Polygenic risk scores: a better cancer predictor for insurers?
Polygenic risk scores are proving to be a powerful tool in the era of genomic medicine. Their application has grown substantially – from predicting common diseases to enhancing risk modelling for various types of cancers. What does this mean for insurers?
Article information and share options
- In recent years, the application of polygenic risk scores has grown substantially – from predicting common disease to enhancing risk modelling for leading cancers.
- PRS enables greater risk stratification which can lead to better prevention, more personalised treatments and improved health outcomes.
- Insurers are increasingly turning towards more sophisticated approaches of scoring risk; polygenic risk scores may have importance to underwriting and pricing.
- Therefore, the insurance industry needs to remain engaged with scientific communities, regulatory bodies and the wider public in managing the role of genetic testing in insurance.
Quick recap – what's a polygenic risk score (PRS)?
A polygenic risk score (PRS) is a quantitative tool to predict an individual's genetic predisposition of developing a disease or a given trait. You can think of it as a snapshot of an individual's genetic "baseline" risk that can predict the probability of future health outcomes.
It is a single number calculated from an array of pooled risks based on single-nucleotide polymorphisms (SNPs), or genetic variants, identified through genome-wide association studies1. Importantly, it remains constant throughout life – even as we grow and develop. By comparing the sum of an individual's genetic risks across the entire genome to individuals of similar ancestry, we can more accurately stratify medical risks at the tail ends of population distribution. This allows us to identify individuals at relatively higher and lower risk of developing certain conditions. Having a more precise understanding of an individual's risk can lead to better, more personalised treatments or preventive measures and, as a result, hopefully improved health outcomes.
- More than 200 different scores available
- Tested once and remain constant throughout life
- Accuracy and reliability highly dependent on ancestry
- Cost: up to $1,000 USD per person's genetic information
Figure 1: Polygenic Risk Scores (PRS) enable more precise risk differentiation by stratifying tail end distribution
Nature vs. nurture?
While discussions around PRS tend to focus on predictive potential, equally as important are its non-deterministic characteristics, such as lifestyle interventions and environmental factors, on health outcomes.
In 2014, a Swedish study of identical twins who share the same DNA sequence were found to have different expressions of genes involved in glucose metabolism. In each pair of twins, only one sibling had developed type 2 diabetes despite sharing an identical genetic code2. This study illustrates two key points. Firstly, that variations in the environment directly impact gene expression through complex molecular pathways. Secondly, that health outcomes correlate with genetic predisposition but can be altered by lifestyle or environmental factors. For PRS, it is not a situation of nature versus nurture but rather, it is a complex and dynamic relationship between nature and nurture.
As a result, PRS shouldn't be seen as just a predictive score; it can also be used as a tool to motivate positive behaviour change. Recent clinical studies are proving this. For example, the Finnish GeneRISK study found that nearly 90% of participants who received information and a PRS on their risk of developing cardiovascular disease (CVD) changed their behaviour to mitigate their risk. Overall, the study also found that risk-reducing behaviour, such as losing weight, quitting smoking or visiting the doctor, increased by over 32% for individuals with a predicted 10% or higher risk of developing CVD and by over 18% for those at lower risks3.
As insurers increasingly adopt wellness programs and look to help policyholders improve their health, will they see PRS as another potential tool in their health management toolkit? PRS bring their own set of challenges for the insurance industry when it comes to underwriting and anti-selection. Many countries have passed regulations limiting insurers' access to genetic information in risk assessment. But they may offer tangible benefits for behaviour change, prevention and risk reduction. More time and research are needed to further investigate the long-term effects PRS will have on consumer behaviour and, importantly, what this means for insurers.
What's new in clinical medicine?
When we investigated PRS last year in our Underwriting Insight, researchers were primarily leveraging the scores to stratify risk for common health conditions like coronary artery disease, atrial fibrillation, and type 2 diabetes4. Since then, the list has grown. PRS are now being used to predict risk across a variety of factors including Alzheimer's disease, alcohol consumption, body mass index (BMI), stroke and several psychiatric conditions. But the greatest strides may have been made in the field of cancer. For example, there are now more than 20 polygenic risk scores that can assign individuals to different categories of breast cancer risk. PRS for lung cancer, prostate cancer and colorectal cancer have also seen increased growth. Could PRS transform the prediction, treatment and even prevention of cancer? We contemplate the possibilities through two case studies – one on breast cancer and the other on lung cancer.
Case study #1 - PRS enhance breast cancer prediction models
The Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) model calculates the lifetime risk of developing breast cancer. It considers family history, mammography, hormonal or reproductive factors, lifestyle factors and genetic variants. It also includes well established gene variants such as BRCA1 and BRCA2 conferring high breast cancer risk. However, these known genetic factors only explain about 45% of the inheritance patterns in breast cancer. The rest cannot be easily quantified but may be captured partially through PRS.
The most recently expanded model of BOADICEA V5 now includes a PRS comprised of 313 SNPs that account for polygenic variance contribution to breast cancer risk5. In this model, PRS results in greater risk stratification than either mammography or clinical risk factors (body mass index, use of hormonal therapy, etc.) alone. However, the most expansive risk stratification was achieved when all the factors were combined, enabling researchers to separate an average risk population with an 11.5% lifetime risk of breast cancer into those with <5% lifetime risk and those with >25% lifetime risk. This represents more than quadruple times the risk in high- vs. low-risk groups and has the potential to lead to more effective screening, surveillance and prevention programs.
Figure 2: More accurate risk assessment for breast cancer when clinical factors are combined with genetic information
What about women with a family history of breast cancer who are already at a higher risk? The addition of PRS in this above-average risk group resulted in recategorization into near-average risk (55.1% of women) and high risk (6.8% of women). In this instance, PRS picks up additional risk missing from family history alone and provides a more precise assessment of the likely outcomes. This shows that in the case of breast cancer, family history is a partial predictor at best – and is more accurate when combined with genetic information in the form of PRS.
Current cancer screening guidelines do recommend earlier and/or more frequent breast imaging for those at higher risk with a positive family history. Prediction tools such as BOADICEA are often used to justify increased screening for those individuals. Clinically, PRS can further stratify the group of women who meet screening guidelines, enabling each woman to receive a customized screening program for their individual risk profile. We explore how this might impact the insurance industry in the concluding sections below.
Case study #2 – A shift towards personalised lung cancer screening in China
The application of PRS for breast cancer is widely known and well-studied. Lung cancer on the other hand presents a new territory. Most lung cancer screening guidelines look at clinical risk factors of smoking history and age, and split the population into average versus high-risk categories. In 2019, a group of researchers developed a PRS based on 19 SNPs that stratified the risk of lung cancer into low, intermediate and high genetic risk groups in a Chinese population6. Combining this 19-SNP PRS with the smoking history, the researchers were able to stratify the study population into a greater number of risk brackets that reflected genetic and lifestyle risk combinations for lung cancer.
Figure 3: Using PRS in addition to smoking history may help better identify high risk patients
The researchers found that a light smoker at a higher genetic risk (HR 2.93) is actually at a greater risk of developing lung cancer than a heavy smoker with low genetic risk (HR 2.08).* This discovery was important because current lung cancer screening programs rely heavily on smoking history to determine who should be screened and consider a high risk to be 30 pack-years* or more. Incorporating PRS as part of the screening protocol – in addition to existing determinants – may help better identify high risk patients, especially those who are at increased risk even though they are a light smoker. This scenario illustrates the interplay between genes and lifestyle factors, and it underscores the importance of examining both in risk assessment.
*A pack-year is the multiplicative sum of the number of packs of cigarettes smoked per day by the number of years the person has smoked. For example, 1 pack- year is equal to smoking 1 pack per day for 1 year, or 2 packs per day for half a year.
What does this mean for L&H insurers?
- PRS enables greater precision in risk stratification for several leading cancers.
- PRS enhances predictive scores in addition to clinical risk factors. If genetic information can be used in the underwriting process, improved accuracy in risk stratification for select common conditions can be expected.
- While family history can only account for some inherited risks, it currently remains the best proxy for genetic risk available to insurers where access to genetic data is restricted.
- Underwriters need to comply with local laws and regulations and/or agreed industry standards.
- The successful use of PRS is likely to result in earlier detection of cancer in at-risk populations, which could lead to earlier interventions and higher cancer survival rates. Over time, this could contribute to overall improvement in mortality rates and influence future product pricing if this changes the aggregate level of expected mortality improvement.
- For critical illness (CI) products, the continued trend towards earlier detection of cancer should be reflected in a greater proportion of predicted claims for early-stage disease and higher age-specific incidence of cancer diagnosis. However, strong cancer definitions, including a wider differentiation of payouts between major (late-stage) and minor (early-stage) CI, will limit the impact from increased early-stage cancer diagnoses.
- The risk of anti-selection due to asymmetry of information – whether it be from PRS or other genetic tests or tools – represents a significant challenge to managing a sustainable portfolio. We discuss this in more detail in our previous publication on genetic testing.
- Increasing popularity of PRS as part of a shift toward personalised cancer screening will result in changes in who is screened, but not in how cancers are ultimately diagnosed. Therefore, PRS in its current domain of application is expected to have minimal impact on cancer definitions and claims assessment.
Cancer screening… then what?
So far, research have focused on the predictive power of PRS and how it can improve the screening portion of personalised medicine. There has not been enough exploration into using PRS to create personalised, targeted treatment therapies. More research is needed to understand the potential of PRS to predict treatment response and prognosis, which will also have an impact on insurance products.
PRS have shown their value in enabling greater differentiation of cancer risk pools. As healthcare standards move towards personalised medicine, we anticipate that PRS as a cancer risk predictor will become more prevalent. Therefore, the insurance industry needs to remain engaged with the medical and scientific communities, regulatory bodies, and the wider public in managing the role of genetic testing in risk protection. As more insurers turn towards a more sophisticated approach of scoring risk, PRS may prove to be a valuable component. Swiss Re continues to monitor the global developments in polygenic risk scoring and cancer prediction. To learn more, get in touch with your local Swiss Re contact.
- Babb de Villiers, C., M. Kroese, and S. Moorthie, Understanding polygenic models, their development and the potential application of polygenic scores in healthcare. Journal of Medical Genetics, 2020: p. jmedgenet-2019-106763.
- Nilsson, E., et al., Altered DNA methylation and differential expression of genes influencing metabolism and inflammation in adipose tissue from subjects with type 2 diabetes. Diabetes, 2014. 63(9): p. 2962-76.
- Science Daily. Individual access to genomic disease risk factors has a beneficial impact on lifestyles. 2018.
- Khera, A.V., et al., Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nature Genetics, 2018. 50(9): p. 1219-1224.
- Lee, A., et al., BOADICEA: a comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genet Med, 2019. 21(8): p. 1708-1718.
- Dai, J., et al., Identification of risk loci and a polygenic risk score for lung cancer: a large-scale prospective cohort study in Chinese populations. Lancet Respir Med, 2019. 7(10): p. 881-891