Association of Fibrinogen with Glycemic Control and Albumin Excretion Rate in Patients with Non-Insulin-Dependent Diabetes Mellitus

Methods

We identified our study base on 1 October 1988. The initial cohort consisted of all 1967 persons with non–insulin-dependent diabetes mellitus who lived in the primarily rural area of Casale Monferrato in northern Italy. According to the 1987 intercensual estimate, the total population of Casale Monferrato at that time was 93 477 persons [13]. In a previous survey [14], by using the multiple-sources capture-recapture method, we estimated that the completeness of ascertainment was high (80%). In the same survey, we found that after adjustment for undercounting, the prevalence of non–insulin-dependent diabetes mellitus was 2.67%. Approximately 10% of patients were receiving dietary therapy, 76% were being treated with oral hypoglycemic drugs, 10% were being treated with insulin, and 3.5% were being treated with hypoglycemic drugs and insulin.

In 1991-1992, we reinvestigated this cohort. As previously reported [15], patients with non–insulin-dependent diabetes mellitus were invited by telephone to be examined at the diabetes clinic; 284 patients could not go to the clinic and were examined at home. Of the 1967 persons in the original cohort, 1574 (80%) were reinvestigated. Of the remaining 393 patients, 280 had died (160 women and 120 men), 7 had left the area, and 106 could not be located. At the time the cohort was identified, the duration of diabetes was slightly shorter in patients who became study participants than in patients who did not (mean ±SD, 8.2 ± 6.6 years compared with 9.5 ± 8.2 years). Study patients also had a lower frequency of hypertension (47.7% compared with 51.0%).

To standardize data collection and blood pressure measurement, three investigators (after receiving training) interviewed and examined all patients. We collected information on smoking habits, personal medical history, family history of diabetes and hypertension, and use of medications. (Copies of the questionnaire used to collect this information are available from the authors). Patients were asked to refrain from smoking and strenuous activity for at least 24 hours before the visit. A beam balance and stadiometer were used to measure height and weight while patients were wearing indoor clothing and no shoes.

Blood pressure was measured with mercury sphygmomanometers to the nearest 2 mm Hg while patients were seated. Measurements were obtained from the right arm at the start of the examination and then three consecutive times after a 5-minute rest. Reported values are the average of the second and the third readings [phase 1 for systolic pressure and phase 5 for diastolic pressure]. Hypertension was defined as 1) systolic blood pressure at least 140 mm Hg or diastolic blood pressure at least 90 mm Hg or 2) treatment with antihypertensive drugs. Smoking habit was classified into one of three categories: nonsmoker, for patients who had never smoked; ex-smoker, for patients who had stopped smoking at least 1 month before the visit; and smoker, for patients who currently smoked.

All but 53 patients collected an overnight, timed urine sample (individual oral and written instructions were provided). When evidence (such as dysuria, urgency, presence of nitrites, and positive bacterial urine culture) suggested urinary tract infection, the urine was discarded and replaced by sterile urine that was collected after the patient received antibiotic treatment. Urinary albumin concentration was measured by using the nephelometric method (Behring Nephelometer Analyzer, Behring Institute, Marburg, Germany; interassay and intraassay coefficients of variation, 4.08% and 5.06%, respectively). Albumin excretion rate was calculated as the timed urine volume multiplied by the albumin concentration. Normoalbuminuria was defined as albumin excretion less than 20 µg/min, microalbuminuria was defined as albumin excretion of 20 to 200 µg/min, and macroalbuminuria was defined as albumin excretion greater than 200 µg/min.

We collected venous blood samples while patients were fasting to determine levels of plasma glucose, triglycerides, total cholesterol (measured by enzymatic-colorimetric methods), high-density lipoprotein (HDL) cholesterol (measured by enzymatic-colorimetric method after precipitation with Mn⁺⁺), fibrinogen (measured by the Clauss method [16] [Baxter Diagnostic AG, Dudingen, Switzerland]; interassay and intraassay coefficients of variation, 3.5% and 4.2%, respectively; normal range in our patients, 1.5 to 3.5 g/L), and hemoglobin A_1c (measured by high-performance liquid chromatography, Daiichi, Menarini, Japan; laboratory reference range, 3.8% to 5.5%). All plasma and urine measurements were done in a central location.

We used SAS software (version 6.03, SAS Institute, Cary, North Carolina) for all statistical analyses. We obtained mean age- and sex-adjusted fibrinogen levels using the least-squares mean option of the general linear models procedure, with age as a continuous variable. Variables that were not normally distributed (albumin excretion rate and triglyceride levels) were analyzed after they were transformed into logarithms. We did a multiple linear regression analysis to examine whether the association of fibrinogen level with hemoglobin A_1c value and albumin excretion rate was independent of the association between fibrinogen level and cardiovascular risk factors. Hemoglobin A_1c value; body mass index; and levels of total cholesterol, HDL cholesterol, and triglycerides were included as continuous variables. Sex was a dichotomous variable coded as 0 for men and 1 for women, and age was used as a continuous or categorical variable. We also evaluated the effect of the interaction between age and sex on the estimates of the variables. We used albumin excretion rate as a continuous or categorical variable (<20, 20 to 200, or >200 µg/min). Hypertensive status was coded as a dichotomous variable (0 if absent and 1 if present), and smoking habit was defined as 0 if the patient was a nonsmoker, 1 if the patient was an ex-smoker, and 2 if the patient was a current smoker. Results of multiple linear regression analysis are shown as the regression coefficient ± SE (b values; the estimated difference in fibrinogen levels as a result of a unit increase in the independent variable) and P values of the regression coefficient (derived using a t-test). Means are expressed ±SD.

Results

Plasma fibrinogen levels were measured in 1525 of the 1574 study patients (669 men and 856 women). Mean age at study entry was 67.3 ± 10.3 years for men and 70.7 ± 10.7 years for women; mean age at diagnosis was 56.4 ± 11.1 years for men and 58.8 ± 12.0 years for women; and mean duration of diabetes was 10.8 ± 6.3 years for men and 11.9 ± 7.5 years for women. The prevalence of hypertension was 81.0%, and the prevalence of diabetic nephropathy was 49.7%. The slightly skewed distribution of plasma fibrinogen levels ranged from 0.7 to 9.6 g/L (mean, 3.6 ± 0.9 g/L; mode, 3.2 g/L). A sample of 200 normal persons (who came from the same area in Italy as the study patients) had lower fibrinogen values (mean, 2.5 ± 0.5 g/L) than did study patients. Plasma fibrinogen levels were less than 2.50 g/L in 8.1% of patients, 2.50 to 3.49 g/L in 41.6% of patients, 3.50 to 4.49 g/L in 36.3% of patients, 4.50 to 5.49 g/L in 10.9% of patients, and 5.50 g/L or greater in 3.1% of patients. In 50.3% of patients, plasma fibrinogen levels were greater than 3.5 g/L. Fibrinogen levels slightly differed between men and women (3.6 ± 0.9 g/L in men and 3.7 ± 0.9 g/L in women).

In men, plasma fibrinogen level increased with age (P < 0.001). In women, plasma fibrinogen level increased by almost 0.4 g/L only between premenopausal and postmenopausal ages (age groups 30 to 49 years and ≥ 50 years), but no significant linear trend was detected (Table 1).

After adjustment for age and sex, fibrinogen level was significantly and linearly related to hemoglobin A_1c value and albumin excretion rate (P < 0.001 for both comparisons) (Table 2).

We did a multiple linear regression analysis to determine whether both hemoglobin A_1c value and albumin excretion rate were associated with plasma fibrinogen level independently of the classic cardiovascular risk factors (age; sex; hypertensive status; levels of total cholesterol, HDL cholesterol, and triglycerides; body mass index; and smoking habit). Data on all variables were available for 1397 patients. Fibrinogen level was significantly associated with both hemoglobin A_1c value (b ± SE = 0.06 ± 0.01; P < 0.001) and albumin excretion rate (b = 0.09 ± 0.03; P = 0.005), independently of age (b = 0.01 ± 0.002; P < 0.001), body mass index (b = 0.01 ± 0.005; P = 0.006), sex (b = 0.11 ± 0.06; P = 0.06), smoking habit (b = 0.08 ± 0.04; P = 0.05), hypertensive status (b = 0.13 ± 0.07; P = 0.05), and total cholesterol level (b = 0.04 ± 0.02; P = 0.07). This model explained 6% of the variability in plasma fibrinogen levels. Adding HDL cholesterol and triglyceride values to previous independent variables did not significantly modify the estimates of the variables. Results were similar when the multivariate analysis included age as a categorical variable and when the interaction between age and sex was examined.

Discussion

We found that patients with non–insulin-dependent diabetes mellitus had an elevated prevalence of hyperfibrinogenemia and that plasma fibrinogen level was independently associated with hemoglobin A_1c value and albumin excretion rate.

Some studies [5, 6] but not others [7] found a positive association between plasma fibrinogen level and diabetes or blood glucose level. Bias cannot be ruled out, however, because none of these studies involved a population-based cohort of persons with non–insulin-dependent diabetes mellitus. Hyperfibrinogenemia in diabetes has been reported to be caused by an increased synthesis of fibrinogen that is not compensated for by a proportional increase in clearance of fibrinogen. These abnormalities have been associated with insulin deficiency and have been corrected with insulin [17], suggesting that hyperfibrinogenemia is an expression of poor glycemic control. It has been reported [18] that fibrinopeptide A (a peptide that is released from fibrinogen when it is transformed into fibrin) is positively related to blood glucose. One study [19] recently suggested that hyperfibrinogenemia is one way by which hyperglycemia activates coagulation. Therefore, both epidemiologic and clinical findings support the hypothesis that poor glycemic control may lead to thrombophilia, a condition that might be involved in the increased cardiovascular risk in patients with diabetes.

We found that albumin excretion rate was also associated with plasma fibrinogen level, independent of other cardiovascular risk factors (for example, age, sex, hypertensive status, total cholesterol level, smoking habit, and body mass index). The association between albumin excretion rate and plasma fibrinogen level has been described in patients with non–insulin-dependent diabetes mellitus [8-10], but the evidence was not conclusive because no multivariate analyses were done to control for confounding variables, such as age, body mass index, hypertension, and smoking. The positive association that was seen between albumin excretion rate and fibrinogen level could help explain the increased cardiovascular-related morbidity and mortality in diabetic patients with microalbuminuria and macroalbuminuria [11, 12]. The follow-up evaluation of mortality in our population-based cohort (in progress) may help to evaluate the role of hyperfibrinogenemia as a cardiovascular risk factor.

Ours is the first study of plasma fibrinogen levels in non–insulin-dependent diabetes mellitus to include a large population-based cohort for which detailed clinical and metabolic characteristics are available. Studies done on selected groups of patients who attend diabetes clinics are generally biased because these groups are not representative of the overall population of persons with diabetes. For example, one study done in Casale Monferrato [20] found that general practitioners selectively referred younger patients and patients with worse glycemic control to the diabetes clinic. In our study, the high completeness of ascertainment in the study base and the many patients recruited suggest that selection bias was limited.

An unresolved issue is whether fibrinogen is causally related to atherothrombosis or merely reflects the inflammatory reaction that occurs in progressive atherosclerosis. The resolution of this issue is not important for prognosis, but a causal relation would have implications for intervention strategies. Furthermore, intervention studies are needed to establish whether the reduction of elevated fibrinogen levels must be included in the targets for decreasing the number of vascular complications of diabetes.

On the basis of our study findings, we conclude that hyperfibrinogenemia could be a mechanism of the increased cardiovascular risk faced by patients with non–insulin-dependent diabetes mellitus.

Dr. Bargero: Servizio di Diabetologia, Ospedale Santo Spirito, viale Giolitti 2, I-15033, Casale Monferrato, Alessandria, Italy.