Βeta-fibrinogen gene promoter A −455 allele associated with poor longterm survival among 55–71 years old Caucasian women in Finnish stroke cohort

The Helsinki Stroke Aging Memory (SAM) cohort comprises of a consecutive series of all Finnish (Caucasian) patients with suspected stroke admitted to Helsinki University Central Hospital (n = 1622) between the 1^st of December 1993 and the 30^th of March 1995. A total of 486 (246 men, 240 women) living patients aged 55 to 85 years (mean age 71.3 years) were recruited for the study 3 months after experiencing an ischemic stroke - more detailed descriptions of the study series and methods have been described elsewhere [17, 18, 31, 32].

A structured medical and neurological history was recorded during hospitalization for the index stroke that was also the time point of laboratory sample collection and blood pressure measurements in addition to MRI investigation. History of cardiac risk factors (myocardial infarction, cardiac failure, atrial fibrillation), arterial hypertension, peripheral arterial disease, and diabetes was investigated by reviewing all available hospital charts, in addition to a structured interview of the subject and a knowledgeable informant. Hypertension was defined as blood pressure ≥160/95 mm Hg. Smoking habit was scored as non-smokers or smokers, with non-smokers only those who had never smoked. Diabetes was defined as previously documented diagnosis, current use of insulin or oral antidiabetic medication, or fasting blood glucose >7.0 mmol/L. Laboratory analyses included total and HDL cholesterol, triglycerides, and fasting blood glucose. Total cholesterol was considered high at >6.5 mmol/L. Carotid stenosis was defined as >50% stenosis of vessel diameter. 383 patients (78.8%) of the SAM cohort underwent a brain MRI investigation, and 371 patients had both MRI and genotype data. Infarct subtype definitions for lacunar (LAI) and large-vessel infarcts (LVI) have been presented earlier [33]. There were 116 (23.8%) of patients with missing values in MRI, cause of death, biochemical or genotype data. The final follow-up study population did not differ in risk factor, stroke type, demographic, or medical parameters from the original cohort. The analyzed and excluded patients did not differ in either vascular risk factors or stroke type (Table 1). Data on survival were collected over a 12.5 year period during which 347 (71.4%) of the patients deceased. MRI, genotype and cause of death data was available for 262 (53.9%) patients. The study was approved by the ethics committee of the Helsinki University Central Hospital, Helsinki, Finland. The study was explained to the patients, and informed consent was obtained.

DNA was separated from frozen blood samples according to standard procedures. The genotypes of the FGB −455 G/A polymorphism were detected by polymerase chain reaction (PCR) and restriction enzyme digestion, followed by poly-acrylamide gel electrophoresis. Primer sequences and the PCR protocol have been previously described in detail [34].

Data was analysed using SPSS/WIN (version 20.0, SPSS Inc.) software. The number of A/A homozygotes of the FGB −455 G/A polymorphism was small among SAM cohort patients and therefore A/A homozygotes and G/A heterozygotes were grouped together to form the A + genotype variable of the of the FGB −455 G/A polymorphism. The data was analysed as whole and in subgroups that were formed using sex and the mean age (71 years) to divide the data into age groups (55–71 and 72–85 years old) in intention to study the sex and age dependence. To study the association between risk factors including FGB −455 G/A A + genotype and survival we first analysed the data by Kaplan-Meier (log rank) survival analysis and graphical approach was used in verification that the proportional hazards assumption was met. Subsequently, the Cox regression proportional hazards model was used to study the association between FGB −455 G/A A + genotype and survival. First we obtained crude hazard ratios (HRs) and confidence intervals (CIs) and then we used multivariate Cox regression analysis in further statistical testing of genotype-survival association with age, history of previous arrhythmias, atrial fibrillation (AF), carotid stenosis (CS), elevated CRP (>10 mg/l), diabetes, hypercholesterolemia, hypertension, hypertriglyceridemia, history of myocardial infarction (MI), sex and smoking pooled into forced and forward stepwise likelihood ratio (LR) models. To further adjust the analyses we stratified Cox regression models (forced and forward stepwise) with each modifier (history of previous arrhythmias, AF, CS, elevated CRP (>10 mg/l), diabetes, hypercholesterolemia, hypertension, hypertriglyceridemia, history of MI and smoking) separately and used age and sex as confounders. The data concerning the history of arrhythmias, AF, CS, diabetes, hypertension, hypercholesterolemia and hypertriglyceridemia was missing in 118 (24.3%) of patients. Furthermore, based on Little’s MCAR (missing completely at random) test (p < 0.05) we concluded that the data was not missing completely at random. Overall summary of missing values showed that each case with missing values had, on average, missing values on roughly 3.37 of the 10 variables suggesting that listwise deletion would lose much of the information in the data. The data missingness pattern was nonmonotone and the random number generator was used to simulate missing values for each observation that were imputed with fully conditional specification (FCS) Markov Chain – Monte Carlo (MCMC) method assuming multivariate normality and five imputations was suitable in graphic check for FCS convergence. Then we analyzed the complete data with Cox regression model to check differences in survival in total population and in subgroups with pooled parameter estimates and we also checked of the variation in the regression coefficient estimates from imputation to imputation, and against the original data. The results were not altered after imputations. Survival analyses for LAI or LVI subgroups, stroke types (TOAST) and specific causes of death were also completed. According to power calculations our study population has the power to show hazard ratios >2.5 to be statistically significant with 0.80 probability (power) when the Type I error probability is 0.05 [35].

There were 371 patients with the FGB −455 G/A polymorphism genotype and clinical data available. The study population comprised 50.7% (186) male patients. The mean age was 71.0 years (SD ± 7.81) and the data was normally distributed. There were 186 smokers (50.1%) in the study population (84.9% men and 15.1% women). Allele distributions were in Hardy-Weinberg equilibrium (p = 0.74). Genotype distributions were 64.7% for GG, 32.1% for GA, and 3.2% for AA. These frequencies closely correspond to the population frequencies among whites; in the ECTIM study, the allele frequencies were 65.9%, 29.1%, and 4.9%, respectively [27]. The main demographic parameters are represented in Table 1.

Of the series 262 (70.6%) patients with genotype and MRI data died during an average of 12.5 years of follow-up. Cardiovascular diseases were the cause of death in the majority of cases (33.6% of men and 30.5% of women). There were no statistically significant differences between sexes in the causes of death; 37.7% of women and 29.5% of men had brain related causes of death and 34.6% of women and 27.3% of men had ischemic brain insults. However, 72–85 years old men had more often brain related (35.4% vs. 20.8%, p = 0.007) and ischemic brain insults (32.9% vs. 18.9%, p = 0.01) as cause of death compared to 55–71 years old men (Table 2). In addition, ischemic brain insults and cardiovascular diseases were more often the cause of death among women aged 72–85 years compared to younger age group (36.7% vs. 23.8%, p = 0.06; 33.9% vs. 19.0, p = 0.04, respectively). Furthermore, cerebral bleeding was the cause of death more often among 55–71 years old women compared to older age group (14.3% vs. 2.8%, p < 0.001) (Table 2).

As expected, age was associated (p < 0.001) with survival in Kaplan-Meier log rank analyses. In the multivariate Cox regression model, age (HR = 1.08, 95% CI: 1.06-2.00, p = 0.005) and diabetes (HR = 1.39, 95% CI: 1.05-1.83, p = 0.03) associated significantly with poor survival. However, significant association between diabetes and risk of cardiac death (HR = 1.59, 95% CI: 0.97-2.59) was not observed in multivariate Cox regression analysis. Smoking had no statistically significant (p = 0.06) effect on survival in the Kaplan-Meier log rank analysis that was also observed in the multivariate Cox Regression model (HR = 1.26, 95% CI: 0.97-1.65) and it was not dependent on sex. In contrast, hypercholesterolemia or elevated cholesterol levels known to increase risk for further cardiovascular events, but paradoxically associates with lower mortality in old age had a positive (HR = 0.65, 95% CI: 0.43-0.99, p = 0.05) impact on longterm post stroke survival in Cox regression model [36].

We did not observe statistically significant (p-values >0.05) association between the FGB −455 G/A A-allele and poor survival in the total study population, nor in men or women separately (Kaplan-Meier log rank and Cox regression model). In addition, genotype-by-smoking interaction on survival in the total study population was not observed in the Kaplan-Meier log rank analysis that remained insignificant in Cox Regression model (Table 3). However, patients with FGB −455 G/A A + genotype of the younger age group (55–71 years old) had worse survival (non-smoker HR = 2.37, 95% CI: 1.24-4.41; smoker HR = 2.33, 95% CI: 1.20-4.52) despite of smoking status suggesting an association between the carrier status of the FGB −455 G/A A-allele and elevated risk for death (Table 3). In subgroup analysis we detected a significant association between the A + genotype of the FGB −455 G/A polymorphism and worse post stroke survival in 55–71 year old women in Kaplan-Meier (p = 0.013) log rank analysis and multivariate Cox Regression forced model (non-smoker, crude HR = 5.21, 95% CI: 1.38-19.7, p = 0.015; smoker, crude HR = 7.03, 95% CI: 1.81-27.3, p = 0.005) (Figure 1 and Table 3). When history of arrhythmias, AF, CS, CRP, diabetes, hypercholesterolemia, hypertension, hypertriglyceridemia, and history of MI were forced into Cox regression model, the association between FGB −455 A-allele and worse survival persisted, however wider CIs were observed reflecting smaller sample size (non-smoker HR = 41.0, 95% CI: 3.23-520, p = 0.004; smoker HR = 37.2, 95% CI: 2.50-548, p = 0.009). Similar results was obtained in Cox regression forward stepwise LR (likelihood ratio) model (non-smoker HR = 26.9, 95% CI: 2.58-279, p = 0.006; smoker HR = 36.1, 95% CI: 3.42-381, p = 0.003) (Figure 1). In Cox regression analysis with stratification for each modifier showed that elevated CRP strengthened the effect of smoking and FGB −455 G/A polymorphism on worse survival in women aged 55–71 years (GG smokers, adjusted HR: 9.83, 95% CI 1.20-80.1; GA/AA non-smokers, adjusted HR: 15.4, 95% CI 1.91-124; GA/AA smokers, adjusted HR: 24.9, 95% CI 3.00-206), however the number of patients decreased considerably in this model (Figure 2). This association was not seen in 72–85 years old women or in men. In addition, we found no genotype-by-smoking interaction with specified cause of death in any age group of men or women.