing availability of stored citrated plasma samples, which was dependent solely on the amount of blood obtained from each subject and was not subject to selective bias (for details, see Sweetnam et al13–17). In preliminary analyses, fibrinogen, assayed by heat nephelometry, and D-dimer, assayed by the specific (Gold) assay, were found to be more closely associated with subsequent CVD than the alternative assays, and only these assays are reported further.
Preexisting CHD and Stroke and Prospective Cardiovascular Events
Follow-up was at a median interval of 13.4 years (interquartile range 10.1 to 14.8). All men were flagged with the National Health Service Central Registry, and we used death certificates coded to International Classification of Diseases-9th revision (ICD-9) 410 to 414 and to ICD-9 430 to 438 as our definition of fatal CHD and stroke, respectively. Preexisting CHD was defined from a past clinical history of myocardial infarction or angina with the LSHTM chest pain questionnaire, and subjects who had ECG evidence of past myocardial infarction were also included as described previously.13 Past history of stroke was obtained from the clinical history for each subject. The LSHTM questionnaire was also administered at the follow-up examinations with additional questions on hospital admissions. These, together with lists from Hospital Activity Analysis of all men admitted to local hospitals with a diagnosis of ICD 410 to 414, were used as the basis for a search of hospital notes for events that met standard World Health Organization (1996) criteria for acute myocardial infarction. Nonfatal ischemic stroke events were obtained by questionnaire and by checking hospital discharge diagnoses; these were then validated by presenting individual case vignettes, using all relevant clinical information from hospital and general practitioner records, to an expert clinical panel, as described elsewhere.18
Statistical Analysis
The end points used for analysis were (1) a prospective CVD event (stroke or CHD), (2) a prospective stroke event, or (3) a prospective CHD event. A prospective event as defined here means either the first event of CVD after baseline in individuals free of disease at baseline, or a recurrent event in individuals with clinical or historical evidence of CVD before baseline. The latter group was included in the analysis because their exclusion would have considerably reduced the statistical power of the study and would less accurately reflect the situation in clinical practice.
A stratified Cox proportional hazards model was used to estimate hazard ratios for the hemostatic variables while taking account of possible differences in the underlying CVD hazard between men with and men without CVD at baseline. For other variables, the assumption of hazard proportionality was assessed with cumulative hazard plots and by tests of interaction between categorized variables and time in the Cox model. Because the linearity of relationships could not be assumed, the distributions for the variables were divided by sample tertiles into thirds, and the lowest third was used as the reference category. Initial analyses were adjusted only for age. Subsequent analyses added the following conventional risk factors as potential confounders: smoking (never/past/current), diabetes (yes/no), systolic blood pressure, total cholesterol, HDL cholesterol, total triglycerides, body mass index, and family history of CHD before 55 years of age (yes/no). It is possible that some of the pathophysiological effects of these conventional risk factors may operate through hemostatic variables. In this case, their "direct" effect will be underestimated in any multivariable model owing to the inclusion of hemostatic intermediaries that contribute to their "indirect" pathway.
Tests for linear trend in hazard ratios across the thirds of each hemostatic variable were obtained, and tests for deviation from linearity of trend were also examined. A retrospective assessment of power used methodology derived for the log rank test19 applied to the ratio of hazards in the top third of the distribution of a hemostatic factor relative to the bottom third. Because there was variation in the numbers of subjects with data available for the various hemostatic factors, only rather general statements about power can be made. For the analysis of the CVD end point (CHD and stroke combined), the study was of sufficient size to have at least 90% power to detect a hazard ratio of between 1.50 and 1.67, depending on the completeness of data. Likewise, for analysis of the CHD end point, there was at least 80% power to detect a hazard ratio of between 1.50 and 1.67. For the less common stroke end point, there was at least 80% power to detect a hazard ratio of between 2.0 and 2.5.
To identify the most predictive subset of the hemostatic variables that were independently associated with CVD in the presence of conventional risk factors, we used a forward stepwise approach with conventional risk factors forced into the Cox model. Tests for linear trend across the 3 categories were used to assess the significance of the hemostatic variables in this model.
All analyses were performed with SPSS version 11.0 using 2-sided tests at the 5% significance level.
Results
Thirteen men with a hemorrhagic stroke during follow-up were removed from the analysis, as were 4 men with no available follow-up data after baseline, which left a total of 2381 men; 648 of these died during follow-up. Only those who had fasted before blood sampling and for whom complete information on conventional risk factors was available were included in the analysis, which resulted in 2208 subjects being included in the present analysis.
Demographic and nonhemostatic conventional risk factors were summarized and their distributions examined. The distributions of all hemostatic variables are positively skewed, requiring logarithmic transformation before analysis, and have been summarized with geometric means and interquartile ranges. The baseline distributions of conventional risk factors, lipids (total triglycerides required logarithmic transformation) and those hemostatic factors examined, are given in Table 1 for the entire study population. Interrelationships between the hemostatic markers, lipids and age, were examined by correlation analysis and are shown in Table 2. The strongest interrelationships were noted for hemostatic variables that were closely biologically related, eg, factor VIIIc, vWF, and APPT, as well as tPA and PAI-1, but triglycerides and total cholesterol showed relationships with factor VIIc, tPA, and PAI-1 with associated correlation coefficients of sufficient magnitude to suggest physiological rather than purely statistical significance.
During the follow-up period, 486 subjects had a CVD event, including 353 cases who had a CHD event first and 133 who had a stroke event first. A Cox proportional hazards analysis was conducted for these outcome events across the thirds of the distribution of the hemostatic variables. The analysis was stratified into 2 strata by the presence or absence of preexisting CVD. Tertile cutpoints and numbers of events in each category are shown in Table 3. Two models were fitted for each hemostatic variable; the first only adjusted for age (crude), whereas the second also adjusted for smoking, diabetes, systolic blood pressure, total cholesterol, HDL cholesterol, triglycerides, body mass index, and family history of premature CHD (adjusted).
For CVD events, significant linear trends over thirds of the distribution are shown for fibrinogen, APC ratio, D-dimer, and PAI-1 whether adjusted for age only or additionally for the other risk factors. For tPA, only the crude model showed a significant trend, whereas for factor VIIc, only the adjusted model showed a significant trend. The patterns between CVD and CHD were essentially similar, reflecting the greater number of CHD events than stroke events For prospective CHD, significant linear trends over thirds of the distribution are shown for D-dimer and APC ratio in both the crude and adjusted models. For fibrinogen, APTT, tPA, and PAI-1, only the crude models showed a significant trend. With factor VIIc, only the adjusted model showed a significant trend. All other hemostatic variables were nonsignificant. For the prospective stroke end point, D-dimer had a significant linear trend in both crude and adjusted analyses, whereas a significant trend was also recorded for PAI-1 for the crude analysis only.
Tests for deviation from linearity were mostly nonsignificant, which indicates that most of the associations with hemostatic variables could be satisfactorily explained by linear trends across the thirds of the distributions. Exceptions were the adjusted F1+2 analysis for CVD, the adjusted factor VIIc analysis, the crude and adjusted factor VIIIc analyses, and the adjusted TAT analysis for CHD. All tests for deviation from linearity were nonsignificant for stroke.
The analyses were repeated for subjects without evidence of CVD at baseline (n=1523). The pattern of results was broadly similar to those shown for the entire population in Table 3, but several results no longer achieved statistical significance, possibly because of the reduced numbers of subjects. However, certain variables (including TAT, D-dimer, and PAI-1) showed divergent results for CHD and ischemic stroke, which suggests a higher level of prediction for ischemic stroke rather than for CHD. These results are shown in the online-only Data Supplement.
Fibrinogen, D-dimer, PAI-1 (all positively associated), and factor VIIc (negatively associated) were found to be independent significant risk factors for a CVD event in the final model. Systolic blood pressure, diabetes, and total cholesterol were also significant risk factors. Age and smoking habit became nonsignificantly associated with risk of a CVD event owing to their associations with the hemostatic variables (particularly with D-dimer). HDL cholesterol, triglycerides, body mass index, and a family history of premature CHD were also nonsignificant in the model.
In the stratified Cox analysis, the 2 strata represented subjects with and without preexisting CVD. In subjects without preexisting CVD for the model that contained only the conventional risk factors, the CVD hazard ratio in the top third of the distribution of predicted risk relative to the bottom third was 2.7. In the model that additionally included the 4 hemostatic variables, this ratio increased to 3.7. In subjects with preexisting CVD, the model was less predictive overall, and the respective hazard ratios were 2.1 and 3.1.
A sensitivity analysis on the model described in Table 4 was performed for subjects without preexisting CVD at baseline. This barely altered the results: D-dimer Gold 1.31, (95% CI 1.05 to 1.64), factor VIIc 0.71, (0.56 to 0.91), and PAI-1 1.23 (0.97 to 1.57), except for fibrinogen 1.16 (0.93 to 1.46), which was attenuated.
Discussion
Previous reports have tended to provide information on the contribution of individual5 or small groups4 of hemostatic markers in relation to subsequent risk of CHD. In the present report, we have examined the contribution of a range of hemostatic factors considered simultaneously, which represents the spectrum of coagulation and fibrinolysis, as well as their capacity to add to the prediction of subsequent CVD when conventional risk factors have been taken into account. Table 4 shows that 4 of the 11 hemostatic variables were independently predictive of CVD after adjustment for conventional risk factors; the small probability values for 3 of these variables suggest that these are unlikely to be chance findings arising from multiple analyses. The probability value associated with PAI-1, although significant at the conventional 5% level, was less extreme than those for the other 3 variables in the context of multiple testing, and the PAI-1 results must be considered preliminary until replicated in other studies. The present data suggest that not only does a "hypercoagulable state" precede these 2 common clinical presentations of arterial thrombosis, but also that these markers of activated coagulation and fibrinolysis may potentially add to clinical prediction of both CHD and ischemic stroke events and increase their prognostic ability, because the combined end point contributes to a larger burden of disease.
Fibrinogen
The present results show that in this study, plasma fibrinogen continues to be a strong predictor of CHD risk13,14 and shows a similar pattern for ischemic stroke, although this failed to achieve statistical significance. Other prospective studies20–23 show positive associations between fibrinogen and ischemic stroke. However, when restricted to subjects with no preexisting CVD, the sensitivity analysis indicated that fibrinogen was no longer significant, which suggests that some of the effect in Table 3 may be due to the inclusion of subjects with preexisting CVD. A study based on patients with cerebrovascular disease (transient ischemic attack) showed that fibrinogen predicted less strongly for subsequent stroke than it did for subsequent CHD.24 However, in the present study, blood pressure and fibrinogen were not correlated. The Fibrinogen Studies Collaboration25 is currently performing a meta-analysis of individual data from prospective studies on the associations of fibrinogen with both CHD and ischemic stroke.
Fibrin D-Dimer, Markers of Thrombin Activation, and Factor VIIc
The results of the present study show that in the Caerphilly Prospective Study, D-dimer continues to be a strong predictor of CHD events16,17 and is also predictive of prospective stroke. This is consistent with findings in the Edinburgh Artery Study26 and a study of primary care patients in Glasgow with atrial fibrillation27 and supports possible roles for increased fibrin turnover in the prediction and pathogenesis of arterial disease and thrombosis.5,28 As in our previous report,17 markers of thrombin generation (F1+2, TAT) and factor VIIc were not significantly associated with risk of CHD on univariable analyses, nor were they associated with risk of ischemic stroke. An unexpected finding from the final multivariable analysis was an inverse association of factor VIIc with CVD risk (Table 4). In this analysis, factor VIIc achieved significance with the inclusion of the lipids in the model depicted in Table 4. Although this may be a chance finding, a similar inverse association was observed in the Second Northwick Park Study.29 In general, however, studies of factor VIIc and CVD risk have produced conflicting findings.1,2
Factor VIIIc and vWF
In a 13-year follow-up, factor VIIIc and vWF activity and antigen, which are strongly correlated, were not significantly associated with risk of CVD end points in the Caerphilly cohort.15 A recent, updated meta-analysis of prospective studies (including the large Reykjavik study) showed a weak association.30 Further studies of factor VIIIc are of interest because of its association with venous thrombosis8 and because of the reduced, dose-dependent risk of CHD in hemophiliacs31 and hemophilia carriers,32 but studies of risk of ischemic stroke have shown conflicting results.26,33
tPA Antigen and PAI Activity
Both of these variables showed significant associations with CHD risk in the present report with extended follow-up, but tPA became nonsignificant after multivariable adjustment. This is consistent with a recent meta-analysis7 of prospective studies of tPA antigen, which also suggested that further studies of PAI-1 are required. The present data suggest that one possible explanation for the association of tPA antigen levels with risk of CHD is that tPA antigen primarily measures tPA:PAI-1 complexes, which are determined in part by increasing PAI-1 levels.7 Previous reports have implicated both tPA antigen and PAI-1 activity in the risk of ischemic stroke.26,34,35 In the present report, after multivariable analyses, PAI-1 remained an independent predictor of CVD, whereas tPA did not. Further prospective studies (and meta-analyses) of PAI-1 and tPA and CVD risk are required.
APC Resistance and APTT
APC resistance phenotype (low APC ratio) was a significant independent risk factor for the combined end point of CVD events when adjusted for conventional risk factors, including smoking habit, which is associated with increased APC ratio.9 However, in the final model of multivariable analyses (Table 4), APC resistance was not independently associated with CVD risk, probably because of associations with other coagulation factors. We were unable to test whether or not this association was due to different frequencies of the factor V Leiden mutation in the event groups (owing to lack of both DNA for mutation analysis and unthawed stored plasma to repeat the APC resistance assay with factor V-depleted plasma).9 However, such an explanation appears unlikely, because large studies have shown only weak associations between the factor V Leiden mutation and risk of CHD or ischemic stroke.1,8 Hence, the association is more likely to reflect "acquired" APC resistance.9 Although a low APC ratio is associated with high factor VIIIc levels,9 factor VIIIc was not associated with CHD or ischemic stroke events in the present analysis. For all fasting cases in the Caerphilly data set, the correlation of factor VIIIc with APTT was –0.57, and its correlation with APC was –0.23, both of which were significant at P<0.01. Low APC ratio is also associated with increasing age, obesity, and hyperlipidemia9; however, multivariable analysis that included these variables did not affect the independent associations of APTT and APC ratio with risk of subsequent CVD. It is possible that these associations may reflect mutual associations with other coagulation factors, such as factor V or factor IX,9 which we were unable to assay in the present study because of lack of unthawed stored plasma. Together with data from observational studies of arterial disease risk factors,9,10 CHD,10,11 and stroke,12 the present data suggest that further prospective studies be performed of APC resistance phenotypes, genetic determinants of APC resistance, and risks of CVD.
In this study, the primary aim is risk stratification and predicting the risk of disease events over future time periods. The Framingham risk equation has gained much prominence, although it has been demonstrated to overestimate risk in British males36 and may not be applicable to other populations in other countries. This has led to a flurry of research activity in devising regional or country-specific prognostic algorithms, as well as possible genotypic identities.37 These scores, however, require direct comparison of the assumptions, definitions, and methodologies used for risk factor analysis. They do not include hemostatic variables, whereas the results of the present study show that several hemostatic variables may be helpful and may even supersede existing conventional risk factors. We have shown previously38,39 that several lifestyle factors are independently associated with hemostatic variables, and others have shown associations with diabetes.40 Inclusion of these conventional risk factors is likely to underestimate the overall contribution of hemostatic factors that may provide a pathogenic pathway for atherothrombotic risk. This has to be balanced against the cost of and access to such tests. Future work should compare the discrepancies in risk, both relative and absolute, between a range of models with and without specific hemostatic variables. Future risk models may also investigate other biological pathways of risk, such as inflammatory markers, for example, C-reactive protein and interleukin-6, and natriuretic hormones, which may be relevant in signifying early vascular damage.
Finally, it is acknowledged that the present study has a number of limitations that restrict the potential generalizability of the findings. First, the population comprises only white males in a UK region at higher-than-average risk of cardiovascular disease. Next, there was a limited availability of results for some hemostatic variable (missing values). Finally, the study has limited statistical power, which has particularly restricted comparison of predictive models for the separate CHD and stroke end points. In addition, validation of stroke events was limited by restricted availability of computerized tomography, particularly at the beginning of the study, and the possibility exists that some hemorrhagic strokes are included among those assumed from the medical history and hospital notes to be ischemic. Repeated measures for the hemostatic variables may have improved their predictive ability but were not available for variables studied in the present report. In a previous report, repeated measures did not usefully add to prediction when lipids and a restricted range of hemostatic factors were considered.41
Conclusions
This updated report from the Caerphilly Prospective Study identifies a "hypercoagulable profile" of hemostatic variables that is associated with risk of ischemic stroke and CHD. We confirm previously observed significant associations of fibrinogen and fibrin D-dimer with risk of CHD and extend the latter to risk of ischemic stroke and both to the combined end point of CVD. We also confirm no significant associations between markers of thrombin activation and risk of CHD or ischemic stroke. New observations include the association of APC resistance with the combined end point of CVD (although not independently associated in the final model that included all other hemostatic factors) and the association of PAI-1 activity (independent of tPA antigen) with both CHD and ischemic stroke. Finally, our previous observations of significant associations between factor VIIIc, vWF, and risk of CHD were not confirmed on further follow-up of this cohort, nor were they significantly associated with risk of ischemic stroke. Further follow-up of this cohort will provide additional information on the associations of several hemostatic variables with risks of CVD, as will collaborative meta-analyses of individual data from this and other prospective studies, such as the Fibrinogen Studies Collaboration.25
Acknowledgments
This work was supported by a grant from the British Heart Foundation and by the Medical Research Council (MRC). The latter has established the database and steering group for the former MRC Epidemiology Unit (South Wales) at the Department of Social Medicine, University of Bristol.
Footnotes
The online-only Data Supplement can be found with this article at http://circ.ahajournals.org/cgi/content/full/112/20/3080/DC1.
References
Reiner AP, Siscovick DS, Rosendaal FR. Hemostatic risk factors and arterial thrombotic disease. Thromb Haemost. 2001; 85: 584–595.
Lowe GDO, Rumley A, Whincup PH, Danesh J. Hemostatic and rheological variables and risk of cardiovascular disease. Semin Vasc Med. 2002; 2: 429–440.
Danesh J, Collins R, Appleby P, Peto R. Association of fibrinogen, C-reactive protein, albumin, or leukocyte count with coronary heart disease: meta-analyses of prospective studies. JAMA. 1998; 279: 1477–1482.
Danesh J, Collins R, Peto R, Lowe GDO. Haematocrit, viscosity, erythrocyte sedimentation rate: meta-analyses of prospective studies of coronary heart disease. Eur Heart J. 2000; 21: 515–520.
Danesh J, Whincup P, Walker M, Lennon L, Thomson A, Appleby P, Rumley A, Lowe GD. Fibrin D-dimer and coronary heart disease: prospective study and meta-analysis. Circulation. 2001; 103: 2323–2327.
Whincup P, Danesh J, Walker M, Lennon L, Thomson A, Appleby P, Rumley A, Lowe GD. Von Willebrand factor and coronary heart disease: prospective study and meta-analysis. Eur Heart J. 2002; 23: 1764–1770.
Lowe GDO, Danesh J, Lewington S, Walker M, Lennon L, Thomson A, Rumley A, Whincup PH. Tissue plasminogen activator antigen and coronary heart disease: prospective study and meta-analysis. Eur Heart J. 2004; 25: 252–259.
Kim RJ, Becker RC. Association between factor V Leiden, prothrombin G20210A, and methylenetetrahydrofolate reductase C677T mutations and events of the arterial circulatory system: a meta-analysis of published studies. Am Heart J. 2003; 146: 948–957.
Lowe GDO, Rumley A, Woodward M, Reid E, Rumley J. Activated protein C resistance and the FV: R506Q mutation in a random population sample: associations with cardiovascular risk factors and coagulation variables. Thromb Haemost. 1999; 81: 918–924.
Kiechl S, Muigg A, Santer P, Mitterer M, Egger G, Oberhollenzer M, Oberhollenzer F, Mayr A, Gasperi A, Poewe W, Willeit J. Poor response to activated protein C as a prominent risk predictor of advanced atherosclerosis and arterial disease. Circulation. 1999; 99: 614–619.
Zahn R, Beeck H, Winkelmann BR, Seidl K, Schneider S, Hellstern P, Senges J. Prospective cross-sectional study of haemostatic factors in patients with and without coronary heart disease. Blood Coagul Fibrinolysis. 2002; 13: 81–87.
van der Bom JG, Bots ML, Haverkate F, Slagboom PE, Meijer P, de Jong PT, Hofman A, Grobbee DE, Kluft C. Reduced response to activated protein C is associated with increased risk for cerebrovascular disease. Ann Intern Med. 1996; 125: 265–269.
Sweetnam PM, Thomas HF, Yarnell JW, Beswick AD, Baker IA, Elwood PC. Fibrinogen, viscosity and the 10-year incidence of ischaemic heart disease: The Caerphilly and Speedwell studies. Eur Heart J. 1996; 17: 1814–1820.
Sweetnam PM, Yarnell JW, Lowe GD, Baker IA, O’Brien JR, Rumley A, Etherington MD, Whitehead PJ, Elwood PC. The relative power of heat-precipitation nephelometric and clottable (Clauss) fibrinogen in the prediction of ischaemic heart disease: the Caerphilly and Speedwell studies. Br J Haematol. 1998; 100: 582–588.
Rumley A, Lowe GDO, Sweetnam PM, Yarnell JWG, Ford RP. Factor VIII, von Willebrand factor and the risk of major ischaemic heart disease in the Caerphilly Study. Br J Haematol. 1999; 105: 110–116.
Lowe GDO, Yarnell JWG, Sweetnam PM, Rumley A, Thomas HF, Elwood PC. Fibrin D-dimer, tissue plasminogen activator, plasminogen activator inhibitor, and the risk of major ischaemic heart disease in the Caerphilly Study. Thromb Haemost. 1998; 79: 129–133.
Lowe GDO, Rumley A, Sweetnam P, Yarnell JWG, Rumley J. Fibrin D-dimer, markers of coagulation activation and the risk of major ischaemic heart disease in the Caerphilly Study. Thromb Haemost. 2001; 86: 822–827.
Baker IA, Pickering J, Elwood PC, Bayer A, Ebrahim S. Fibrinogen, viscosity and white blood cell count predict myocardial, but not cerebral infarction: evidence from the Caerphilly and Speedwell cohort. Thromb Haemost. 2002; 87: 421–425.
Freedman LS. Tables of the number of patients required in clinical trials using the logrank test. Stats Med. 1982; 1: 121–129.
Wilhelmsen L, Svardsudd K, Korsan-Bengsten K, Larsson B, Welin L, Tibblin G. Fibrinogen as a risk factor for stroke and myocardial infarction. N Engl J Med. 1984; 311: 501–505.
Kannel WB, Wolf PA, Castelli WP, D’Agostino RB. Fibrinogen and risk of cardiovascular disease: the Framingham Study. JAMA. 1987; 258: 1183–1186.
Lowe GDO, Lee AJ, Rumley A, Price JF, Fowkes FG. Blood viscosity and risk of cardiovascular events: the Edinburgh Artery Study. Br J Haematol. 1997; 96: 168–173.
Maresca G, Di Blasio A, Marchioli R, Di Minno G. Measuring plasma fibrinogen to predict stroke and myocardial infarction: an update. Arterioscler Thromb Vasc Biol. 1999; 19: 1368–1377.
Rothwell PM, Howard SC, Power DA, Gutnikov SA, Algra A, van Gijn J, Clark TG, Murphy MF, Warlow CP. Fibrinogen concentration and risk of ischemic stroke and acute coronary events in 5113 patients with transient ischemic attack and minor ischemic stroke. Stroke. 2004; 35: 2300–2305.
Fibrinogen Studies Collaboration. Collaborative meta-analysis of prospective observational studies of plasma fibrinogen and cardiovascular disease. Euro J Cardiovasc Prev Rehabil. 2003; 11: 9–17.
Smith FB, Lee AJ, Fowkes FGR, Price JF, Rumley A, Lowe GDO. Haemostatic factors as predictors of ischaemic heart disease and stroke in the Edinburgh Artery Study. Arterioscler Thromb Vasc Biol. 1997; 17: 3321–3325.
Tait RC, Murdoch DL, O’Neill KF, McMahon A, Wallace IW, Rumley A, Jackson J, Lowe GDO. Fibrin D-dimer is an independent predictor of future stroke in patients with atrial fibrillation. Circulation. 2003; 108 (suppl IV): IV-787. Abstract.
Fowkes FGR, Lowe GDO, Housley E, Fowkes FG, Lowe GD, Housley E, Rattray A, Rumley A, Elton RA, MacGregor IR, Dawes J. Cross-linked fibrin degradation products, progression of peripheral arterial disease, and risk of coronary heart disease. Lancet. 1993; 342: 84–86.
Cooper JA, Miller GJ, Bauer KA, Morrissey JH, Meade TW, Howarth DJ, Barzegar S, Mitchell JP, Rosenberg RD. Comparison of novel haemostatic factors and conventional risk factors for prediction of coronary heart disease. Circulation. 2000; 102: 2816–2822.
Danesh J, Wheeler JG, Hirschfield GM, Eda S, Eiriksdottir G, Rumley A, Lowe GD, Pepys MB, Gudnason V. C-reactive protein and other circulating markers of inflammation in the prediction of coronary heart disease. N Engl J Med. 2004; 350: 1387–1397.
Rosendaal FR, Briet E, Stibbe J, van Herpen G, Leuven JA, Hofman A, Vandenbroucke JP. Haemophilia protects against ischaemic heart disease: a study of risk factors. Br J Haematol. 1990; 75: 525–530.
Sramek A, Kriek M, Rosendaal FR. Decreased mortality of ischaemic heart disease among carriers of haemophilia. Lancet. 2003; 362: 351–54.
Folsom AR, Rosamond WD, Shahar E, Cooper LS, Aleksic N, Nieto FJ, Rasmussen ML, Wu KK. Prospective study of markers of hemostatic function with risk of ischemic stroke. Circulation. 1999; 100: 736–742.
Ridker PM, Hennekens CH, Stampfer MJ, Manson JE, Vaughan DE. Prospective study of endogenous tissue plasminogen activator and risk of stroke. Lancet. 1994; 343: 940–943.
Nilsson TK, Jansson JH, Bowman K. t-PA–PAI-1 complex as a risk factor for the development of a first stroke. Fibrinolysis. 1998; 12 (suppl 1): 21.Abstract.
Brindle P, Emberson J, Lampe F, Walker M, Whincup P, Fahey T, Ebrahim S. Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study. BMJ. 2003; 327: 1267–1270.
Topol EJ, Lauer MS. The rudimentary phase of personalised medicine: coronary risk scores. Lancet. 2003; 362: 1776–1777.
Yarnell JWG, Sweetnam PM, Rumley A, Lowe GD. Lifestyle and hemostatic risk for ischemic heart disease: the Caerphilly study. Arterioscler Thromb Vasc Biol. 2000; 20: 271–279.
Yarnell JWG, Sweetnam PM, Rumley A, Lowe GDO. Lifestyle factors and coagulation activation markers: the Caerphilly study. Blood Coagul Fibrinolysis. 2001; 12: 721–728.
Mansfield MW, Heywood DM, Grant PJ. Sex differences in coagulation and fibrinolysis in white subjects with non–insulin-dependent diabetes mellitus. Arterioscler Thromb Vasc Biol. 1996; 16: 160–164.
Yarnell JWG, Patterson CC, Sweetnam PM, Lowe GDO. Haemostatic/inflammatory markers predict 10-year risk of IHD at least as well as lipids: the Caerphilly collaborative studies. Eur Heart J. 2004; 25: 1049–1056.(Ann Smith, PhD; Chris Pat)
the Cardiovascular and Medical Division (A.R., G.L.), University of Glasgow, Glasgow, Scotland
the Department of Social Medicine (Y.B.-S.), University of Bristol, Bristol, United Kingdom.
Abstract
Background— Few studies have examined whether hemostatic markers contribute to risk of coronary disease and ischemic stroke independently of conventional risk factors. This study examines 11 hemostatic markers that reflect different aspects of the coagulation process to determine which have prognostic value after accounting for conventional risk factors.
Methods and Results— A total of 2398 men aged 49 to 65 years were examined in 1984 to 1988, and the majority gave a fasting blood sample for assay of lipids and hemostatic markers. Men were followed up for a median of 13 years, and cardiovascular disease (CVD) events were recorded. There were 486 CVD events in total, 353 with prospective coronary disease and 133 with prospective ischemic stroke. On univariable analysis, fibrinogen, low activated protein C ratio, D-dimer, tissue plasminogen activator (tPA), and plasminogen activator inhibitor-1 (PAI-1) were associated significantly with risk of CVD. On multivariable analyses with conventional risk factors forced into the proportional hazards model, fibrinogen, D-dimer, and PAI-1 were significantly associated with risk of CVD, whereas factor VIIc showed an inverse association (P=0.001). In a model that contained the conventional risk factors, the hazard ratio for subsequent CVD in the top third of the distribution of predicted risk relative to the bottom third was 2.7 for subjects without preexisting CVD. This ratio increased to 3.7 for the model that also contained the 4 hemostatic factors.
Conclusions— Fibrinogen, D-dimer, PAI-1 activity, and factor VIIc each has potential to increase the prediction of coronary disease/ischemic stroke in middle-aged men, in addition to conventional risk factors.
Key Words: coagulation fibrinolysis fibrinogen coronary disease stroke
Introduction
There is increasing evidence that elevations in plasma markers of activated coagulation and fibrinolysis are associated with increased risk of coronary heart disease (CHD), but there are limited data for ischemic stroke.1,2 Meta-analyses of prospective cohort studies have shown significant associations of CHD risk with plasma levels of fibrinogen,3 hematocrit, viscosity and erythrocyte sedimentation rate,4 fibrin D-dimer5 (a measure of thrombin and plasmin activity that generates turnover of cross-linked fibrin), and the endothelial markers von Willebrand factor6 (vWF, which plays an important role in platelet adhesion and aggregation) and tissue plasminogen activator antigen7 (tPA, which plays a major role in endogenous fibrinolysis). At present, there are limited data from prospective studies of other hemostatic variables, including coagulation factors VII and VIII; markers of thrombin formation (eg, prothrombin fragments [F1+2], thrombin-antithrombin [TAT] complexes); plasminogen activator inhibitor (PAI-1); or activated protein C (APC) resistance.1,2,7 The latter phenotype is associated with genetic mutations including factor V Leiden, which is strongly associated with risk of venous thrombosis but only weakly with risk of CHD or stroke.1,8 However, the APC resistance phenotype has been associated in observational studies with risk factors for arterial thrombosis,9 coronary artery disease,10,11 and stroke.12
A major aim of the Caerphilly Prospective Study is to examine the associations of hemostatic variables with risk of CHD and stroke in middle-aged men. In previous reports from this cohort, we have reported significant associations of risk of CHD with plasma fibrinogen and viscosity,13,14 factor VIII and vWF,15 tPA (but not PAI-1),16 and fibrin D-dimer (but not factor VII, F1+2, TAT, or APC resistance).16,17 In these reports, lack of significant associations may have reflected the limited number of CHD events and hence a lack of statistical power. The aims of the present study were therefore (1) to report an updated analysis from further follow-up of the Caerphilly Prospective Study of the association of hemostatic variables with cardiovascular disease (CVD) events occurring after the baseline examination (ie, prospective events) and for its 2 components, CHD and ischemic stroke, and (2) to determine which hemostatic variables add to the prediction of CVD after accounting for conventional risk factors.
Methods
The general design and methods of the Caerphilly Prospective Study have been described elsewhere.13–17 Briefly, at each examination, the men were invited to attend an afternoon or evening clinic at which a detailed medical and lifestyle history was obtained, the London School of Hygiene and Tropical Medicine (LSHTM) chest pain questionnaire was administered, a full 12-lead ECG (ECG) was recorded, and weight and blood pressure were measured. The men were invited to return, fasting, to an early morning clinic where a blood sample was taken.
Study Population
Men from the general population were seen between 1984 and 1988, when they were aged 49 to 65 years. A total of 2398 men attended the evening clinic, and a fasting blood sample was obtained from 2223 (93%).
Blood Collection, Storage, and Analysis
Blood was taken between 7 and 10 AM for 91% of the men. It was taken before 7 AM for 7% and between 10 and 11 AM for the remaining 2%. The blood was collected without venous stasis into evacuated containers with a 19-gauge butterfly needle and Sarstedt monovette adaptors. Centrifugation was performed within 1 hour. Citrated plasma stored at –70°C was used for all samples assayed in the Department of Medicine, University of Glasgow, and fresh plasma samples anticoagulated with edetic acid (EDTA) were used for the measurement of nephelometric fibrinogen (Department of Hematology, Southmead Hospital, Bristol).
Assays for fibrinogen (Clauss and heat-nephelometry),14 factor VIII activity (VIIIc) and vWF activity and antigen,15 tPA antigen and PAI activity,16 fibrin D-dimer by the "original" ELISA assay16 and the more specific (Gold) assay17 (both from AGEN), F1+2, TAT, total factor VII activity (VIIc), activated partial thromboplastin time (APTT), and APC ratio17 were performed as described previously. Not all assays could be performed on all subjects because of the decreasing availability of stored citrated plasma samples, which was dependent solely on the amount of blood obtained from each subject and was not subject to selective bias (for details, see Sweetnam et al13–17). In preliminary analyses, fibrinogen, assayed by heat nephelometry, and D-dimer, assayed by the specific (Gold) assay, were found to be more closely associated with subsequent CVD than the alternative assays, and only these assays are reported further.
Preexisting CHD and Stroke and Prospective Cardiovascular Events
Follow-up was at a median interval of 13.4 years (interquartile range 10.1 to 14.8). All men were flagged with the National Health Service Central Registry, and we used death certificates coded to International Classification of Diseases-9th revision (ICD-9) 410 to 414 and to ICD-9 430 to 438 as our definition of fatal CHD and stroke, respectively. Preexisting CHD was defined from a past clinical history of myocardial infarction or angina with the LSHTM chest pain questionnaire, and subjects who had ECG evidence of past myocardial infarction were also included as described previously.13 Past history of stroke was obtained from the clinical history for each subject. The LSHTM questionnaire was also administered at the follow-up examinations with additional questions on hospital admissions. These, together with lists from Hospital Activity Analysis of all men admitted to local hospitals with a diagnosis of ICD 410 to 414, were used as the basis for a search of hospital notes for events that met standard World Health Organization (1996) criteria for acute myocardial infarction. Nonfatal ischemic stroke events were obtained by questionnaire and by checking hospital discharge diagnoses; these were then validated by presenting individual case vignettes, using all relevant clinical information from hospital and general practitioner records, to an expert clinical panel, as described elsewhere.18
Statistical Analysis
The end points used for analysis were (1) a prospective CVD event (stroke or CHD), (2) a prospective stroke event, or (3) a prospective CHD event. A prospective event as defined here means either the first event of CVD after baseline in individuals free of disease at baseline, or a recurrent event in individuals with clinical or historical evidence of CVD before baseline. The latter group was included in the analysis because their exclusion would have considerably reduced the statistical power of the study and would less accurately reflect the situation in clinical practice.
A stratified Cox proportional hazards model was used to estimate hazard ratios for the hemostatic variables while taking account of possible differences in the underlying CVD hazard between men with and men without CVD at baseline. For other variables, the assumption of hazard proportionality was assessed with cumulative hazard plots and by tests of interaction between categorized variables and time in the Cox model. Because the linearity of relationships could not be assumed, the distributions for the variables were divided by sample tertiles into thirds, and the lowest third was used as the reference category. Initial analyses were adjusted only for age. Subsequent analyses added the following conventional risk factors as potential confounders: smoking (never/past/current), diabetes (yes/no), systolic blood pressure, total cholesterol, HDL cholesterol, total triglycerides, body mass index, and family history of CHD before 55 years of age (yes/no). It is possible that some of the pathophysiological effects of these conventional risk factors may operate through hemostatic variables. In this case, their "direct" effect will be underestimated in any multivariable model owing to the inclusion of hemostatic intermediaries that contribute to their "indirect" pathway.
Tests for linear trend in hazard ratios across the thirds of each hemostatic variable were obtained, and tests for deviation from linearity of trend were also examined. A retrospective assessment of power used methodology derived for the log rank test19 applied to the ratio of hazards in the top third of the distribution of a hemostatic factor relative to the bottom third. Because there was variation in the numbers of subjects with data available for the various hemostatic factors, only rather general statements about power can be made. For the analysis of the CVD end point (CHD and stroke combined), the study was of sufficient size to have at least 90% power to detect a hazard ratio of between 1.50 and 1.67, depending on the completeness of data. Likewise, for analysis of the CHD end point, there was at least 80% power to detect a hazard ratio of between 1.50 and 1.67. For the less common stroke end point, there was at least 80% power to detect a hazard ratio of between 2.0 and 2.5.
To identify the most predictive subset of the hemostatic variables that were independently associated with CVD in the presence of conventional risk factors, we used a forward stepwise approach with conventional risk factors forced into the Cox model. Tests for linear trend across the 3 categories were used to assess the significance of the hemostatic variables in this model.
All analyses were performed with SPSS version 11.0 using 2-sided tests at the 5% significance level.
Results
Thirteen men with a hemorrhagic stroke during follow-up were removed from the analysis, as were 4 men with no available follow-up data after baseline, which left a total of 2381 men; 648 of these died during follow-up. Only those who had fasted before blood sampling and for whom complete information on conventional risk factors was available were included in the analysis, which resulted in 2208 subjects being included in the present analysis.
Demographic and nonhemostatic conventional risk factors were summarized and their distributions examined. The distributions of all hemostatic variables are positively skewed, requiring logarithmic transformation before analysis, and have been summarized with geometric means and interquartile ranges. The baseline distributions of conventional risk factors, lipids (total triglycerides required logarithmic transformation) and those hemostatic factors examined, are given in Table 1 for the entire study population. Interrelationships between the hemostatic markers, lipids and age, were examined by correlation analysis and are shown in Table 2. The strongest interrelationships were noted for hemostatic variables that were closely biologically related, eg, factor VIIIc, vWF, and APPT, as well as tPA and PAI-1, but triglycerides and total cholesterol showed relationships with factor VIIc, tPA, and PAI-1 with associated correlation coefficients of sufficient magnitude to suggest physiological rather than purely statistical significance.
During the follow-up period, 486 subjects had a CVD event, including 353 cases who had a CHD event first and 133 who had a stroke event first. A Cox proportional hazards analysis was conducted for these outcome events across the thirds of the distribution of the hemostatic variables. The analysis was stratified into 2 strata by the presence or absence of preexisting CVD. Tertile cutpoints and numbers of events in each category are shown in Table 3. Two models were fitted for each hemostatic variable; the first only adjusted for age (crude), whereas the second also adjusted for smoking, diabetes, systolic blood pressure, total cholesterol, HDL cholesterol, triglycerides, body mass index, and family history of premature CHD (adjusted).
For CVD events, significant linear trends over thirds of the distribution are shown for fibrinogen, APC ratio, D-dimer, and PAI-1 whether adjusted for age only or additionally for the other risk factors. For tPA, only the crude model showed a significant trend, whereas for factor VIIc, only the adjusted model showed a significant trend. The patterns between CVD and CHD were essentially similar, reflecting the greater number of CHD events than stroke events For prospective CHD, significant linear trends over thirds of the distribution are shown for D-dimer and APC ratio in both the crude and adjusted models. For fibrinogen, APTT, tPA, and PAI-1, only the crude models showed a significant trend. With factor VIIc, only the adjusted model showed a significant trend. All other hemostatic variables were nonsignificant. For the prospective stroke end point, D-dimer had a significant linear trend in both crude and adjusted analyses, whereas a significant trend was also recorded for PAI-1 for the crude analysis only.
Tests for deviation from linearity were mostly nonsignificant, which indicates that most of the associations with hemostatic variables could be satisfactorily explained by linear trends across the thirds of the distributions. Exceptions were the adjusted F1+2 analysis for CVD, the adjusted factor VIIc analysis, the crude and adjusted factor VIIIc analyses, and the adjusted TAT analysis for CHD. All tests for deviation from linearity were nonsignificant for stroke.
The analyses were repeated for subjects without evidence of CVD at baseline (n=1523). The pattern of results was broadly similar to those shown for the entire population in Table 3, but several results no longer achieved statistical significance, possibly because of the reduced numbers of subjects. However, certain variables (including TAT, D-dimer, and PAI-1) showed divergent results for CHD and ischemic stroke, which suggests a higher level of prediction for ischemic stroke rather than for CHD. These results are shown in the online-only Data Supplement.
Fibrinogen, D-dimer, PAI-1 (all positively associated), and factor VIIc (negatively associated) were found to be independent significant risk factors for a CVD event in the final model. Systolic blood pressure, diabetes, and total cholesterol were also significant risk factors. Age and smoking habit became nonsignificantly associated with risk of a CVD event owing to their associations with the hemostatic variables (particularly with D-dimer). HDL cholesterol, triglycerides, body mass index, and a family history of premature CHD were also nonsignificant in the model.
In the stratified Cox analysis, the 2 strata represented subjects with and without preexisting CVD. In subjects without preexisting CVD for the model that contained only the conventional risk factors, the CVD hazard ratio in the top third of the distribution of predicted risk relative to the bottom third was 2.7. In the model that additionally included the 4 hemostatic variables, this ratio increased to 3.7. In subjects with preexisting CVD, the model was less predictive overall, and the respective hazard ratios were 2.1 and 3.1.
A sensitivity analysis on the model described in Table 4 was performed for subjects without preexisting CVD at baseline. This barely altered the results: D-dimer Gold 1.31, (95% CI 1.05 to 1.64), factor VIIc 0.71, (0.56 to 0.91), and PAI-1 1.23 (0.97 to 1.57), except for fibrinogen 1.16 (0.93 to 1.46), which was attenuated.
Discussion
Previous reports have tended to provide information on the contribution of individual5 or small groups4 of hemostatic markers in relation to subsequent risk of CHD. In the present report, we have examined the contribution of a range of hemostatic factors considered simultaneously, which represents the spectrum of coagulation and fibrinolysis, as well as their capacity to add to the prediction of subsequent CVD when conventional risk factors have been taken into account. Table 4 shows that 4 of the 11 hemostatic variables were independently predictive of CVD after adjustment for conventional risk factors; the small probability values for 3 of these variables suggest that these are unlikely to be chance findings arising from multiple analyses. The probability value associated with PAI-1, although significant at the conventional 5% level, was less extreme than those for the other 3 variables in the context of multiple testing, and the PAI-1 results must be considered preliminary until replicated in other studies. The present data suggest that not only does a "hypercoagulable state" precede these 2 common clinical presentations of arterial thrombosis, but also that these markers of activated coagulation and fibrinolysis may potentially add to clinical prediction of both CHD and ischemic stroke events and increase their prognostic ability, because the combined end point contributes to a larger burden of disease.
Fibrinogen
The present results show that in this study, plasma fibrinogen continues to be a strong predictor of CHD risk13,14 and shows a similar pattern for ischemic stroke, although this failed to achieve statistical significance. Other prospective studies20–23 show positive associations between fibrinogen and ischemic stroke. However, when restricted to subjects with no preexisting CVD, the sensitivity analysis indicated that fibrinogen was no longer significant, which suggests that some of the effect in Table 3 may be due to the inclusion of subjects with preexisting CVD. A study based on patients with cerebrovascular disease (transient ischemic attack) showed that fibrinogen predicted less strongly for subsequent stroke than it did for subsequent CHD.24 However, in the present study, blood pressure and fibrinogen were not correlated. The Fibrinogen Studies Collaboration25 is currently performing a meta-analysis of individual data from prospective studies on the associations of fibrinogen with both CHD and ischemic stroke.
Fibrin D-Dimer, Markers of Thrombin Activation, and Factor VIIc
The results of the present study show that in the Caerphilly Prospective Study, D-dimer continues to be a strong predictor of CHD events16,17 and is also predictive of prospective stroke. This is consistent with findings in the Edinburgh Artery Study26 and a study of primary care patients in Glasgow with atrial fibrillation27 and supports possible roles for increased fibrin turnover in the prediction and pathogenesis of arterial disease and thrombosis.5,28 As in our previous report,17 markers of thrombin generation (F1+2, TAT) and factor VIIc were not significantly associated with risk of CHD on univariable analyses, nor were they associated with risk of ischemic stroke. An unexpected finding from the final multivariable analysis was an inverse association of factor VIIc with CVD risk (Table 4). In this analysis, factor VIIc achieved significance with the inclusion of the lipids in the model depicted in Table 4. Although this may be a chance finding, a similar inverse association was observed in the Second Northwick Park Study.29 In general, however, studies of factor VIIc and CVD risk have produced conflicting findings.1,2
Factor VIIIc and vWF
In a 13-year follow-up, factor VIIIc and vWF activity and antigen, which are strongly correlated, were not significantly associated with risk of CVD end points in the Caerphilly cohort.15 A recent, updated meta-analysis of prospective studies (including the large Reykjavik study) showed a weak association.30 Further studies of factor VIIIc are of interest because of its association with venous thrombosis8 and because of the reduced, dose-dependent risk of CHD in hemophiliacs31 and hemophilia carriers,32 but studies of risk of ischemic stroke have shown conflicting results.26,33
tPA Antigen and PAI Activity
Both of these variables showed significant associations with CHD risk in the present report with extended follow-up, but tPA became nonsignificant after multivariable adjustment. This is consistent with a recent meta-analysis7 of prospective studies of tPA antigen, which also suggested that further studies of PAI-1 are required. The present data suggest that one possible explanation for the association of tPA antigen levels with risk of CHD is that tPA antigen primarily measures tPA:PAI-1 complexes, which are determined in part by increasing PAI-1 levels.7 Previous reports have implicated both tPA antigen and PAI-1 activity in the risk of ischemic stroke.26,34,35 In the present report, after multivariable analyses, PAI-1 remained an independent predictor of CVD, whereas tPA did not. Further prospective studies (and meta-analyses) of PAI-1 and tPA and CVD risk are required.
APC Resistance and APTT
APC resistance phenotype (low APC ratio) was a significant independent risk factor for the combined end point of CVD events when adjusted for conventional risk factors, including smoking habit, which is associated with increased APC ratio.9 However, in the final model of multivariable analyses (Table 4), APC resistance was not independently associated with CVD risk, probably because of associations with other coagulation factors. We were unable to test whether or not this association was due to different frequencies of the factor V Leiden mutation in the event groups (owing to lack of both DNA for mutation analysis and unthawed stored plasma to repeat the APC resistance assay with factor V-depleted plasma).9 However, such an explanation appears unlikely, because large studies have shown only weak associations between the factor V Leiden mutation and risk of CHD or ischemic stroke.1,8 Hence, the association is more likely to reflect "acquired" APC resistance.9 Although a low APC ratio is associated with high factor VIIIc levels,9 factor VIIIc was not associated with CHD or ischemic stroke events in the present analysis. For all fasting cases in the Caerphilly data set, the correlation of factor VIIIc with APTT was –0.57, and its correlation with APC was –0.23, both of which were significant at P<0.01. Low APC ratio is also associated with increasing age, obesity, and hyperlipidemia9; however, multivariable analysis that included these variables did not affect the independent associations of APTT and APC ratio with risk of subsequent CVD. It is possible that these associations may reflect mutual associations with other coagulation factors, such as factor V or factor IX,9 which we were unable to assay in the present study because of lack of unthawed stored plasma. Together with data from observational studies of arterial disease risk factors,9,10 CHD,10,11 and stroke,12 the present data suggest that further prospective studies be performed of APC resistance phenotypes, genetic determinants of APC resistance, and risks of CVD.
In this study, the primary aim is risk stratification and predicting the risk of disease events over future time periods. The Framingham risk equation has gained much prominence, although it has been demonstrated to overestimate risk in British males36 and may not be applicable to other populations in other countries. This has led to a flurry of research activity in devising regional or country-specific prognostic algorithms, as well as possible genotypic identities.37 These scores, however, require direct comparison of the assumptions, definitions, and methodologies used for risk factor analysis. They do not include hemostatic variables, whereas the results of the present study show that several hemostatic variables may be helpful and may even supersede existing conventional risk factors. We have shown previously38,39 that several lifestyle factors are independently associated with hemostatic variables, and others have shown associations with diabetes.40 Inclusion of these conventional risk factors is likely to underestimate the overall contribution of hemostatic factors that may provide a pathogenic pathway for atherothrombotic risk. This has to be balanced against the cost of and access to such tests. Future work should compare the discrepancies in risk, both relative and absolute, between a range of models with and without specific hemostatic variables. Future risk models may also investigate other biological pathways of risk, such as inflammatory markers, for example, C-reactive protein and interleukin-6, and natriuretic hormones, which may be relevant in signifying early vascular damage.
Finally, it is acknowledged that the present study has a number of limitations that restrict the potential generalizability of the findings. First, the population comprises only white males in a UK region at higher-than-average risk of cardiovascular disease. Next, there was a limited availability of results for some hemostatic variable (missing values). Finally, the study has limited statistical power, which has particularly restricted comparison of predictive models for the separate CHD and stroke end points. In addition, validation of stroke events was limited by restricted availability of computerized tomography, particularly at the beginning of the study, and the possibility exists that some hemorrhagic strokes are included among those assumed from the medical history and hospital notes to be ischemic. Repeated measures for the hemostatic variables may have improved their predictive ability but were not available for variables studied in the present report. In a previous report, repeated measures did not usefully add to prediction when lipids and a restricted range of hemostatic factors were considered.41
Conclusions
This updated report from the Caerphilly Prospective Study identifies a "hypercoagulable profile" of hemostatic variables that is associated with risk of ischemic stroke and CHD. We confirm previously observed significant associations of fibrinogen and fibrin D-dimer with risk of CHD and extend the latter to risk of ischemic stroke and both to the combined end point of CVD. We also confirm no significant associations between markers of thrombin activation and risk of CHD or ischemic stroke. New observations include the association of APC resistance with the combined end point of CVD (although not independently associated in the final model that included all other hemostatic factors) and the association of PAI-1 activity (independent of tPA antigen) with both CHD and ischemic stroke. Finally, our previous observations of significant associations between factor VIIIc, vWF, and risk of CHD were not confirmed on further follow-up of this cohort, nor were they significantly associated with risk of ischemic stroke. Further follow-up of this cohort will provide additional information on the associations of several hemostatic variables with risks of CVD, as will collaborative meta-analyses of individual data from this and other prospective studies, such as the Fibrinogen Studies Collaboration.25
Acknowledgments
This work was supported by a grant from the British Heart Foundation and by the Medical Research Council (MRC). The latter has established the database and steering group for the former MRC Epidemiology Unit (South Wales) at the Department of Social Medicine, University of Bristol.
Footnotes
The online-only Data Supplement can be found with this article at http://circ.ahajournals.org/cgi/content/full/112/20/3080/DC1.
References
Reiner AP, Siscovick DS, Rosendaal FR. Hemostatic risk factors and arterial thrombotic disease. Thromb Haemost. 2001; 85: 584–595.
Lowe GDO, Rumley A, Whincup PH, Danesh J. Hemostatic and rheological variables and risk of cardiovascular disease. Semin Vasc Med. 2002; 2: 429–440.
Danesh J, Collins R, Appleby P, Peto R. Association of fibrinogen, C-reactive protein, albumin, or leukocyte count with coronary heart disease: meta-analyses of prospective studies. JAMA. 1998; 279: 1477–1482.
Danesh J, Collins R, Peto R, Lowe GDO. Haematocrit, viscosity, erythrocyte sedimentation rate: meta-analyses of prospective studies of coronary heart disease. Eur Heart J. 2000; 21: 515–520.
Danesh J, Whincup P, Walker M, Lennon L, Thomson A, Appleby P, Rumley A, Lowe GD. Fibrin D-dimer and coronary heart disease: prospective study and meta-analysis. Circulation. 2001; 103: 2323–2327.
Whincup P, Danesh J, Walker M, Lennon L, Thomson A, Appleby P, Rumley A, Lowe GD. Von Willebrand factor and coronary heart disease: prospective study and meta-analysis. Eur Heart J. 2002; 23: 1764–1770.
Lowe GDO, Danesh J, Lewington S, Walker M, Lennon L, Thomson A, Rumley A, Whincup PH. Tissue plasminogen activator antigen and coronary heart disease: prospective study and meta-analysis. Eur Heart J. 2004; 25: 252–259.
Kim RJ, Becker RC. Association between factor V Leiden, prothrombin G20210A, and methylenetetrahydrofolate reductase C677T mutations and events of the arterial circulatory system: a meta-analysis of published studies. Am Heart J. 2003; 146: 948–957.
Lowe GDO, Rumley A, Woodward M, Reid E, Rumley J. Activated protein C resistance and the FV: R506Q mutation in a random population sample: associations with cardiovascular risk factors and coagulation variables. Thromb Haemost. 1999; 81: 918–924.
Kiechl S, Muigg A, Santer P, Mitterer M, Egger G, Oberhollenzer M, Oberhollenzer F, Mayr A, Gasperi A, Poewe W, Willeit J. Poor response to activated protein C as a prominent risk predictor of advanced atherosclerosis and arterial disease. Circulation. 1999; 99: 614–619.
Zahn R, Beeck H, Winkelmann BR, Seidl K, Schneider S, Hellstern P, Senges J. Prospective cross-sectional study of haemostatic factors in patients with and without coronary heart disease. Blood Coagul Fibrinolysis. 2002; 13: 81–87.
van der Bom JG, Bots ML, Haverkate F, Slagboom PE, Meijer P, de Jong PT, Hofman A, Grobbee DE, Kluft C. Reduced response to activated protein C is associated with increased risk for cerebrovascular disease. Ann Intern Med. 1996; 125: 265–269.
Sweetnam PM, Thomas HF, Yarnell JW, Beswick AD, Baker IA, Elwood PC. Fibrinogen, viscosity and the 10-year incidence of ischaemic heart disease: The Caerphilly and Speedwell studies. Eur Heart J. 1996; 17: 1814–1820.
Sweetnam PM, Yarnell JW, Lowe GD, Baker IA, O’Brien JR, Rumley A, Etherington MD, Whitehead PJ, Elwood PC. The relative power of heat-precipitation nephelometric and clottable (Clauss) fibrinogen in the prediction of ischaemic heart disease: the Caerphilly and Speedwell studies. Br J Haematol. 1998; 100: 582–588.
Rumley A, Lowe GDO, Sweetnam PM, Yarnell JWG, Ford RP. Factor VIII, von Willebrand factor and the risk of major ischaemic heart disease in the Caerphilly Study. Br J Haematol. 1999; 105: 110–116.
Lowe GDO, Yarnell JWG, Sweetnam PM, Rumley A, Thomas HF, Elwood PC. Fibrin D-dimer, tissue plasminogen activator, plasminogen activator inhibitor, and the risk of major ischaemic heart disease in the Caerphilly Study. Thromb Haemost. 1998; 79: 129–133.
Lowe GDO, Rumley A, Sweetnam P, Yarnell JWG, Rumley J. Fibrin D-dimer, markers of coagulation activation and the risk of major ischaemic heart disease in the Caerphilly Study. Thromb Haemost. 2001; 86: 822–827.
Baker IA, Pickering J, Elwood PC, Bayer A, Ebrahim S. Fibrinogen, viscosity and white blood cell count predict myocardial, but not cerebral infarction: evidence from the Caerphilly and Speedwell cohort. Thromb Haemost. 2002; 87: 421–425.
Freedman LS. Tables of the number of patients required in clinical trials using the logrank test. Stats Med. 1982; 1: 121–129.
Wilhelmsen L, Svardsudd K, Korsan-Bengsten K, Larsson B, Welin L, Tibblin G. Fibrinogen as a risk factor for stroke and myocardial infarction. N Engl J Med. 1984; 311: 501–505.
Kannel WB, Wolf PA, Castelli WP, D’Agostino RB. Fibrinogen and risk of cardiovascular disease: the Framingham Study. JAMA. 1987; 258: 1183–1186.
Lowe GDO, Lee AJ, Rumley A, Price JF, Fowkes FG. Blood viscosity and risk of cardiovascular events: the Edinburgh Artery Study. Br J Haematol. 1997; 96: 168–173.
Maresca G, Di Blasio A, Marchioli R, Di Minno G. Measuring plasma fibrinogen to predict stroke and myocardial infarction: an update. Arterioscler Thromb Vasc Biol. 1999; 19: 1368–1377.
Rothwell PM, Howard SC, Power DA, Gutnikov SA, Algra A, van Gijn J, Clark TG, Murphy MF, Warlow CP. Fibrinogen concentration and risk of ischemic stroke and acute coronary events in 5113 patients with transient ischemic attack and minor ischemic stroke. Stroke. 2004; 35: 2300–2305.
Fibrinogen Studies Collaboration. Collaborative meta-analysis of prospective observational studies of plasma fibrinogen and cardiovascular disease. Euro J Cardiovasc Prev Rehabil. 2003; 11: 9–17.
Smith FB, Lee AJ, Fowkes FGR, Price JF, Rumley A, Lowe GDO. Haemostatic factors as predictors of ischaemic heart disease and stroke in the Edinburgh Artery Study. Arterioscler Thromb Vasc Biol. 1997; 17: 3321–3325.
Tait RC, Murdoch DL, O’Neill KF, McMahon A, Wallace IW, Rumley A, Jackson J, Lowe GDO. Fibrin D-dimer is an independent predictor of future stroke in patients with atrial fibrillation. Circulation. 2003; 108 (suppl IV): IV-787. Abstract.
Fowkes FGR, Lowe GDO, Housley E, Fowkes FG, Lowe GD, Housley E, Rattray A, Rumley A, Elton RA, MacGregor IR, Dawes J. Cross-linked fibrin degradation products, progression of peripheral arterial disease, and risk of coronary heart disease. Lancet. 1993; 342: 84–86.
Cooper JA, Miller GJ, Bauer KA, Morrissey JH, Meade TW, Howarth DJ, Barzegar S, Mitchell JP, Rosenberg RD. Comparison of novel haemostatic factors and conventional risk factors for prediction of coronary heart disease. Circulation. 2000; 102: 2816–2822.
Danesh J, Wheeler JG, Hirschfield GM, Eda S, Eiriksdottir G, Rumley A, Lowe GD, Pepys MB, Gudnason V. C-reactive protein and other circulating markers of inflammation in the prediction of coronary heart disease. N Engl J Med. 2004; 350: 1387–1397.
Rosendaal FR, Briet E, Stibbe J, van Herpen G, Leuven JA, Hofman A, Vandenbroucke JP. Haemophilia protects against ischaemic heart disease: a study of risk factors. Br J Haematol. 1990; 75: 525–530.
Sramek A, Kriek M, Rosendaal FR. Decreased mortality of ischaemic heart disease among carriers of haemophilia. Lancet. 2003; 362: 351–54.
Folsom AR, Rosamond WD, Shahar E, Cooper LS, Aleksic N, Nieto FJ, Rasmussen ML, Wu KK. Prospective study of markers of hemostatic function with risk of ischemic stroke. Circulation. 1999; 100: 736–742.
Ridker PM, Hennekens CH, Stampfer MJ, Manson JE, Vaughan DE. Prospective study of endogenous tissue plasminogen activator and risk of stroke. Lancet. 1994; 343: 940–943.
Nilsson TK, Jansson JH, Bowman K. t-PA–PAI-1 complex as a risk factor for the development of a first stroke. Fibrinolysis. 1998; 12 (suppl 1): 21.Abstract.
Brindle P, Emberson J, Lampe F, Walker M, Whincup P, Fahey T, Ebrahim S. Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study. BMJ. 2003; 327: 1267–1270.
Topol EJ, Lauer MS. The rudimentary phase of personalised medicine: coronary risk scores. Lancet. 2003; 362: 1776–1777.
Yarnell JWG, Sweetnam PM, Rumley A, Lowe GD. Lifestyle and hemostatic risk for ischemic heart disease: the Caerphilly study. Arterioscler Thromb Vasc Biol. 2000; 20: 271–279.
Yarnell JWG, Sweetnam PM, Rumley A, Lowe GDO. Lifestyle factors and coagulation activation markers: the Caerphilly study. Blood Coagul Fibrinolysis. 2001; 12: 721–728.
Mansfield MW, Heywood DM, Grant PJ. Sex differences in coagulation and fibrinolysis in white subjects with non–insulin-dependent diabetes mellitus. Arterioscler Thromb Vasc Biol. 1996; 16: 160–164.
Yarnell JWG, Patterson CC, Sweetnam PM, Lowe GDO. Haemostatic/inflammatory markers predict 10-year risk of IHD at least as well as lipids: the Caerphilly collaborative studies. Eur Heart J. 2004; 25: 1049–1056.(Ann Smith, PhD; Chris Pat)