Ready or not, routine use of genetic information is imminent in everyday cardiology practice. This impending reality stems partly from the declining cost of whole-genome sequencing that is now approaching the $1,000 barrier. Clinicians undoubtedly could use such insight in managing atrial fibrillation (AF), an extremely common and costly condition. However, analyzing and applying the vast amount of data housed in the human genome will clearly present huge challenges. With this backdrop, Parvez et al. (1) report in this issue of the Journal that response to certain antiarrhythmic drugs (AADs) may be genotype dependent.

Using a single institution registry, the authors examined success of AAD therapy in 676 Caucasian patients with AF. Patients were included in the study if they had a documented history of AF with concurrent use of at least 1 conventional AAD (Vaughan-Williams class I or class III) agent. About two-thirds had typical, but nonvalvular, AF—that is, AF in conjunction with hypertension, coronary artery disease, congestive heart failure, or diabetes—with the remainder having lone AF. Response to therapy was analyzed with a questionnaire (2) that yielded a symptomatic AF burden score by quantifying AF symptom frequency, duration, and severity. A patient was deemed a responder if she or he remained on the same AAD therapy for a minimum of 6 months and had a >75% reduction in their AF burden score. The investigators examined the association between AAD responder status and the presence of 4 single nucleotide polymorphisms (SNPs) that had previously been linked with AF in genome-wide association (GWA) studies.

Motivation for this study built from the recent recognition that family history is a risk factor for nonvalvular AF on top of traditional risk factors, such as age, hypertension, congestive heart failure, compromised respiratory function, and hyperthyroidism. Like many complex polygenic disorders, AF is inherited only rarely in Mendelian fashion, yet having at least 1 parent with AF doubles the risk of AF (3). For lone AF, family history confers an even greater risk (4). Several GWA studies (5) have recently identified specific genetic variants associated with AF, including 2 common SNPs on chromosome 4q25 (rs2200733 and rs10033464) near the transcription factor PITX2 (6). Subsequent studies have found a more definitive association for rs2200733 than for rs10033464 (7- 9). Two additional SNPs on chromosomes 1q21 and 16q22 with modest effects have also been identified (10- 11).

With the rationale that the multiple genetic variants associated with AF might indicate variable mechanisms contributing to AF susceptibility, or variable subtypes of AF, Parvez et al. (1) hypothesized that response to AAD therapy might also be genotype dependent. Testing first in a discovery cohort of 478 patients, the authors identified an association between a response to AAD therapy and the rs10033464 SNP, but not with the other SNPs tested. The association with rs10033464 was predominantly, if not exclusively, seen in patients with typical AF as opposed to the patients with lone AF. Using multivariate regression, the tendency to be a nonresponder to AAD therapy for patients carrying a minor allele at rs10033464 persisted after controlling for clinical variables such as age, sex, hypertension, coronary artery disease, heart failure, and diabetes. Interestingly, the investigators further found that the response rate to class I AADs was higher in patients with 1 or more minor alleles at rs10033464. On the other hand, the response rate to class III AADs was higher in patients with only wild-type alleles. In a validation cohort of 178 Caucasian patients with either lone or typical AF from Vanderbilt, the association of rs10033464 with response to ADDs was also significant; subgroup analysis of typical versus lone AF was not performed because of sample size.

Although this innovative study associating the SNP rs10033464 with AAD response breaks new ground, it should be considered hypothesis generating until confirmed by a randomized, double-blinded trial. We share the authors' optimism that the genomic revolution will yield opportunities to inform and tailor treatment, but this study raises several questions. First, the AAD response rate of 72% to 83% is unusually high, even if amiodarone contributed heavily (data not provided), and reflects relatively nonrigorous criteria for AAD success. Although freedom from any AF may be too stringent and miss useful clinical benefit of AAD (or ablation), success rates for AAD in a recent randomized trial was only 16% (12). Even with a more lenient endpoint in the AFFIRM (Atrial Fibrillation Follow-up Investigation of Rhythm Management) antiarrhythmic substudy, the 1-year success was only 23% for class I AADs and only 38% for sotalol (13). Since AADs alone or in combination with rate control agents may render AF less symptomatic or even asymptomatic, the 75% reduction in AF symptoms used to determine success in this study may underestimate actual AF prevalence. The relative subjectivity of the endpoint reduces confidence in the differences observed for the rs10033464 SNP and for different AADs. Concerning the secondary endpoint, AF recurrence, the intensity of monitoring was less than that advocated by the Heart Rhythm Society/European Heart Rhythm Association/European Cardiac Arrhythmia Society Expert Consensus Statement on Catheter and Surgical Ablation of AF, namely, a 24-h Holter monitor every 3 to 6 months for 1 to 2 years (14).

A second weakness of the study is that the AAD selection was not random, and the decision to continue it relatively subjective, depending upon the patient's and physician's preferences and the AF burden score. The patient-completed AF symptom questionnaire may be dependent upon patient and physician bias, and may not identify actual AF burden.

Third, the specific associations in this study are somewhat at odds with findings from the large studies that first identified these AF-related SNPs. The SNP highlighted in this study, rs10033464, has been the polymorphism less strongly associated with AF in GWA studies than its neighbor on 4q25 (rs2200733). Since both SNPs are thought to involve PITX2, the biological plausibility of the association is perhaps weakened. It is curious also that rs10033464 was more clearly associated with AAD response in typical AF patients in this study, whereas SNPs at 4q25 had previously been more strongly associated with lone AF (6).

Fourth, moving from statistical associations between rs10033464 SNP status and AAD response, to using the SNP information to guide therapy is problematic. For example, the positive predictive value of minor allele carrier status for predicting nonresponse to AAD is only 23%; that is, 25 of 109 total MAC patients in the discovery cohort were nonresponders per Table 2 (1). Analogously, the differential response to AADs is interesting and hypothesis generating, but difficult to act on with the degree of differences found. Lastly, as with any initial report, one needs to consider the possibility that the associations arose by chance; such positive associations may be more likely to be reported than negative ones (15).

In summary, Parvez et al. (1) are to be commended for extending the epidemiological associations between several SNPs and AF to an exploration of potential therapeutic ramifications. Lacking patient-specific predictors for efficacy, current AF drug selection considers only comorbidities in an effort to merely minimize proarrhythmia (16). Pharmacogenetic insights could theoretically help predict efficacy and toxicity. Although we are closer to the beginning than to everyday use in the clinic, Parvez et al. (1) have started us on an exciting journey.