Polygenic risk and echocardiographic markers of left atrial dysfunction improve prediction of incident atrial fibrillation: insights from the akershus cardiac examination (ACE) 1950 Study
M Solberg, A Shrestha, A Sekkelsten, P Ronningen, S Enger, E Orstad, M Lyngbakken, H Rosjo, K Steine, M Mattingsdal, A Tveit, T Berge, I ChristophersenAbstract
Background
Left atrial (LA) dysfunction is prognostic for incident atrial fibrillation (AF) and can be assessed by echocardiographic strain of the reservoir (LASr) and contractile (LASct) phases. Genome-wide association studies (GWAS) have revealed >350 genetic loci associated with AF. Aggregating genetic risk variants in polygenic risk scores for AF (AF PRS) and adding these to clinical risk factors improves prediction of AF over clinical risk factors alone. Combining AF PRS and LA strain with clinical risk scores for incident AF may further improve the prediction of incident AF.
Purpose
To assess whether the addition of AF PRS and LA strain improves the prediction of incident AF beyond clinical risk factors.
Methods
The Akershus Cardiac Examination (ACE) 1950 Study is a prospective cohort of women and men born in 1950, residing in Akershus County, Norway. A baseline examination was conducted between 2012-2015 and included echocardiography with LA strain analysis and blood samples. The Illumina HumanCoreExome array was used for genotyping. Individual SNP effect sizes were retrieved from one of the latest GWAS meta-analyses for AF [1] and used to calculate AF PRS scores using the --score function in Plink. Incident AF diagnoses were obtained from national health registries until Dec 31st, 2022 and validated by ECG. The Cohorts for Heart and Aging Research in Genomic Epidemiology for AF (CHARGE-AF) risk score and sex were used as a baseline risk model for incident AF. The predictive value of AF PRS and LA strain measures for incident AF were assessed by Cox proportional hazards and C-statistics.
Results
The baseline examination included 3706 participants, and AF PRS was calculated in 3332 (51.2% females). Due to prevalent AF, 159 participants were excluded from follow-up. The remaining 3173 participants were followed for a median of 8.3 (7.9-8.9) years, during which incident AF was identified and validated in 178 (5.3%). CHARGE-AF score, sex, AF PRS, LASr and LASct were all associated with incident AF in univariable analyses. AF PRS, LASr and LASct retained significant predictive value for incident AF after adjustment for baseline AF risk by CHARGE-AF score and sex. (Table 1) Adding AF PRS to the baseline AF risk model significantly improved C-statistics (Figure 1), but the predictive value was not improved further by addition of LASr (p=0.12) or LASct (p=0.15). The model including CHARGE-AF, sex, AF PRS and LASct showed the strongest predictive value for incident AF.
Conclusion
In a general population born in 1950, CHARGE-AF score, sex and AF PRS combined with LA strain predicted incident AF better than CHARGE-AF score and sex.