New genomic risk score improves prediction of coronary artery disease

Genomic Risk Prediction of Coronary Artery Disease in 480,000 Adults Implications for Primary Prevention

Literature - Inouye M, Abraham G, Nelson CP et al. - J Am Coll Cardiol 2018;72:1883–93

Introduction and methods

Family history is a known risk factor for coronary artery disease (CAD), and the heritability of CAD has been estimated to be 40% to 60%. Existing genomic risk scores (GRS) have limited use due to a number of reasons, including insufficient accuracy for CAD [1-3]. In this study, a novel GRS for CAD was developed, and its potential as a screening tool for primary prevention was evaluated.

For this purpose, a meta-analytic strategy was used that captures the totality of information from the largest previous genome-wide association studies (GWAS). Subsequently, the external performance of this meta score (meta-GRS) was assessed in stratifying CAD risk in >480,000 individuals (aged 40-69 years) from the UK Biobank (UKB) [4].

The meta-GRS was constructed based on 3 GRSs:

a previously published score (GRS46K) of 46,000 genetic variants derived from a genetic association study using Metabochip, a genotyping array with a focus on cardio-metabolic genetic loci,
a score of 202 genetic variants significantly associated with CAD at false discovery rate <0.05 (FDR202) in a recent GWAS from CARDIoGRAMplusC4D,
a genome-wide polygenic score (1000Genomes) based on the same GWAS.

Main results

The external UKB validation set included 22,242 CAD cases before age 75 years and 460,387 non-cases in total, with 9,729 prevalent cases of CAD at the time of recruitment, and 12,513 incident cases of CAD during a mean follow-up of 6.2 years, at the censoring age of 75 years in 2017.
The meta-GRS consisted of 1,745,180 genetic variants, together explaining 26.8% of CAD heritability.
The area under the receiver-operating curve in the external UKB validation set was 0.79 (+2.8% over the reference logistic model consisting of gender, age at assessment, genotyping array, and 10 PCs).
The meta-GRS offered greater positive predictive value at any given sensitivity and, thus, greater area under the precision recall/sensitivity curve compared with the reference model (0.161 vs. 0.123).
The meta-GRS stratified individuals into significantly different life course trajectories of CAD risk, with those in the top 20% of meta-GRS distribution having an HR of 4.17 (95%CI:3.97-4.38), compared to those in the bottom 20%.
Individuals in the top 20% of meta-GRS distribution on lipid-lowering or antihypertensive therapy had an HR of 2.83 (95%CI: 2.61-3.07), compared to those in the bottom 20%.
The meta-GRS was significantly but weakly associated with BMI, diabetes, hypertension, current smoking, family history of heart disease, and self-reported high cholesterol.
Combining the meta-GRS with all 6 conventional risk factors led to a model with a C-index of 0.696 (95%CI: 0.688-0.703), corresponding to an increase of 2.6% over the model consisting of the 6 conventional risk factors.
For men in the top 20% of meta-GRS with >2 conventional factors, 10% cumulative risk of CAD was reached by 48 years of age.

Conclusion

The meta-GRS developed and evaluated in the present study achieved greater risk discrimination than previously available genetic risk scores that were based on selected genetic variants. Meta-GRS provides the opportunity to stratify individuals for different trajectories of CAD risk in general populations and highlights the potential for genomic screening in early life to complement conventional risk prediction.

Editorial comment

In his editorial article, Natarajan [5] discusses the need for improvement of CAD risk prediction, and states that current polygenic risk scores are lacking the full spectrum of genetic variation that influences CAD risk, since only whole genome sequencing can identify genomic variation. He also points out the need for relevant prospective randomized controlled studies, which would trigger guidelines changes in this direction. He concludes that health systems currently insufficiently manage to identify those likely to sustain premature CHD. ‘Inouye et al. show that incorporation of CHD polygenic risk with clinical risk factors can improve risk prediction and may help identify individuals who are candidates for earlier preventive therapies. Additionally, this single genetic test (currently <$100) only needs to be performed once, and this framework can be applied to calculate polygenic risk for virtually any trait.’

References

Show references

Find this article online at JACC