Predicting atrial fibrillation (AF) risk in otherwise healthy women with no known cardiovascular disease may require only easily obtainable information that is readily available to physicians and other healthcare practitioners, according to a study published online Feb. 26 in the European Heart Journal. Adding certain biomarkers to the model increased its predictive value slightly, but not enough to result in significant reclassification of AF risk.
AF is now the most common arrhythmia, and it increases the risk of stroke, heart failure and death. Its prevalence is increasing in populations with no known heart disease, and therefore effectively directing prevention efforts is challenging. Recently several predictive models have been developed, but all require baseline electrocardiograms (EKGs), which may not be available in a population with no known heart disease. Some of the models use genetic information to predict AF risk.
Brendan Everett, MD, of Brigham and Women’s Hospital in Boston, and colleagues set out to design a model to predict the likelihood of AF among 20,822 participants in the Women’s Genome Health Study, a subset of the Women’s Health Study (WHS) of healthy female medical professionals. The participants were all of European ancestry, their genetic information was available, and none had cardiovascular disease, heart failure or AF at baseline.
The participants were followed for a mean of 14.5 years. The primary endpoint was a reported, verified incident of AF.
The 32 covariables the researchers considered for their predictive model included demographic information, basic clinical measurements such as blood pressure and cholesterol levels, lifestyle factors such as smoking and alcohol use history, and available genetic biomarkers that have been associated with AF. After excluding participants for whom information on the covariables was incomplete, the study population was divided into a derivation data set group (13,061 women) and an independent validation data set group (6,879 women). Baseline characteristics of the two groups were substantially similar.
The researchers used Cox proportional hazard models and Bayes Information Criteria minimization to select the covariables for their model. The final model included the following six factors: age (adjusted hazard ratio 239.79), weight (1.17), height (1.36), systolic blood pressure (1.17), two or more alcoholic drinks per day (1.63) and current or former smoker (1.29).
After testing this predictive model in the validation cohort, Everett et al found that it was a significantly better predictor of AF than age alone, was well-calibrated and “substantially improved classification into 10-year risk categories of less than 1 percent, between 1 and less than 5 percent, and 5 percent or greater.” Predicting risk with both the age-only model and the WHS AF model, the latter was able to more accurately reclassify the AF risk of 22.5 percent of the study population.
In a secondary part of the study, the participants’ genetic risk score (GRS) was calculated according to recently published AF risk allelles (Nat Genet 2012;44:670-675). Incorporating both weighted and unweighted GRS improved the ability of the WHS AF model to discriminate, but not to the extent that it had a significant impact on the ability of the WHS AF model to classify participants into 10-year risk categories, the researchers reported.
Everett et al pointed out that this simple model has good predictive value for AF, uses information that is readily available and, importantly, several of the variables are lifestyle factors that can be changed. Thus, clinicians can use this tool to counsel currently healthy women about their future AF risk and ways to reduce it. In addition, the tool has promise in targeting populations for AF screening or risk-reduction interventions, the authors concluded.
The authors noted that although testing the inclusion of biomarkers was a strength of the study, B-type natriuretic peptide levels were not available, and this biomarker has been strongly associated with AF incidence. They recommended further study on the predictive value of B-type natriuretic peptide levels on AF risk.
As limitations, the authors cited the fact that all study participants were women of European ancestry and therefore the results may not be generalizable to other female populations or to men. The authors also stated that the absence of baseline EKGs limited their ability to determine the predictive value of EKG data in the AF WHS model or to compare the effectiveness of the AF WHS model to other recently developed models that use EKG data.