Transcription
November 2020ReCentMedical NewsPolygenic risk scores are being vigorouslyresearched across different fields of medicine.Clinical utility of polygenic risk scoresIntroductionThe practice of preventive medicine requires estimatingthe risks of developing chronic diseases so as to enablerisk-mitigating measures such as diet, lifestyle and medicalinterventions to be implemented.Similarly, in insurance underwriting we assess applicants’risks of developing life-shortening common chronicdiseases in order to correctly assort each person to theappropriate risk classification for pricing purposes.Both clinical and insurance underwriting approaches usedto derive such disease risk estimates are well known andincludes consideration of the following: Demographic characteristics such as age and gender Lifestyle criteria, for example BMI, smoking status,alcohol consumption and physical exercise habits Clinical risk factors such as blood pressure, bloodchemistries and biomarkersConspicuously absent from these lists is routine genetictesting. The emergence of polygenic risk scores forcommon adult-onset diseases aims to change this andultimately enhance available risk estimation tools.Genetic conceptsOur genes serve as blueprints to make molecules calledproteins, the building blocks for everything in our body. Insingle-gene (monogenic) disease, a mutation in just one ofour estimated 20,000 genes is responsible for disease,1See El-Fishawy P. (2013) Common Disease-Common Variant Hypothesis. In:Volkmar F.R. (eds) Encyclopedia of Autism Spectrum Disorders. Springer, NYalthough one should note that the presence of a mutationdoes not guarantee disease penetrance.Despite the relative success in identifying genesresponsible for many monogenic diseases, the majority ofdiseases do not trace back to a single genetic cause. Thecommon-disease, common-variant hypothesis posits that‘if a disease that is heritable is common in the population,then the genetic contributors to the disease must also becommon in the population.1Each cell nucleus contains around 3 billion nucleobasepairs with any two people differing by on average 3 millionpositions in their DNA. The majority of these differences –or ‘single nucleotide polymorphisms’ (SNPs) – do notappear to have any effect, but a small number ermine the significance of the SNP by comparinggenetic sequences of individuals with a trait or disease(phenotype) to those without it. SNPs present in those withthe phenotype and absent in the control are considered tobe ‘associated’ with the phenotype and the degree ofassociation is termed the ‘effect size’. These ‘genome-wideassociation studies’ (GWAS) have identified hundreds ofthousands of SNPs associated with various phenotypes andwhile GWAS have been undertaken for over 15 years,availability of data and enhanced processing capabilitieshas accelerated research in this space in more recent times.Further accelerating the usability of GWAS is the advent ofpolygenic risk scoring (PRS); a weighted sum of the
number of risk variants for a particular disease, distilledinto a very transparent score fitting a normal distribution.Essentially, you can think of PRS as a debit/credit modelfor genetic profiles, with the highest PRS correlating withthe highest relative risk of developing the particulardisease or trait over your lifetime. Arguably, PRSrepresents one of the most digestible means of reflectinggenetic risk, one which we can readily transpose over thetraditional insurance paradigm approach to riskclassification where the majority of lives are average risk(standard rates) and lives either side of the normal curverepresent the lowest and highest (preferred or substandard)risks respectively. While the ability to predict lifetimelikelihood of disease increases as more phenotypeassociated SNPs are identified and the PRS moves towardsand into the 90th percentile, development of the diseaseoften still relies upon other stimuli. Indeed, common,complex diseases appear to occur as a result of manygenomic variants with small effect sizes, interacting withoften modifiable environmental influences such as diet,sleep, stress and smoking; i.e. the genetics are notdeterministic.Fig. 1: The clinical impact of a high polygenic risk scorefor coronary artery disease6Odds Ratio95% CIRemaining 80%2.552.43-2.67Top 10%Remaining 90%2.892.74-3.05Top 5%Remaining 95%3.343.12-3.58Polygenic risk scores and coronary artery diseaseTop 1%Remaining 99%4.834.25-5.46Familial hypercholesterolaemia (FH) – an autosomaldominant disease – is most commonly caused by amutation of the low-density lipoprotein receptor (LDLR)gene.2 While other monogenic mutations have been found,approximately 15% of FH cases appear to be caused bymonogenic mutations of undetermined prevalence ormultiple genes interacting additively to influence the trait(polygenic disease). Roughly one in 250 members of theglobal population develop FH3 and they are predicted tohave up to 3.9 times more cardiovascular events over theirlifetime than non-familial hypercholesterolemia patientswith an otherwise similar risk profile.4 While much of thisrisk can be attenuated by early and aggressivecardiovascular risk factor modification, the challenge isoften one of timely identification – bearing in mind thatexposure to elevated low-density lipoproteins begins inutero for those with FH.5Top 0.5%Remaining 99.5%5.174.34-6.122See Chial, H. (2008) Rare Genetic Disorders: Learning About Genetic DiseaseThrough Gene Mapping, SNPs, and Microarray Data. Nature Education3 See Henderson, Raymond et al. “The genetics and screening of familialhypercholesterolaemia.” Journal of biomedical science vol. 23 39. 16 Apr. 20164 See Guillermo Villa, et al., Prediction of cardiovascular risk in patients withfamilial hypercholesterolaemia, European Heart Journal - Quality of Care andClinical Outcomes, Volume 3, Issue 4, (2017)5 See Gidding S.S., et al. (2015) The agenda for familial hypercholesterolemia: ascientific statement from the American Heart Association.Hannover Re 2PRS - CADReference GroupTop 20%Coronary artery disease polygenic risk scores in the top5% have odds ratios for CAD at a similar level tomonogenic disease. PRS appears to identify a differentsubset of lives at risk for CAD than those identified throughmonogenic sequencing and it has very low correlation withtraditional cardiovascular risk factors. 7 Considering thataround 15% of first heart attacks are in the context of notraditional major cardiovascular risk factors8, could it bethat PRS can help identify a significant sub-set of theselives early enough that intervention could mitigate the riskof a heart attack? Given that roughly 2% of early heartattack patients are identified as having a monogenicmutation, versus 17% of patients having a PRS in the top5%, the scope to intervene appears significant. Hazardratios for CAD in the top 20% of PRS are around 4 times6See Khera, Amit V et al. “Genome-wide polygenic scores for common diseasesidentify individuals with risk equivalent to monogenic mutations.” Naturegenetics vol. 50,9 (2018)7 See Biobank UK; Genome-wide polygenic scores to stratify risk for commondiseases. (2018)8 See Canto, John G et al. “Number of coronary heart disease risk factors andmortality in patients with first myocardial infarction.” JAMA vol. 306,19 (2011)
greater than for the bottom 20% and this pattern isobserved across numerous diseases, showing significantpotential for screening and treatment protocol change.9However, other results are mixed. The authors of a 2020study 10 published in JAMA journal found that the samepolygenic risk score used in the study by Khera et al.11 didnot improve on the risk stratification of incident coronaryheart disease in middle-aged Caucasian populations whencompared to traditional clinical risk scores, suggesting thatthe clinical utility of PRS wears off as people age.Polygenic risk scores and psychiatryGiven the polygenic architectures of psychiatric disorders,polygenic risk scoring seems primed to fundamentallyinform and change our understanding and indeedclassification of psychiatric disorders. 12 The advent ofpolygenic risk scores has generated significant interest inthe field of psychiatry, largely due to the lack of reliablebiomarkers in the field. Apart from family history, there arecurrently no clinical or laboratory predictors of theprobabilistic risk of psychiatric disorders in healthypopulations.However, recent psychiatric research has begun to explorethe use of PRS and this has highlighted two areas ofdevelopment:1. Estimating the latent risk of various psychiatricdisorders in healthy individuals – there is data to showthat in a general population of college students, a highPRS score for schizophrenia and neuroticism canidentify individuals at an increased risk of developinganxiety and depression.132. Predicting mental health outcomes in those with anexisting psychiatric diagnosis: both bipolar polygenicrisk scores and schizophrenia polygenic risk scoreshave been shown to predict likely outcomes insubstance addiction disorders.149See Inouye, Michael et al. Genomic risk prediction of coronary artery disease innearly 500,000 adults: implications for early screening and primary prevention.bioRxiv 25071210 See Mosley JD et al. Predictive Accuracy of a Polygenic Risk Score ComparedWith a Clinical Risk Score for Incident Coronary Heart Disease. JAMA (2020)11 See Khera AV et al. Whole-Genome Sequencing to Characterize Monogenic andPolygenic Contributions in Patients Hospitalized with Early-Onset MyocardialInfarction. Vol 139 (2019)12 See Anderson JS et al. Polygenic risk scoring and prediction of mental healthoutcomes. Current Opinion in Psychology (2019)13 See Docherty AR et al. Polygenic prediction of the phenome across ancestry, inemerging adulthood. Psychological Medicine 8/2018Hannover Re 3Polygenic risk scores and breast cancerDifferent studies 15 , 16 have shown that women withpolygenic risk scores for breast cancer in the highest 1%have a lifetime risk of developing breast cancer almostequivalent to the lifetime risk seen in women with high-riskmonogenic mutations such as BRCA1 and BRCA2. Thesepolygenic high-risk women should therefore be offered thesame intensive risk-reducing strategies as if they werecarrying a high-risk BRCA mutation.Low polygenic risk scores for breast cancer may also beuseful to determine the subset of asymptomatic middleaged women for whom mammographic screening – and itsassociated risks of false positive diagnoses andunnecessary treatments – may not be required.17Polygenic risk scores as predictors of all-causemortalityA few studies have emerged that have looked for geneticvariants associated with lifespan. 18 , 19 Wright et al.reported on eleven loci (the positions of a gene or mutationon a chromosome) significantly associated with paternallifespan and four loci significantly associated with maternallifespan. 20 Not surprisingly, a number of these lifespanassociated loci are also significantly associated with lifeshortening diseases. In their estimation, only 10% ofvariation in observed human lifespan is due to geneticvariants, which does not bode well for the development ofrobust polygenic risk scores for all-cause mortality.Is there a polygenic risk score to predicta person’s lifespan?Rather than pursue a study of polygenic risk scores for allcause mortality using putative lifespan genetic variants,the authors of an intriguing study published in September14See Reginsson GW et al. Polygenic Risk Scores for Schizophrenia and BipolarDisorder associate with Addiction. Addiction Biology Vol 2315 See footnote 216 See Mavaddat N et al. Polygenic Risk Scores for Prediction for Breast Cancerand Breast Cancer Subtypes. The American Journal of Human Genetics (2019)17 See footnote 218 See Wright KM et al. A Prospective Analysis of Genetic Variants Associatedwith Human Lifespan. G3 : Genes, Genomes, Genetics (2019)19 See Timmers PRHJ et al. Genomics of 1 million parent lifespans implicatesnovel pathways and common diseases and distinguishes survival chances. eLife2019;20 See footnote 18
2020 in The American Journal of Human Geneticsdeveloped a composite all-cause mortality polygenic riskscore incorporating polygenic risk scores for 13 commondiseases and 12 established risk factors.21 These diseasesand risk factors are known to have some geneticcomponent and have been shown to be significantlyassociated with mortality. They conducted their analyseson a large dataset from the UK Biobank. Mortality data wasobtained from death and cancer registries linked to the UKBiobank.AuthorsGareth MatthewsChief UnderwriterTel. 44 20 [email protected], their sex-specific polygenic risk score for all-causemortality showed modest predictive ability. At theextremes of score distribution, the PRS may be more usefulas it is able to identify those with both significantly reducedand elevated risks of all-cause mortality. Differences in lifeexpectancy between the top and bottom 5% of thecomposite PRS were estimated to be 4.79 years and 6.75years for women and men, respectively. Comparatively,Timmers et al. found polygenic risk score differences of 5years of life between the top and bottom deciles for bothmales and females. 22 Clearly, such results would be ofinterest to life insurance actuaries and underwritersparticularly in markets offering preferred life insuranceclasses.Nico Van ZylVP, Chief Medical DirectorTel. 1 720 279 [email protected] us on LinkedIn to keep up to date with thelatest Life & Health news.ConclusionPolygenic risk scores are still in their infancy, yet they arebeing vigorously researched across different fields ofmedicine. Overall, these scores’ current predictive abilityfor incident phenotypic traits appears to be modest withincreased clinical utility for individuals scoring at thehigher and lower ends of the score distributions. Recentstudies suggest that PRS in the presence of monogenicmutation significantly modifies the penetrance of thedisease risk variants, circling us back to a continuum ofcommon low-risk to rare high-risk genetic variants actingcumulatively to drive overall risk in any individual. 23Polygenic risk scores for all-cause mortality have yieldedresults that would be of interest to actuaries andunderwriters although access to incorporate these newgenomic tools is likely to be significantly restricted due tothe legislative and regulatory restrictions that exist inseveral countries and regions.21See Meisner A et al. Combined Utility of 25 Disease and Risk Factor PolygenicRisk Scores for Stratifying Risk of All-Cause Mortality. The American Journal ofHuman Genetics (2020)Hannover Re 4The information provided in this document does in no waywhatsoever constitute legal, accounting, tax or other professionaladvice. While Hannover Rück SE has endeavoured to include inthis document information it believes to be reliable, complete andup-to-date, the company does not make any representation orwarranty, express or implied, as to the accuracy, completeness orupdated status of such information. Therefore, in no casewhatsoever will Hannover Rück SE and its affiliated companies ordirectors, officers or employees be liable to anyone for anydecision made or action taken in conjunction with the informationin this document or for any related damages.22See Timmers PRHJ et al. Genomics of 1 million parent lifespans implicatesnovel pathways and common diseases and distinguishes survival chances. eLife2019;8:e39856 pp. 1-4023 See Fahed, A.C., et al. Polygenic background modifies penetrance ofmonogenic variants for tier 1 genomic conditions. Nat Commun 11, 3635 (2020).
Hannover Rück SE. All rights reserved. Hannover Re is theregistered service mark of Hannover Rück SEwith Early-Onset Myocardial Infarction. Circulation 2019 Vol 139pp. 1593-1602BibliographyMavaddat N et al. Polygenic Risk Scores for Prediction for BreastCancer and Breast Cancer Subtypes. The American Journal ofHuman Genetics January 3, 2019, Vol 104 pp. 21-34Anderson JS et al. Polygenic risk scoring and prediction of mentalhealth outcomes. Current Opinion in Psychology 2019, 22 pp. 7781Biobank UK; Genome-wide polygenic scores to stratify risk small-1.pdf, Retrieved on 2 September 2020.Canto, John G et al. “Number of coronary heart disease risk factorsand mortality in patients with first myocardial infarction.” JAMAvol. 306,19 (2011): 2120-7. doi:10.1001/jama.2011.1654Chial, H. (2008) Rare Genetic Disorders: Learning About GeneticDisease Through Gene Mapping, SNPs, and Microarray Data.Nature Education 1(1):192Docherty AR et al. Polygenic prediction of the phenome acrossancestry, in emerging adulthood. Psychological Medicine 8/2018Vol 48 (11) pp. 1814-1823El-Fishawy P. (2013) Common Disease-Common VariantHypothesis. In: Volkmar F.R. (eds) Encyclopedia of //doi.org/10.1007/978-1-4419-1698-3 1998Meisner A et al. Combined Utility of 25 Disease and Risk FactorPolygenic Risk Scores for Stratifying Risk of All-Cause Mortality.The American Journal of Human Genetics September 3, 2020 Vol107, pp. 1–14Melzer D et al. The genetics of human ageing. Nature ReviewsGenetics 11/5/2019 Vol 21(2) pp 88-101Mosley JD et al. Predictive Accuracy of a Polygenic Risk ScoreCompared With a Clinical Risk Score for Incident Coronary HeartDisease. JAMA February 18, 2020 Volume 323, Number 7 pp. 627635Reginsson GW et al. Polygenic Risk Scores for Schizophrenia andBipolar Disorder associate with Addiction. Addiction Biology Vol23 (1) pp. 485-492Timmers PRHJ et al. Genomics of 1 million parent lifespansimplicates novel pathways and common diseases anddistinguishes survival chances. eLife 2019;8:e39856 pp. 1-40Wright KM et al. A Prospective Analysis of Genetic VariantsAssociated with Human Lifespan. G3 : Genes, Genomes, GeneticsSeptember 2019 Vol 9 pp. 2863-2878Fahed, A.C., Wang, M., Homburger, J.R. et al. Polygenicbackground modifies penetrance of monogenic variants for tier 1genomic conditions. Nat Commun 11, 3635 idding S.S., Champagne M.A., de Ferranti S.D., et al. (2015) Theagenda for familial hypercholesterolemia: a scientific statementfrom the American Heart Association. Circulation 132:2167–2192.Guillermo Villa, Bruce Wong, Lucie Kutikova, Kausik K. Ray, PedroMata, Eric Bruckert, Prediction of cardiovascular risk in patientswith familial hypercholesterolaemia, European Heart Journal Quality of Care and Clinical Outcomes, Volume 3, Issue 4, October2017, Pages 274–280,Henderson, Raymond et al. “The genetics and screening of familialhypercholesterolaemia.” Journal of biomedical science vol. 23 39.16 Apr. 2016, doi:10.1186/s12929-016-0256-1Inouye, Michael et al. Genomic risk prediction of coronary arterydisease in nearly 500,000 adults: implications for early tps://doi.org/10.1101/250712Khera, AV et al. “Genome-wide polygenic scores for commondiseases identify individuals with risk equivalent to monogenicmutations.” Nature genetics vol. 50,9 (2018): 1219-1224.doi:10.1038/s41588-018-0183-zKhera AV et al. Whole-Genome Sequencing to CharacterizeMonogenic and Polygenic Contributions in Patients HospitalizedReCent medical news editions relating to this topicGenetic tests: are they all equal?Personal genomic testing, the consumer and thelife insurance industryhr equariumFind out which solutions on hr equarium focuson genetics.www.hannover-re.com
Coronary artery disease polygenic risk scores in the top 5% have odds ratios for CAD at a similar level to monogenic disease. PRS appears to identify a different subset of lives at risk for CAD than those identified through
