Risk Model Aids Prediction of Atrial Fibrillation in At-Risk Patient Population

Previous data show that the risk for new-onset atrial fibrillation is almost double with the presence of hypertension, while diabetes is linked to higher risk of suffering a cardiovascular disease.

Investigators, led by Rosa Abellana, Biostatistics, Department of Basic Clinical Practice, University of Barcelona, aimed to develop and validate a diagnostic predictive model to calculate risk of AF at 5 years in hypertensive, diabetic population.

A tool to detect AF may be crucial for patients, as stroke constitutes the first manifestation of undiagnosed AF in 25% of cases, investigators noted.

Study

For AF detection, the team created a prediction model to determine probability of AF occurrence in the patient population.

Then, it was validated using bootstrap techniques and externally with data from a comparable patient sample.

The derivation cohort included patients with hypertension and diabetes attended in 52 primary healthcare centers in Barcelona, with inclusion criteria of >50 years and without an AF diagnosis as of January 2013.

Exclusions include patients with chronic inflammatory diseases, malignant neoplasm, and dementia. A 5-year follow-up was performed from January 2013 – December 2017.

Further, the validation cohort was collected from 11 primary healthcare centers in Catalonia South, with the information collected from January 2013 – December 2017.

Information on medical history and demographic characteristics were collected to categorize prognostic variables, including gender, age, smoking status, alcohol risk, and laboratory information.

In addition, comorbidities included hypercholesterolemia, myocardial infarction, peripheral vascular disease, valvular heart disease, heart failure, thromboembolism, stroke, chronic renal disease and treatment related to antihypertensive drugs.

Investigators used multivariable cox regression to identify clinical risk factors associated with development of AF.

Results

The collected data was composed of 54,575 hypertensive diabetic patients, with a median follow-up of 60 months.

The team noted that AF developed in 3745 patients, at an incidence rate of 15.3 per 1000 person-year. Patient demographics included 51.7% women and a mean age of 72.1 years.

Investigators noted the most prevalent comorbidities at baseline were hypercholesterolemia (57.5%), chronic renal disease (18.8%), and stroke (11.3%). They found 11.9% of patients took ≥3 antihypertensives daily.

Univariate analysis showed the potential predictive variables for AF included gender, age, smoking status, BMI, SPB, DPB, heart rate, all cardiovascular commodities, chronic renal disease, and number of anti-hypertensive drugs.

Further, multivariate analysis identified gender, age, BMI, SBP, DBP, heart rate, myocardial infarction, peripheral artery disease, valvular heart disease, heart failure, thromboembolism, stroke, and chronic kidney disease.

In addition, investigators noted the AF predictive model demonstrated good discrimination ability with an apparent c-index at 5 years of 0.692 (95% CI, 0.684 – 0.700), as well as an optimism-corrected c-index of 0.69.

The validation cohort showed a lower discrimination, with a c-index of 0.670.

Investigators observed the number of antihypertensive drugs for blood pressure control had an associated with increased AF risk.

Obesity was related to an increased risk of AF, while cardiovascular disease also had an association with AF in the study. Higher mortality rates were also found in patients with heart failure.

Conclusion

Investigators concluded their risk model may help clinicians evaluate the risk of AF development, particularly in high-risk populations.

“The model accurately predicts future atrial fibrillation in a population with both diabetes and hypertension,” investigators wrote. “Early detection allows the prevention of possible complications arising from this disease.”

The study, “Predictive model for atrial fibrillation in hypertensive diabetic patients,” was published online in the European Journal of Clinical Investigation.