Fibrinogen Is a Marker for Nephropathy and Peripheral Vascular Disease in Type 1 Diabetes: Studies of plasma fibrinogen and fibrinogen gene polymorphism in the DCCT/EDIC cohort

Study subjects were recruited from the DCCT/EDIC cohort of patients with type 1 diabetes (26,27) as part of a collaborative project between investigators at the Medical University of South Carolina (MUSC) and the DCCT/EDIC research group. The primary goal of the project is to identify risk factors and mechanisms for macrovascular disease in type 1 diabetes.

A 105.5-ml blood sample was collected after an overnight fast of at least 8 h from 957 patients consecutively appearing for their biennial exam according to the DCCT/EDIC research protocol (27). The initial 63-ml aliquot was collected and used for nonrelated research protocols. The subsequent 11-ml aliquot (nine volumes) of blood was collected using 0.13 mol/l trisodium citrate, pH 4 (one volume), as an anticoagulant. The blood sample and the anticoagulant were mixed by inverting the collection tube, which was maintained at 4°C for a period not exceeding 2 h before further processing. Erythrocytes were sedimented by centrifugation (3,000g) for 15 min. The plasma was transferred into polypropylene cryovials (Sarstedt, Newton, NC) and immediately stored at −20°C for a period not exceeding 30 days. Samples were then shipped on dry ice to the study site via overnight courier and were stored at − 70°C for up to 2 months until fibrinogen levels were determined. Samples were thawed one time and the assays run immediately. Results are reported for the analysis of 909 samples. Results from 48 analyses were unavailable due to instrument failure, insufficient sample quantity, or severe sample hemolysis. Samples taken at the same venipuncture were shipped on dry ice to the DCCT/ EDIC trial’s Central Biochemistry Laboratory, Department of Laboratory Medicine and Pathology, University of Minnesota, Minneapolis, for analysis of lipid profile, serum creatinine, and HbA_1c (26). Plasma for lipoprotein-related analyses and blood cells for DNA analyses were sent to MUSC. The studies were approved by the Institutional Review Boards at all DCCT/EDIC Clinical Study Centers and at MUSC, and all patients signed an informed consent.

Each EDIC subject undergoes a standardized annual history and physical examination, resting electrocardiograms, and blood pressure determinations. In the DCCT/EDIC cohort, HbA_1c is measured annually using high-performance liquid chromatography (26), and the fasting lipid profile (total cholesterol, total triglyceride, HDL cholesterol, calculated LDL cholesterol levels) is determined in alternate years.

Renal function was assessed in alternate years by measurement of urine albumin excretion rates from 4 h urine collections (27) as described previously (28). Ninety-eight percent of assessments were conducted within 1 year of the fibrinogen sample collection.

Seven-field stereoscopic fundus photographs were taken by certified photographers and were centrally assessed by graders unaware of the original DCCT treatment-group assignments or current clinical status. The graders used the protocol of the Early Treatment Diabetic Retinopathy Study (ETDRS) (29). The overall level of severity of retinopathy in both eyes was determined for each patient according to the ETDRS interim scale (30). Sixty percent of ETDRS assessments were done within 1 year, and 39% within 2 years, of sample collection for fibrinogen determination.

Ankle-brachial index (ABI), a measure of peripheral vascular status, is measured annually in EDIC subjects using both posterior tibial and dorsalis pedis arteries (26). A normal ABI is taken as 0.8–1.4, with high (>1.4) levels reflecting vascular rigidity and low (<0.8) levels reflecting vascular occlusion (26).

Carotid intimal-medial thickness (IMT) was measured a maximum of 3 years before this study using B-mode ultrasonography as previously described (31).

Quantitative determinations of fibrinogen level in plasma were made using a commercially available assay kit (Diagnostica Stago) based on the clotting method of Von Clauss (32) in an ST4 agglutinometer (Diagnostica Stago). Assay reagents and plasma samples were prepared according to the manufacturer’s protocol. Fibrinogen control samples (Fibrinogen Control N, Fibrinogen Control P; Diagnostica Stago) were included with each batch of samples analyzed. Interassay coefficient of variation averaged 6.4%; intra-assay coefficient averaged 3.9%.

Genomic DNA was isolated from peripheral blood using a DNA isolation kit (Gentra Systems, Minneapolis, MN). Genotyping of the G^{− 455}→A polymorphism (numbered from the transcriptional start site) in the 5′ untranslated region (UTR) was carried out by amplification of the polymorphic region by PCR followed by digestion with the restriction enzyme HaeIII. A 757-bp amplification product was generated by PCR of genomic DNA using the following primers: Fibrinogen5′UTRf: 5′ -TTTGTTGAACATTTTACCTTATGTGA-3′ and Fibrinogen5′UTRr: 5′-CATTGTCGTTGACACCTTCC-3′. Taq DNA polymerase was purchased from Gibco-BRL (Gaithersburg, MD). Reaction samples (20 μl) were denatured (3 min, 95°C), and PCR was carried out for 40 cycles (9600 thermal cycler; Perkin-Elmer, Foster City, CA). Each cycle consisted of three segments; denaturation (95°C, 30 s), primer annealing (60°C, 60 s), and primer extension (72°C, 60 s, 1.5 mmol/l MgCl₂); an additional extension period (5 min, 72°C) was applied upon completion of the 40 cycles. This PCR product was directly digested with HaeIII according to the manufacturer’s instructions (New England Biolabs, Beverly, MA), in 25-μl volumes for 6 h at 37°C. Digested products were resolved on 2% agarose gels and visualized with ethidium bromide under ultraviolet light. Restriction digestion of the 757-bp PCR product with HaeIII results in two DNA fragments of 570 and 187 bp in GG homozygotes and a single fragment of 757 bp in AA homozygotes, reflecting lack of the restriction site.

Urinary albumin excretion rates were log transformed because they were not normally distributed. Simple Pearson correlation was performed between fibrinogen and other continuous variables. Single and multiple linear regression models were used to assess the association between fibrinogen and other continuous, clinically relevant variables, where the clinical variables were considered as dependent and fibrinogen as independent. Multiple linear regression analyses tested whether fibrinogen was significantly associated with the clinical variables after adjusting for covariates such as age, sex, duration of diabetes, DCCT treatment group, waist-to-hip ratio, HbA_1c, BMI, triglycerides, and total and LDL cholesterol. Deviation from Hardy-Weinberg equilibrium was assessed by χ² (2 degrees of freedom [df]). The hypothesis that clinical variables, fibrinogen level, and vascular complications differed among β-fibrinogen genotypes was tested using ANOVA and by ANCOVA after considering the confounding effects of concomitant variables while main effects were tested. ANCOVA F test was used to determine if two or more adjusted least-square means were significantly different between or across genotypes. Bonferroni adjustment was applied for multiple comparisons. SAS software (version 8; SAS Institute, Cary, NC) was used in all statistical analyses. All significance tests for the comparisons were two sided, and results were considered significant at P < 0.05.

Clinical characteristics of the 909 patients are delineated in Table 1. These characteristics do not differ significantly from those in the 1,301 patients who comprise the overall DCCT/EDIC cohort (data not shown). The study group comprised nearly equal numbers of men and women of similar age who were randomized equally to the intensive and conventional treatment subgroups during the DCCT. Proportionately more men than women were hypertensive, and men exhibited significantly higher systolic and diastolic blood pressure levels (P < 0.0001). Plasma levels of triglycerides and LDL cholesterol were significantly higher in men than in women; HDL cholesterol levels were significantly lower (all P < 0.0001). The severity of retinopathy did not differ between men and women; however, nephropathy, measured as the log albumin excretion rate, was marginally increased in men (P = 0.01). Peripheral vascular disease was more advanced in men than in women as demonstrated by significantly increased IMT in both the common carotid and internal carotid arteries, and by a significantly elevated ABI (P < 0.001). The percentage distribution of fibrinogen genotypes was similar for both men and women.

Table 2 shows the relationship between plasma fibrinogen and other parameters using univariate regression analysis. When all patients were considered, statistically significant correlations between fibrinogen and age, BMI, triglycerides, total and LDL cholesterol, current HbA_1c, diastolic blood pressure, log albumin excretion rate, and the average ABI were observed. However, close examination of the univariate regressions revealed marked differences between men and women. In men, all of these correlations remained statistically significant except for blood pressure. In addition, the relationship with waist-to-hip ratio was statistically significant, and, in fact, was the strongest predictor of plasma fibrinogen level (r = 0.20; P < 0.001). In women, fibrinogen was correlated with BMI, total and LDL cholesterol, HbA_1c, and both diastolic and systolic blood pressure, but not with waist-to-hip ratio, age, or triglycerides.

Table 3 shows discontinuous analyses for several variables. Female patients had higher mean plasma fibrinogen level than males (321 ± 5 vs. 300 ± 5 mg/dl; P = 0.0045). In both sexes, the relationship between fibrinogen and HbA_1c was nonlinear, with lower fibrinogen levels in patients in the lower quartile with HbA_1c values <7.3%, and a higher mean value in the highest quartile with HbA_1c ≥9.0% (Table 3). There was no significant effect of smoking status on plasma fibrinogen in these patients.

Univariate regression analyses also demonstrated that fibrinogen was related to markers of nephropathy and peripheral vascular disease and that the relationship was dependent on sex. Fibrinogen was positively correlated with urinary albumin excretion rate in the group as a whole (Table 2); however, this relationship was statistically significant only in men (r = 0.13; P = 0.003) and not in women (Fig. 1). It was also instructive to examine this relationship with albuminuria as a discontinuous variable in Fig. 2. Whereas there was a tendency for fibrinogen to increase with the development of microalbuminuria (albumin excretion rate 40–299 mg/24 h) compared with normal individuals (<40 mg/24 h), this difference did not achieve statistical significance. In contrast, there was a marked 1.5-fold elevation in plasma fibrinogen noted in patients with proteinuria (albumin excretion rate ≥300 mg/24 h). These differences in mean fibrinogen level as a function of albuminuria category were statistically significant in women and in the group as a whole, but did not quite achieve significance in men. Thus, there were indications in both men (Fig. 1) and women (Fig. 2) that fibrinogen was associated with nephropathy. The clear and pronounced elevation in patients with proteinuria (and not with microalbuminuria) suggests that high fibrinogen levels were acquired with more severe degrees of albuminuria and did not precede the development of microalbuminuria. Thus, elevated fibrinogen could contribute to the acceleration of atherogenesis that accompanies proteinuria. There was no relationship between fibrinogen level and the severity of another microvascular complication, retinopathy, as assessed by ETDRS retinopathy score in Table 2.

Plasma fibrinogen was also correlated with the ABI, a measure of peripheral vascular disease, and this relationship also exhibited sex dimorphism (Table 2). In women, the relationship between fibrinogen and the average value (Fig. 3) of the four ABI measurements in each individual (i.e., right and left dorsalis pedis and posterior tibial arteries) was highly significant. High fibrinogen levels were associated with lower ABI measurements (i.e., more severe peripheral vascular disease) with respect to both the dorsalis pedis (r = −0.12; P = 0.0160) and posterior tibial (r = −0.15; P = 0.0038) arteries (data not shown). However, in men, plasma fibrinogen was not significantly correlated with ABI measurements (Fig. 3, Table 2).

We determined whether plasma fibrinogen concentration was independent of various clinical parameters and of vascular disease complications. The higher fibrinogen level in women and the relationship between fibrinogen and HbA_1c remained highly significant after adjusting for multiple covariables (Table 3).

Multiple regression analyses were conducted to determine whether plasma fibrinogen was independently correlated with various clinical parameters and diabetes complications. Multiple regression equations for albumin excretion rate and the average ABI are depicted in Table 4. Plasma fibrinogen was found to be an independent predictor of the albumin excretion rate in men (P = 0.016) but not in women. Other factors that were statistically significant in the multiple regression equation for albumin excretion were age, DCCT treatment group, HbA_1c, systolic blood pressure, triglycerides, and LDL cholesterol. As regards peripheral vascular disease, plasma fibrinogen was independently associated with the average ABI measurement in women (P = 0.009), but not in men. Other risk factors that achieved statistical significance in the regression equation for ABI included BMI, waist-to-hip ratio, and systolic blood pressure in men, but only BMI in women. Age, DCCT treatment group, current HbA_1c, diastolic blood pressure, triglycerides, LDL and HDL cholesterol, and smoking status did not otherwise significantly contribute to the multiple regression equation.

The relationship between fibrinogen and carotid wall thickness, another quantitative measure of macrovascular disease, was also investigated. Unlike the marker for peripheral vascular disease, there was no association between plasma fibrinogen and either common carotid IMT or internal carotid IMT determined by duplex ultrasonography in men or women (Table 2). Similarly, fibrinogen was not significantly related to common or internal carotid IMT after adjusting for multiple covariables (including age, DCCT treatment group, HbA_1c, BMI, waist-to-hip ratio, blood pressures, and lipid panel) in multiple regression analyses (data not shown).

The distribution of β-fibrinogen G⁻⁴⁵⁵→A genotypes was 60% for GG, 35% for GA, and 6% for AA% for all subjects and did not differ significantly between sexes (Table 1). These frequencies conformed to Hardy-Weinberg equilibrium and were similar to those previously reported in nondiabetic subjects. This polymorphism in the 5′ promoter region of the β -fibrinogen gene did not influence fibrinogen levels. There were no significant differences in mean plasma fibrinogen levels according to genotype, either before or after adjusting for covariables including age, sex, BMI, current HbA_1c, and smoking status (Table 5). Furthermore, in measured genotype analyses (Table 5), the fibrinogen gene polymorphism was not associated with urinary albumin excretion rate, retinopathy score, ABI, or common carotid IMT. However, the association with IMT of the internal carotid artery did just reach statistical significance (ANOVA P = 0.050; ANCOVA P = 0.045). Homozygosity for the more common G allele was associated with increased IMT (0.687 ± 0.008 mm) compared with homozygosity for the less common A allele (0.642 ± 0.026 mm), with heterozygotes having an intermediary mean value (0.659±.010 mm) (post hoc pairwise comparisons GG versus GA, P = 0.037; GG versus AA, P = 0.1046; GA versus AA, P = 0.544).

Our cross-sectional analyses indicated that fibrinogen levels are related to microvascular and macrovascular disease in type 1 diabetes. Fibrinogen levels were associated with nephropathy status in type 1 diabetic subjects (Table 2; Figs. 1 and 2). In men but not women, there was a significant correlation between fibrinogen levels and albumin excretion rate across the entire range of albumin loss (r = 0.10; P = 0.0039), with no evidence of a threshold for increased fibrinogen. In women, discontinuous analysis demonstrated that, whereas fibrinogen levels were similar in normoalbuminuric and microalbuminuric subgroups, there was a significant increase in plasma fibrinogen with macroalbuminuria. Therefore, in both men and women, fibrinogen was related to nephropathy as manifested by the severity of albuminuria. The results of previous studies in diabetic subjects have not been consistent and have shown varying associations between the presence of micro- and/or macroalbuminuria and elevated fibrinogen concentrations (5–11, 33–35), most probably as a result of the varying influence of multiple additional factors and drug treatments.

In the present study, analysis of covariance was used to control for effects of multiple other cardiovascular risk factors (including HbA_1c) and showed that hyperfibrinogenemia is independently associated with nephropathy status and albumin excretion rate. It is not clear, however, whether hyperfibrinogenemia occurs secondary to the onset of nephropathy or is a primary factor that antedates microalbuminuria. Further prospective evidence is required to confirm hyperfibrinogenemia as an independent predictor of future diabetic nephropathy. Our longitudinal follow-up in the EDIC cohort (in progress) will address this issue. However, even if hyperfibrinogenemia is a secondary abnormality, it could still contribute to a worsening of existing nephropathy and to the associated increase in vascular disease risk associated with albuminuria.

Regarding other covariables, smoking status did not affect plasma fibrinogen (P = 0.37) in our patients, contrary to previous studies in both the general population and type 1 diabetic subjects (2,5, 14,15). It is possible that this represents a type 2 error due to the limited number of current smokers in the EDIC study. Sex did affect fibrinogen levels, with higher values observed in women, as reported in other studies. Multiple reports (36–38), but not all (39,40), have demonstrated that fibrinogen levels are increased in diabetes, although the mechanism has not been clarified. The results of the present study show that one factor contributing to this relationship may be hyperglycemia, since fibrinogen levels are elevated in patients as a function of poor glycemic control (Tables 2 and 3).

In contrast to the association with diabetic nephropathy, fibrinogen levels were not related to another microvascular complication, retinopathy (Table 2). Although retinopathy and nephropathy are often concomitant, this is not always the case, and the data indicate that different risk factors may influence the development of microvascular complications at different sites.

With respect to macrovascular disease, the association between high plasma fibrinogen and peripheral vascular disease in type 1 diabetes is a novel observation. We found an inverse relationship between fibrinogen and the ABI, which is a measure of subclinical peripheral vascular disease (Table 2). This relationship was statistically significant in the group as a whole, but this was due to a significant correlation in women only. Furthermore, hyperfibrinogenemia was independently associated with ABI in women after controlling for conventional vascular risk factors and other covariables (Table 4). Although the magnitude of the independent association is not large, it is possible that this may be an underestimate of the true impact of fibrinogen. This is because the range of ABI values was quite narrow, with few values <0.8, indicating that, as a whole, the DCCT/EDIC cohort had not yet developed clinically overt peripheral vascular disease. Additional measurements of fibrinogen as the incidence of more severe and clinically overt peripheral arterial disease increases may help refine the degree of risk conferred by fibrinogen. Furthermore, while it is statistically valid to determine the independent association with ABI, biologically, fibrinogen may interact with other risk factors to contribute to the overall vascular disease risk. These results in type 1 diabetes are consistent with previous reports in nondiabetic patients with peripheral vascular disease, where an inverse correlation between fibrinogen and ABI values has also been demonstrated (41–46).

We did not observe an association between fibrinogen levels and carotid artery IMT determined by ultrasound. Carotid IMT has previously been shown to be associated with conventional vascular risk factors in the EDIC cohort, including total cholesterol, LDL cholesterol, BMI, and waist-to-hip ratio in men, systolic blood pressure, and smoking (31). It should be pointed out that the measurement of carotid IMT was performed up to 3 years before measurement of fibrinogen and was also in the relatively normal range, which are potential sources of error. However, carotid IMT has been shown to be associated with the presence of coronary artery disease, and it is of interest that no association between fibrinogen and electocardiographic or symptomatic evidence of coronary artery disease was demonstrated in one previous study in type 1 diabetes (5). Further measurements of plasma fibrinogen over time, combined with ongoing assessment of events and measurement of coronary artery disease by the DCCT/EDIC study, will more directly examine whether there is a relationship between fibrinogen and coronary artery disease in type 1 diabetes.

The results of the present study show that the β-fibrinogen G⁻⁴⁵⁵→A polymorphism does not influence fibrinogen levels in type 1 diabetes (Table 5). We chose to examine the G⁻⁴⁵⁵→A polymorphism because of recent in vitro evidence suggesting that this polymorphism is associated with altered transcriptional activity in β-fibrinogen promoter/luciferase reporter constructs and influences the binding of a specific, but as yet unidentified, nuclear protein that regulates transcription (47). Our results are in contrast with those of previous studies (22, 23,25, 48–50) that have demonstrated an association between this polymorphism and fibrinogen levels. The explanation for the difference in the results of these studies may involve heterogeneity in the genetic factors regulating fibrinogen levels in diabetic and nondiabetic subjects. Moreover, multiple other variables in diabetes could influence the interaction between genotype and circulating fibrinogen levels. Along these same lines, fibrinogen gene polymorphisms have been shown to influence the effect of smoking on fibrinogen levels in nondiabetic subjects (21,49); however, we found no evidence of an association between the β-fibrinogen G⁻⁴⁵⁵→A polymorphism and fibrinogen levels in either current smokers or nonsmokers.

Evidence indicates a role for genetic factors in the pathogenesis of nephropathy in type 1 diabetes (51,52). The results of this study show that the β-fibrinogen gene polymorphism does not contribute to this genetic susceptibility (Table 5). Similarly, there was no association between the G⁻⁴⁵⁵→A polymorphism and ABI values in type 1 diabetes. The β-fibrinogen gene polymorphism was, however, independently associated with internal carotid artery IMT. This observation, together with the finding that circulating fibrinogen was not affected by this promoter polymorphism, suggests that the relationship with internal carotid IMT is the result of linkage disequilibrium with some other adjacent causal gene.

Last, an insulin resistance syndrome score has been developed based on clinical risk factors in patients with type 1 diabetes and was validated using euglycemic-hyperinsulinemic clamp studies (53). Fibrinogen levels were significantly associated (P < 0.0001) with this insulin resistance syndrome score (data not shown). Waist-to-hip ratio, triglycerides, hypertension status, and glycemic control are factors used to define this score and predict glucose disposal rates in type 1 diabetic patients. In the present study, fibrinogen levels were correlated with BMI, waist-to-hip ratio, HbA_1c level, and triglycerides in men, and with BMI, HbA_1c level, and both systolic and diastolic blood pressure in women. Thus, while there is considerable overlap of the traits, sex seems to influence which are presented. The present study demonstrates conclusively, however, that elevated plasma fibrinogen correlates with several traits that comprise the insulin resistance syndrome in patients with type 1 diabetes and suggests that elevated fibrinogen may be a component of the trait cluster.

This work was supported by a program project grant (principal investigator: W.T.G.) cofunded by the National Institutes of Health/National Heart, Lung, and Blood Institute (PO1-HL55782) and the Juvenile Diabetes Foundation International. Additional funding was supplied by the Department of Veterans Affairs Research Service (R.L.K.) and by a research contract from the Division of Diabetes, Endocrinology, and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the National Institutes of Health (DCCT/EDIC).

The authors gratefully acknowledge Kirby Smith for DNA extractions, Jenny Smith and Leslie Nicholson for study coordination, and John Bercik, the General Clinical Research Center (GCRC) (M01-RR-1070) computer systems manager, for computer database support. The authors are also grateful to the patients in the DCCT/EDIC for their long-term participation in this important trial.