Prospective Epidemiologic Data

Seven epidemiologic studies have so far provided prospective data on fibrinogen and cardiovascular disease. Table 1 summarizes their methods and Table 2 shows their main results. In the Northwick Park Heart Study, white men aged 40 to 64 years were tested for a range of clotting factors, including fibrinogen. The sample was drawn from factory workers, from civil servants, and from postal workers; the participation rate was approximately 80%. Pre-existing disease was not an exclusion criterion. Follow-up of 4 years (a review by an independent panel of physicians [7]), showed that 49 persons had died, 27 from cardiovascular disease. Fatal coronary events and fibrinogen were significantly associated, which was independent of other risk factors and stronger than the analogous association for total cholesterol. Other causes of death were not related to fibrinogen levels. Fifteen of the 24 patients who died of ischemic heart disease were in the high tertile of fibrinogen (>3.2 g/L). At 10 years' follow-up [8], 109 men had had a first coronary event. Multiple regression analyses showed an association between fibrinogen and fatal or nonfatal myocardial infarction, which was again independent of other risk factors. Approximately half of all the coronary events occurred in the high tertile of fibrinogen. The association was strongest for events occurring early ( 5 years) after recruitment.

In the Gothenburg Study [12], fibrinogen, blood pressure, total cholesterol, and smoking habits were quantified in a random sample of men born in 1913. Eighty-one percent of the target population, then aged 54 years, were recruited. After a mean follow-up period of 13.5 years, there had been 92 myocardial infarctions, 37 strokes, and 60 deaths due to noncardiovascular causes. These events had been verified by a combination of interviews, review of medical records, death certificates, autopsies, as well as registers for myocardial infarction and stroke. The autopsy rate was more than 80%. Univariate analyses identified smoking, cholesterol, and fibrinogen as risk factors for ischemic heart disease, whereas blood pressure and fibrinogen were risk factors for stroke. In a multivariate analysis (adjusting for blood pressure, cholesterol, and smoking), the association between fibrinogen and cardiovascular disease was weaker but still statistically significant for stroke. The study was recently extended to a 21-year follow-up, during which time 119 myocardial infarctions, 81 strokes, and 333 deaths due to other causes had occurred [13]. Fibrinogen was again positively associated with the incidence of ischemic heart disease in a univariate analysis, whereas in the multivariate evaluation, stroke and total mortality rate were statistically associated with fibrinogen. The strength of this study is the completeness of its data collection. The sampling method used makes it likely that a representative sample of Swedish men was recruited. The study is, however, prone to type 2 error because of the small sample size. No exclusions were made for treated pre-existing disease; therefore the possibility of a confounding effect of therapy on baseline measurements is conceivable.

For the Leigh study, men aged 40 to 69 years, initially free of ischemic heart disease, diabetes, or hypertension, were recruited from one general practice in the United Kingdom [14]. Of all men eligible (n = 505), 76% were examined, whereas the rest were excluded. After a mean follow-up of 7.3 years (range, 0.1 to 16.1 years), 40 cases of myocardial infarction had occurred. Fibrinogen was positively correlated with its incidence. In hypertensive patients, for instance, the incidence was six times higher when fibrinogen levels exceeded 3.5 g/L compared to the subpopulation with values below this threshold. Multivariate analyses showed that the predictive power of all variables, in descending order, were fibrinogen, age, systolic blood pressure, total cholesterol, obesity, number of cigarettes smoked per day, and very-low-density lipoprotein levels. The odds ratio of the high compared with the low fibrinogen tertile was quite large: 21.1 (CI, 4.5 to 64.4). Unfortunately the study is burdened with serious drawbacks: It deals with a highly selected population, includes only 76% of the target population, and has a very nonuniform follow-up period; thus its results are difficult to generalize. The study has therefore been excluded from the meta-analysis.

The tenth biennial examination of the Framingham Study evaluated the interrelation of fibrinogen with smoking. Participants with a history of cardiovascular disease were excluded from this analysis. The average age at baseline was 55 years (range, 47 to 79 years). The results confirmed a dose-dependent increase in fibrinogen with smoking [15]. During a 14-year follow-up period (starting in 1968), the risk for cardiovascular disease in men and women increased quasi-linearly as a function of initial fibrinogen levels. The age-adjusted incidence in male smokers with high fibrinogen levels was doubled compared with a low-fibrinogen subgroup. As in the Northwick Park Heart Study [7, 8], the effect was more pronounced in younger men. Using the same data, fibrinogen was shown to be a risk factor for ischemic heart disease, independent of smoking or other accepted risk factors [16]. In women, the magnitude of the fibrinogen-mediated risk declined with age and had no apparent effect beyond the age of 70 years. Fibrinogen was also a risk factor for stroke in men, but not in women. The relative effect of fibrinogen was comparable to that of high blood pressure, obesity, smoking, or diabetes. A further analysis of the Framingham material [17] revealed that in men the fibrinogen risk ratio was greatest for stroke, intermediate for myocardial infarction, and smallest for peripheral arterial occlusive disease. For women, the risk ratio was greatest for ischemic heart disease.

The PROCAM study [18] examined men aged 40 to 65 years who had no history of myocardial infarction or stroke. Its design is similar to that of the Framingham Study. The PROCAM study was started in 1979, and only in 1981 was fibrinogen added to the test battery. Fifteen cardiovascular events were observed in a subsample of 1674 men during 2 years of follow-up. They were verified by questionnaire, re-examination, and by questioning families, family doctors, and hospital representatives. Ten of these events fell into the high fibrinogen tertile. When this trial was extended to 2817 men, also followed for 2 years, 55 coronary events had occurred. Twenty-nine were located in the upper and 10 were in the lowest fibrinogen tertile [19]. These reports are still preliminary in character; so far only subgroup analyses have been published. Because approximately 20 000 persons were included in the original sample, one would expect important data to be published in the near future.

The above studies either did not consider low-density lipoprotein (LDL) in their statistical models or quantified this variable by inadequate methods. Because LDL is one of the strongest predictors of ischemic heart disease, the predictive power of fibrinogen might have been overestimated by previous investigators. The GRIPS Study [20], a prospective cohort study on a random, population-based sample of men aged 40 to 60 years and initially free of cardiovascular disease, attempted to overcome this drawback. One hundred seven myocardial infarctions had occurred after 5 years of follow-up. Fibrinogen was a strong predictor in an univariate model. Using a multivariate regression model, which accounted for LDL, the relationship weakened, although it remained statistically significant. In the final model, the rank order of predictors, in descending order, was as follows: LDL, familial disposition, lipoprotein (a) (Lp[a]), high-density lipoprotein (HDL), fibrinogen, age, smoking, glucose, and blood pressure, according to a multivariate logistic regression analysis.

Meta-Analysis

A meta-analysis of the six prospective epidemiologic studies with samples representative of the general population [8, 11, 12, 16, 18, 20] was done. Because distribution of events into tertiles of fibrinogen is the only type of information available (or calculable) for all studies, the analysis had to be done on the basis of such data. Published quintiles of two studies (Gothenburg and Caerphilly-Speedwell) were recalculated into tertiles using a conservative approach that assumes a linear increase in events within the quintiles. This method tends to underestimate rather than to overestimate the effect.

Figure 1 shows all six studies depicted as fibrinogen tertiles. The event rates per 1000 persons per year differ according to the inclusion criteria of each study. Nevertheless, the impression is one of a uniform, continuous increase in risk from the lowest to the highest tertile.

View larger version (19K): [in this window] [in a new window]   Figure 1. Cardiovascular events by tertiles of fibrinogen levels in six prospective studies. Each study is represented by one symbol. Tertiles are connected by lines. Tertiles are defined as follows. Gothenburg: mean value of re-events: myocardial infarction, 3.56 g/mL; stroke, 3.70 g/mL (no standard deviations, no cut-off points provided); Framingham: cut-off points, 2.65 g/mL, 3.12 g/mL; Prospective Cardiovascular Munster Study (PROCAM): cut-off points, 2.22 g/mL; 3.12 g/mL; Northwick Park Heart Study (NPHS): cut-off points, 2.70 g/mL; 3.19 g/mL; the Caerphilly and Speedwell Collaborative Heart Disease Studies (CSCHDS): mean value, 4.09 ± 0.92 g/mL (no cut-off points provided); the Gottingen Risk, Incidence, and Prevalence Study (GRIPS): cut-off points, 3.22 g/mL; 3.91 g/mL.

View larger version (13K): [in this window] [in a new window]   Figure 2. Odds ratios for cardiovascular events in persons with fibrinogen levels in the upper tertile compared to the lower tertile. Odds ratios and 95% confidence intervals in prospective epidemiologic studies. IHD = ischemic heart disease. CSCHDS = The Caerphilly and Speedwell Collaborative Heart Disease Studies; GRIPS = The Gottingen Risk, Incidence, and Prevalence Study; NPHS = Northwick Park Heart Study; PROCAM = Prospective Cardiovascular Munster Study.

Fibrinogen is related to most other risk factors. On the one hand, this might suggest that its relationship to cardiovascular end points is indirect, merely due to these associations with "true" risk factors. On the other hand, it might indicate that other risk factors could mediate a fibrinogen effect. When, in the above studies, these factors were included in the multivariate analysis, the association between fibrinogen and cardiovascular disease usually weakened but remained statistically significant. Therefore the data are in accordance with the hypothesis that fibrinogen represents one mechanism by which various other risk factors lead to cardiovascular disease.

Clinical Data Suggesting a Causal Relationship

Clinical reports that include information on fibrinogen are abundant. Therefore only some of the most important findings with an emphasis on prospective data were selected for the subsequent discussions.

Ischemic Heart Disease

The fact that an acute myocardial infarction leads to hyperfibrinogenemia has been known since the 1950s [1-4, 24]. Fibrinogen increases progressively with the extent of coronary atherosclerosis [25-27]. Such findings can be interpreted as an acute-phase response [1-4]; however, fibrinogen levels (or plasma viscosity) remain elevated many years after an infarction [28].

A prospective study investigated fibrinogen in 120 survivors of a myocardial infarction [29]. Reinfarction occurred only in patients whose fibrinogen level exceeded 7.5 g/L during the acute event. In another prospective study, 1716 men were followed for 2 years [30]; all had survived a myocardial infarction 6 months before they entered the study. One hundred twenty-six patients had a second ischemic event during follow-up; fibrinogen was significantly elevated in this subgroup. Furthermore, statistically significant differences in fibrinogen existed between patients who survived and those who died. The relative odds for death showed a quasi-linear relationship with fibrinogen.

Stroke

Fibrinogen levels peak after an acute stroke [31]. An early longitudinal trial depicted the time course of these changes and showed that fibrinogen was excessively high in those patients who later died of the disease [32]. In the past, the phenomenon has been attributed almost exclusively to an acute-phase reaction due to brain tissue necrosis. However, plasma viscosity and fibrinogen are significantly increased in patients with transitory ischemic attacks, suggesting that fibrinogen levels are elevated before the stroke [33, 34]. In a prospective study of 625 stroke survivors [35], fibrinogen was measured during stroke rehabilitation when the acute-phase response had subsided. It was statistically higher in patients who had a second cardiovascular event within the next 2 years. The odds ratios were significantly elevated only for fibrinogen (3.67; CI, 1.31 to 11.69) and plasma viscosity (2.86; CI, 1.06 to 8.43); the effect was independent of concomitant risk factors. In another longitudinal trial, patients with progression of carotid artery lesions had significantly higher baseline fibrinogen levels compared with those with nonprogressing lesions as assessed by angiogram [36].

Peripheral Arterial Occlusive Disease

In all stages of this disease, fibrinogen is increased compared with healthy controls [37, 38]. It is often highly abnormal in the absence of overt angiographic narrowing of the arteries. Hyperfibrinogenemia may hinder blood flow by its rheologic effects, a notion that led to the concept of "rheologic claudication" [38]. Longitudinal data from patients with peripheral arterial occlusive disease showed that high fibrinogen was a predictor for reocclusion of femoropopliteal vein grafts [39]. Patients suffering from peripheral arterial occlusive disease have a high incidence of variation of the ß-fibrinogen locus leading to elevated plasma levels [40]. This phenomenon might explain (at least in part) the cause of elevated fibrinogen in these persons.

Determinants of the Fibrinogen Plasma Concentration

Numerous interactions between fibrinogen and other variables have been described in cross-sectional epidemiologic studies [41-45]. The largest such study (n = 15 803) showed fibrinogen to be 0.2 g/L higher in blacks than in whites [46]. It confirmed that women had higher values than did men. Fibrinogen increased with age, smoking, body size, diabetes, fasting serum insulin, LDL, Lp(a), leukocyte count, and menopause. It decreased with ethanol intake, exercise, HDL, and postmenopausal female hormone use. Table 3 shows the present knowledge about "determinants" of fibrinogen levels in the absence of overt disease.

Smoking is the strongest known determinant of fibrinogen levels in healthy persons. The effect is dose related [52] and reversible on cessation of smoking [53]. Fibrinogen and carboxy-hemoglobin levels are statistically correlated [54]. Plasma viscosity was elevated in male but not in female smokers [55]. Thus the relation between fibrinogen and smoking must be considered when interpreting the incidence data on cardiovascular disease. Most of the above studies did this by using multivariate logistic regression assessment. The Framingham Study [15-17] provided detailed analyses of the inter-relation of fibrinogen with smoking and cardiovascular disease and estimated that 50% of the cardiovascular harm caused by chronic smoking is mediated through its effect of increasing fibrinogen. This hypothesis seems worth further investigation.

Fibrinogen levels are higher in patients with essential hypertension than in normotensive controls [56]. Similarly, plasma viscosity is elevated in hypertensive persons, and blood pressure readings are positively correlated with this variable [57]. Even when hypertension is mild, fibrinogen levels are higher than in normotensive controls [58]. Fibrinogen is also elevated in diabetic patients [59]. Patients with microvascular involvement have higher fibrinogen levels than do diabetic patients free of such complications. The Framingham data revealed a correlation between blood sugar levels and fibrinogen [60]. Fasting glucose also correlates with fibrinogen in nondiabetic persons [46]. In diabetics with albuminuria, fibrinogen is higher than in patients without this complication [61]. Finally, fibrinogen has been shown to be an independent predictor of vascular complications in type II diabetes [62].

Pathophysiologic Mechanisms

The mechanisms by which fibrinogen may promote atherosclerosis and thrombosis are still not fully understood. Fibrinogen strongly affects hemostasis, blood rheology, platelet aggregation, and endothelial function. A hypercoagulable state clearly favors the thrombotic aspects of cardiovascular disease. Fibrinogen is the major determinant of plasma viscosity and induces reversible red cell aggregation. Both phenomena limit the fluidity of blood. The hemorheologic consequences of hyperfibrinogenemia might act at various levels: by reducing flow, by predisposing to thrombosis, and by enhancing atherogenesis [63, 64]. Platelet hyperaggregation plays an accepted role in the genesis of an atherosclerotic lesion. Fibrinogen binds to receptors on the platelet membrane which, in turn, is a precondition for aggregation in vivo [65]. Furthermore, fibrinogen is also integrated directly into arteriosclerotic vascular lesions, where it is converted to fibrin and fibrinogen degradation products; it binds low-density lipoproteins and sequesters more fibrinogen. Both fibrinogen and fibrinogen degradation products have been shown to stimulate smooth muscle cell proliferation and migration [66, 67]. These effects suggest that fibrinogen is involved in the earliest stages of plaque formation.

The role of fibrinogen as an acute-phase reactant also deserves consideration. Several aspects of atherosclerosis bear similarities to an inflammatory process [68, 69]. Thus it is conceivable that early atherosclerosis itself leads to a mild inflammatory response that elevates acute-phase proteins [37] and other variables of the acute-phase response [70]. If this is true, elevated fibrinogen levels would not represent a causal risk factor but a risk marker for early atherosclerosis.

The intense research activities in this area are discussed in more detail elsewhere [64-67]. In the absence of more experimental evidence, it seems unwise to speculate on the relative importance of one phenomenon compared to another. It seems likely that the above clinical and epidemiologic findings are the result of complex interactions between these and possibly other phenomena, which we still need to understand more completely.

Causality

Unless causality is established, a risk-related variable is, strictly speaking, only a risk marker and not a risk factor. To answer the question of which of the two fibrinogen represents, one should consider it in view of the accepted criteria for causality (Table 5). Several of these preconditions are fulfilled: The effect is strong and "dose-dependent" [16], it is consistent both in different populations and in different experimental approaches, and it is biologically plausible. The issue of temporality, however, is difficult to clarify. Even though several of the prospective studies had excluded patients with overt cardiovascular disease, the presence of subclinical disease cannot be ruled out completely. Thus it is theoretically possible that fibrinogen represents a mere marker for disease and is not a true risk factor.

As has been noted, many other characteristics exert apparent influences on fibrinogen levels. Thus specificity of the fibrinogen effect is clearly absent. Yet all of the above prospective epidemiologic studies have attempted to control for this by taking into account the associations known at the time. The results invariably demonstrated that fibrinogen was a risk factor independent of these influences.

The ultimate test for causality would, of course, be a randomized trial to therapeutically lower fibrinogen levels in patients at risk and determine the subsequent cardiovascular outcome. Such an investigation would require the availability of a drug that reduces fibrinogen levels safely and selectively; no such substance is known. The list of medications that have been shown to decrease fibrinogen in various clinical settings is long [71]. All of these, however, are marketed for other pharmacological (cardiovascular) actions. Clofibrate and derivatives, for instance, have been reported to induce relatively large fibrinogen reductions [71]. Yet, they primarily reduce blood lipids, which renders them ill suited for examining the effect of the independent fibrinogen effect. Thus the value of lowering fibrinogen in an attempt to reduce the cardiovascular risk is still unknown. Although the ultimate test for the hypothesis of a causal link between fibrinogen and cardiovascular disease does not seem possible at present, most of the crucial preconditions of causality are fulfilled. Therefore, a degree of uncertainty remains.

Conclusions

A link between fibrinogen and atherothrombogenesis is undeniable; it is the nature of this association that is debatable. Prospective epidemiologic investigations indicate that plasma fibrinogen is a powerful, independent predictor for myocardial infarction and stroke. Clinical findings suggest that fibrinogen may also be a risk factor for the sequelae of atherothrombotic events. These and other lines of evidence strongly suggest (but do not prove) that the association between fibrinogen and cardiovascular disease is, in fact, causal.

Currently several determinants of the fibrinogen level in health and disease are known. Most of these lifestyle determinants are amenable to change. Thus the need for adequate lifestyle modifications is further stressed by the recognition of fibrinogen as a risk factor. Future cardiovascular research in this area should consider fibrinogen.

Reprint Requests: Edzard Ernst, MD, University of Vienna, AKH, Wahringer Gurtel 18-20, 1090 Vienna, Austria.