GEN Exclusives

More »

GEN Exclusives

More »
February 21, 2018

Polygenic Risk Scores Show Utility for Stratifying Disease Risk

Polygenic Risk Scores Offer a Glimpse of True Clinical Utility

Polygenic Risk Scores Show Utility for Stratifying Disease Risk

Source: chombosan/Getty Images

  • Investigators from the University of Michigan School of Public Health found that polygenic risk scores (PRS) from a phenome-wide association study (PheWAS) is significantly associated with several cancer traits and non-cancer diagnoses, possibly demonstrating the potential usefulness of this system for risk stratification among certain population subsets.

    “I think this dataset has fantastic opportunities that will help us explore the temporal associations between PRS [polygenic risk score] data and disease as well as the sequences of events that occur in cancer patients,” said one of the study’s investigators, Lars G. Fritsche, Ph.D., who is assistant research scientist in the department of biostatistics at the University of Michigan School of Public Health. Overall, Dr. Fritsche believes clinicians may be able to draw inspiration from these findings for “identifying risk factors previously ignored with the data we currently have in hand.”

    Previous PheWAS investigations have allowed scientists to identify common variations in several different disease states. The polygenic risk score (PRS), defined as genome-wide genotype data culminated into one variable that measures genetic liability to a disease trait, may allow clinicians to have the opportunity to potentially predict disease risk. This information could then guide disease prevention efforts.

  • Study Findings and Rationale

    Study investigators enrolled a total of 28,260 genotyped patients who participated in the longitudinal biorepository Michigan Genomics Initiative. All participants had ICD9-based electronic health record (EHR) data available. Approximately 47.7% had at least one neoplasm prior to or upon study enrollment. Researchers calculated PRS for 12 cancer traits, including breast, prostate, skin, basal cell, bladder, non-Hodgkin's lymphoma, colon, squamous cell, kidney, lung, thyroid, and brain and nervous system carcinomas.

    Phenome-wide associations were found between certain cancer and non-cancer diagnoses. For instance, reduced risk for hypothyroidism correlated to patients who also had a high thyroid cancer PRS, and high risk for actinic keratosis was associated with a high squamous cell carcinoma PRS.

    Investigators of this study relied primarily on data recorded in electronic health records (EHRs), which are typically reserved for clinical practice. “The main point of this study is a push to integrate EHR,” said study investigator Bhramar Mukherjee, Ph.D., who is also from the department of biostatistics at the University of Michigan School of Public Health. “That’s a challenge because of confidentiality issues. EHRs are primarily designed for billing and patient care—not for research.”

    In this study, investigators took findings from an existing study and sought whether those findings were reproducible in an academic setting population. “When we looked, approximately 50% had at least one cancer diagnosis,” said Dr. Mukherjee. “Cancer patients undergo surgery more often, and they actually tend to come to an academic medical center. We had about 14,000 that had a neoplasm diagnosis. We wanted to see what comes out of it.”

  • The Value of EHR Data in Genetics Research

    The utility of EHR data in managing a patient’s treatment course is generally accepted by most practicing clinicians and researchers alike. In some patients, EHR data can aid in determining associations between a patient’s disease and the effects of any administered drugs. According to the investigators, measuring cancer patients’ genomic variation and integrating these data into EHRs may offer greater risk stratification than relying on clinical history alone. Additionally, genomic variation data may improve the rate at which future investigators and/or clinicians can discover associations between specific cancer diagnoses and genome changes.

    “One of the main incentives to EHR access is that you have the entire sequence/phenome-wide description of a patient’s diagnosis,” according to Dr. Mukherjee. “For prostate cancer, we found that there was an association between the prostate cancer PRS and erectile dysfunction as well as urinary incontinence.” What is generating these associations? The answer may come from the EHR. “By looking at patients with EHR data, we saw that EHR showed the sequence [of events], including what happens before and after [diagnosis]. These temporal associations are important.”

  • Application of PRS in the Real-World Setting

    To apply PRS in the clinical setting, the value lies in determining where a patient’s PRS falls on the population distribution. For instance, interventions used to decrease disease risk, such as exercise, dietary restriction, and stress reduction, may be offered to patients in the top PRS values. Furthermore, PRS may help establish screening algorithms in certain forms of cancer, among other illnesses, in order to stratify disease risk. Ultimately, there’s hope that this stratification may result in more targeted, optimized care while helping to improve survival and quality of life.

    In addition, the findings from this study suggest PRS may assist in cancer risk stratification among those who are seeking treatment in the academic medical center setting. The real question is, are these findings ready to be implemented into prime-time clinical practice? “PRS can be predictive of disease association, but how do you communicate it?” commented Dr. Mukherjee. “At the individual level, the risks could be very modest.” Some clinicians may be overly optimistic with PRS findings, resulting in “very aggressive prophylactic methods that are not necessary and that may even be harmful.”

    Jonathan L. Haines, Ph.D., director of the Institute for Computational Biology and chair of the department of population and quantitative health sciences at Case Western Reserve University School of Medicine, has been actively involved in researching applications of genetic risk scores in disease. His previous research has been focused on PRS in age-related macular degeneration and multiple sclerosis. Based on the new PRS findings by Drs. Fritsche and Mukherjee et al., Dr. Haines seems hopeful for its applicability.

    “The novelty here is really in showing that the known polygenic risk scores can be applied to a new dataset collected in a different way and still be very useful,” said Dr. Haines. “In other words, the previously published results do generalize into real-world healthcare data.”

    “Obviously, anything we can do to better understand cancer is critically important, as it is one of the leading causes of death,” added Dr. Haines. “Cancer genomics has been extremely successful in helping us understand how different each kind of cancer is. More importantly, it has led to pre-symptomatic diagnosis that allows preclinical treatments (e.g., BRCA1 mutation testing and breast cancer) and much more effective post-diagnostic treatments of cancer.”

    Calculating a PRS is relatively simple, and findings from these scores may facilitate risk prediction many years prior to the usual age of disease onset. Despite these advantages, many obstacles currently exist that prevent the use of PRS in precision medicine. “This is just the tip of the iceberg; it shows the potential and the challenges with using phenome-wide associations,” said Dr. Mukherjee. “But finding an association isn’t enough. We have very big data with biosampling, so we really need to think carefully about how we’re managing the data, defining the phenotype, and looking at the temporal sequence.”

    Currently, there is a substantial need for clinician and public education on polygenic initiatives to increase genetic understanding as it relates to medicine. Since most polygenic medicine research is focused primarily on individuals of European descent, including this new study, more research focused on patients of non-European ancestry is needed to determine wide-scale applicability.

    “There is still much to learn and many cancers have not yet given up their genetic secrets,” said Dr. Haines. “In many cancers, our genetic knowledge is still not good enough to be used for any sort of prediction.”

Related content