Predicting the Future in Patients with Atrial Fibrillation: who Develops Heart Failure?

This editorial refers to 'Predictors and prognostic implications of incident heart failure following the first diagnosis of atrial fibrillation in patients with structurally normal hearts: the Belgrade Atrial Fibrillation Study', by T.S. Potpara et al., published in this issue on pages 415–424. Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia; the lifetime risk is 1 in 4 for persons over the age of 55 years in The Netherlands.1 AF is associated with a five-fold increase in the risk of stroke, a three-fold increase in the risk of heart failure, and a doubling in the risk of dementia and death.2 Risk prediction models are important to define the individual risk for development of AF.3 However, in patients who have developed AF, prediction of AF-related complications is also of great clinical importance. Risk prediction schemes are useful for risk communication, patient motivation, and clinical decision-making.4,5 The best examples are the stroke risk schemes developed and validated to predict embolic stroke, of which the CHA2DS2-VASc score currently is the most widely used.2,6 Based on the CHA2DS2-VASc score, the guidelines recommend the use of oral anticoagulation in patients with (a history of) AF, to prevent future stroke.2,6 Despite the success of stroke risk schemes, these schemes only predict one deleterious complication of AF. The EuroHeart Survey investigators proposed a more general risk scheme, which attempts to predict the progressive nature of AF, from paroxysmal to permanent AF, and progression of AF led to more adverse cardiovascular events and hospital admissions.7 The HATCH scoring system calculates 1 point for hypertension, age ≥ 75 years, and COPD, and 2 points for transient ischaemic attack or stroke, and heart failure, and allows instant classification of the risk of progression to persistent or permanent AF in patients with paroxysmal AF.7 However, this risk score has not been validated in independent cohorts, and has not found its way into daily clinical practice yet. Until very recently, no risk schemes were available to predict the development of heart failure in patients with AF. In September 2012, Suzuki et al. proposed the H2ARDD risk score (heart diseases = 2 points, anaemia = 1 point, renal dysfunction = 1 point, diabetes = 1 point, and diuretic use = 1 point; range 0–6 points) to predict heart failure in patients with AF. This score had an excellent discriminative c-statistic of 0.84 in their single hospital-based cohort consisting of 1942 Japanese AF patients.8 In the present issue of the journal, the results of an analysis in the Belgrade Atrial Fibrillation Study were presented.9 The authors included a cohort of 842 patients with new-onset AF with structurally normal hearts and investigated risk factors for the development of heart failure. During 5 years of follow-up, 83 AF patients developed heart failure, with a linearized rate of incident heart failure of 0.97% (95% confidence interval 0.78–1.19%) per 100 patient-years. Several covariates were associated with an increased risk of heart failure: a history of hypertension, diabetes mellitus, dilated left atrium (defined as left atrial diameter >40 mm), and low-normal LVEF (defined as LVEF 50–54%). In addition, the authors found that AF patients who developed heart failure were also at increased risk for cardiovascular hospitalizations and mortality. This last finding is in accordance with another recent paper published in the journal by Smit et al. which showed that when patients developed first AF and then heart failure they had an impaired prognosis, with increased cardiovascular hospitalizations and mortality rates, although prognosis was even worse when patients developed first heart failure and then AF.10 As with all observational studies, there are some limitations that need to be considered. First, the authors only included patients with structurally normal hearts, and a LVEF >50% on echocardiography. In addition to questions about the generalizability of results to other AF cohorts, this risk score can only be used after an echocardiography has been performed and may preclude the use of the Belgrade AF risk score outside a hospital setting. In contrast, another heart failure risk score, developed by the Framingham Heart Study investigators, that has also very recently been published in the journal, includes only widely available clinical covariates (age, body mass index, diabetes, significant murmur, electrocardiographic LV hypertrophy, and prevalent myocardial infarction). Secondly, residual and unmeasured confounding needs to be considered. Covariates, and biomarkers (e.g. BNP) that were not measured or collected, may have potentially influenced the outcome of the present study. Thirdly, difficulties in heart failure score derivation in AF should be considered. The variability in cardiovascular conditions underlying AF, in temporal patterns of AF (paroxysmal, persistent, and permanent AF), and in the treatment of AF (e.g. rate control drugs, rhythm control drugs, and ablation) may impact the development of heart failure in AF patients. Furthermore, heart failure diagnosis is based on clinical signs and symptoms, which in part overlap with AF symptoms, and may complicate the diagnosis of heart failure in AF patients, especially in those with a preserved EF. Fourthly, validation of this risk score in independent cohorts has not been performed yet, which is necessary before a risk score can enter daily clinical practice. Despite these limitations, the present study highlights several important points. Risk scores for heart failure may help to classify AF patients into clinically important categories: those at high risk for development of heart failure and also true low-risk patients. This is not only important for risk communication and motivation of the individual patient, but also helps to develop and apply specific therapies for the high-risk patients, and to avoid potential harmful therapies in low-risk patients. For example, the early identification of high-risk patients may help to develop novel therapies for specific patients. Presumably, we may have to start early and aggressively with 'upstream' therapy in these high-risk patients to reverse the modelling process at an early stage. Results of the upstream therapy in patients with early AF—The relevance of the Routine versus Aggressive upstream rhythm Control for prevention of Early atrial fibrillation in heart failure (RACE 3) study—are therefore eagerly awaited.12 Present heart failure risk scores8,9,11 can serve as a benchmark to study novel biomarkers and risk factors of heart failure in AF patients. The addition of novel covariates may improve risk prediction and reclassification for heart failure in AF patients. Accurate risk assessment is the backbone of prevention and may help to pave the way for the ongoing success story comparable with stroke prevention in AF. Conflict of interest: none declared.