Polygenic Risk Score (PRS) Test: Disease Risk, DNA Markers, and Results

A polygenic risk score estimates inherited susceptibility to a common disease by combining the effects of many DNA markers. Unlike a single-gene test that looks for a rare, high-impact variant, a PRS may use hundreds, thousands, or millions of common variants, each contributing a very small amount to risk. The final score usually places a person relative to a reference population—for example, near the 50th percentile or in the highest 5% for coronary artery disease.

A PRS is not a diagnosis and does not predict whether disease will definitely occur. Its meaning depends on the population used to build and calibrate the score, the person’s genetic ancestry, the outcome definition, and how the score is combined with age, family history, lifestyle, imaging, and clinical measurements. Two commercial tests for the same condition can produce different rankings because they use different variants and models. Clinical use remains selective, and a result should change screening or prevention only when an evidence-based care pathway exists.

• A PRS measures relative inherited susceptibility, not current disease and not a guaranteed future outcome.
• A percentile shows rank within a reference group; it is not the same as an absolute lifetime risk percentage.
• Performance can drop when the tested person’s ancestry differs from the score’s development and validation populations.
• A high score may support earlier or more intensive prevention, but only within condition-specific clinical guidance.
• A low score does not cancel smoking, high blood pressure, symptoms, family history, or a pathogenic single-gene variant.
• The exact PRS name, version, reference population, and calibration method are necessary to interpret or repeat the result.

Table of Contents

• How a Polygenic Risk Score Is Built
• What PRS Results Look Like
• Relative Risk and Absolute Risk
• Where PRS May Be Used
• Ancestry, Calibration, and Equity
• Limitations and Common Errors
• Choosing and Using a PRS Test

How a Polygenic Risk Score Is Built

Most common diseases do not arise from one genetic change. Coronary artery disease, type 2 diabetes, common cancers, glaucoma, and many other conditions reflect a combination of many genetic influences plus environment, behavior, age, and chance. A polygenic risk score compresses part of that inherited contribution into one number.

Researchers usually begin with a genome-wide association study, or GWAS. A GWAS compares millions of markers across large groups of people with and without a trait. It identifies variants—often single-nucleotide polymorphisms, or SNPs—that occur more often with the disease. Each association receives an estimated effect size.

A simplified score can be written as:

PRS = sum of the number of risk alleles at each marker × that marker’s statistical weight

The real process is more complex. Nearby DNA markers are correlated through linkage disequilibrium, so methods must avoid counting the same signal repeatedly. Modern approaches may include millions of markers, use Bayesian models, and account for ancestry-specific correlation patterns. Researchers then test the score in an independent validation group.

A clinically credible PRS needs more than a statistically significant association. It should show:

• Analytical validity: the laboratory accurately measures the necessary variants.
• Discrimination: the score separates higher-risk from lower-risk groups to a useful degree.
• Calibration: predicted risks match observed disease rates in the target population.
• Incremental value: the score adds information beyond existing clinical factors.
• Clinical utility: using the score changes care in a way that improves outcomes or meaningfully reduces harm.

These layers are often confused. A score may be strongly associated with disease but add little to an established calculator. It may improve statistical prediction without changing who qualifies for treatment. It may identify high-risk people but lack evidence that acting on the result improves health.

The reference population is built into the result. A raw score has no intuitive meaning until it is standardized or ranked. Laboratories may report a z-score, percentile, odds ratio, hazard ratio, relative risk, or modeled absolute risk. Each format requires different interpretation.

PRS testing is different from a single-gene genetic test. A pathogenic variant in LDLR can cause familial hypercholesterolemia with a large effect on coronary risk. A coronary PRS combines many common variants with small effects. A person can have either, both, or neither. A low PRS does not neutralize a high-impact pathogenic variant.

It is also different from family history. Family history captures shared genes, environment, behavior, and healthcare patterns. PRS measures selected inherited markers. The two can provide independent information, so one should not automatically replace the other.

What PRS Results Look Like

There is no universal PRS report. The same person can receive different numbers from different scores because the models use different variants, weights, populations, and disease definitions.

Percentile rank

A percentile is the most familiar format. A result at the 92nd percentile means the score is higher than approximately 92% of people in the stated reference group. It does not mean a 92% chance of disease. It also does not mean the person’s risk is 92% higher unless the report separately provides that estimate.

Percentiles require a named reference. The 92nd percentile among one ancestry group, age range, or biobank may not equal the 92nd percentile in another. A report that omits its comparison population is incomplete.

Standard deviation or z-score

Some reports state that a person is, for example, 1.5 standard deviations above the population mean. This indicates position on the score distribution. Researchers often express disease odds per one-standard-deviation increase, but that average relationship does not by itself provide an individual probability.

Relative effect

A report may say the person has 1.8 times the odds or risk compared with a reference group. Odds and risk are not identical, especially for common outcomes. The time horizon also matters: five-year risk, ten-year risk, and lifetime risk are different quantities.

Relative effects can sound dramatic when baseline risk is low. Doubling a 1% risk produces a 2% risk, not a near certainty. Conversely, a modest relative increase applied to a high baseline can be clinically important.

Risk category

Commercial and clinical programs may label results low, average, intermediate, high, or very high. Those categories are model specific. Ask what cutoff defines each group and why that cutoff was chosen. The top 20%, top 10%, and top 1% are very different thresholds.

Integrated absolute risk

The most clinically useful reports may combine PRS with age, sex-related factors, family history, blood pressure, cholesterol, breast density, or other established predictors to estimate absolute risk. Such a model must be validated for the relevant population and clinical setting. Adding more variables does not guarantee accuracy if calibration is poor.

A good report should identify:

• the disease or phenotype predicted;
• the exact score name and version;
• the variants or method used;
• the reference and validation populations;
• the person’s percentile or standardized score;
• the type and time horizon of any risk estimate;
• performance by ancestry and sex when relevant;
• known limitations;
• the recommended clinical context; and
• whether the result has been interpreted in a regulated clinical laboratory.

A score based on self-reported disease in a research database may behave differently from one built using confirmed diagnoses, imaging, or pathology. The outcome definition is part of the test, not a minor technical detail.

Relative Risk and Absolute Risk

Absolute risk is the chance that an event will occur within a defined period. Relative risk compares one group with another. PRS reports often emphasize relative position because it is easier to calculate across datasets, but healthcare decisions usually require absolute risk.

Consider two people with the same high coronary PRS. One is 30 years old with normal blood pressure and cholesterol. The other is 70, smokes, has diabetes, and has high cholesterol. The genetic score may be identical, but their ten-year absolute risks are not. Age and current clinical factors dominate the near-term calculation for the older person.

The reverse is also important. A younger person can have low ten-year risk but high lifetime susceptibility. A PRS may identify inherited risk earlier than standard short-term calculators, potentially creating a longer window for prevention. Whether that should change medication or screening depends on evidence, not simply on the fact that risk appears early.

Absolute risk modeling usually requires:

1. A baseline disease rate for a population similar to the patient.
2. A validated relationship between the PRS and disease in that population.
3. Other predictors relevant to the condition.
4. A defined time horizon.
5. Adjustment for competing risks, such as death from other causes, when appropriate.

A percentile alone cannot supply these components.

A worked example

Suppose a report says a PRS is associated with 1.7-fold relative risk compared with the population average. If a clinically similar person has an estimated 10-year baseline risk of 5%, a rough multiplication suggests 8.5%, but real integrated models may produce a different result because they account for interactions, calibration, and competing factors. It would be inappropriate to use this simple calculation for treatment without a validated tool.

The distinction also prevents false reassurance. A person in the 20th percentile for breast cancer PRS may still have substantial risk because of age, a previous high-risk biopsy, chest radiation, a strong family history, or a pathogenic BRCA1 or BRCA2 variant. Standard screening recommendations may still apply.

A PRS should be viewed as one variable in a broader predictive genetic testing assessment. It shifts probability; it rarely settles it.

Where PRS May Be Used

Clinical interest is greatest where risk-stratified prevention already exists. The score must connect to an action such as screening age, screening frequency, preventive medication, or intensity of risk-factor management.

Coronary artery disease

Coronary PRS can identify people with inherited susceptibility that may not be obvious from cholesterol or family history. Potential uses include refining lifetime-risk discussions, encouraging earlier risk-factor assessment, or helping select people for preventive strategies. However, standard factors—LDL cholesterol, blood pressure, diabetes, smoking, age, and existing vascular disease—remain central. A high PRS does not diagnose blocked arteries, and chest pain requires clinical evaluation regardless of score.

Evidence that communicating a high score improves long-term outcomes is still developing. Lifestyle recommendations such as not smoking, exercising, maintaining healthy blood pressure, and managing cholesterol benefit people across PRS categories.

Breast cancer

Breast cancer PRS may refine risk models that also include age, reproductive history, family history, breast density, prior biopsy findings, and rare pathogenic variants. In selected programs, integrated risk could influence when screening begins or whether supplemental imaging or preventive medication is discussed. The score must be validated in the population being tested, and it should not replace BRCA1 and BRCA2 genetic testing when personal or family history suggests hereditary cancer.

Prostate cancer

A prostate cancer PRS may identify men more likely to develop disease and potentially inform the timing of prostate-specific antigen discussions. The challenge is distinguishing clinically significant cancer from indolent disease. A score that increases detection without improving meaningful outcomes can lead to overdiagnosis and unnecessary procedures.

Type 2 diabetes

Diabetes PRS can stratify inherited susceptibility, but body weight, age, family history, pregnancy history, medications, sleep, activity, and measured glucose remain powerful predictors. Clinical utility depends on whether the score changes prevention beyond recommendations already indicated by conventional risk.

Other conditions

Scores are being studied for colorectal cancer, atrial fibrillation, glaucoma, Alzheimer disease, inflammatory disorders, psychiatric conditions, and many traits. Research validity is not the same as routine clinical readiness. For outcomes without proven prevention or with difficult risk communication, testing may create anxiety without a clear benefit.

Medication and treatment response

Some polygenic models aim to predict treatment response or adverse effects. These are different from established single-gene pharmacogenetic testing. Most treatment-response PRS tools remain investigational and should not be assumed to have the same evidence as a guideline-backed gene–drug pair.

A reasonable clinical use case should answer four questions:

• Who should be tested?
• Which validated score should be used?
• What result crosses an action threshold?
• What intervention has evidence of net benefit for that group?

Without those answers, the result may be informative but not actionable.

Ancestry, Calibration, and Equity

Ancestry is one of the most important limitations in current PRS performance. Many large genetic studies have overrepresented people of European genetic ancestry. A score developed in that setting often loses accuracy when applied to people with different ancestry because allele frequencies, linkage disequilibrium, environmental context, and disease rates differ.

Genetic ancestry is not the same as race or ethnicity. Race is a social classification; ethnicity includes culture, geography, and identity. Genetic ancestry estimates statistical similarity to reference populations and can be continuous and mixed. A person may have ancestry from several regions, making a single category inadequate.

Poor transferability can cause several errors:

• Percentiles may be systematically shifted.
• A high-risk cutoff may classify too many or too few people.
• Relative effects may be weaker than expected.
• Absolute-risk estimates may be miscalibrated.
• Health disparities may worsen if the most accurate tools are available mainly to populations already well represented in research.

Calibration can adjust the score distribution and risk estimates for a target population, but it does not automatically recover all lost predictive performance. Multi-ancestry GWAS, ancestry-aware methods, larger diverse cohorts, and local validation can improve fairness. The report should show evidence for the person’s relevant ancestry group rather than simply state that the test is “for everyone.”

This limitation does not mean PRS can never be used in underrepresented populations. It means the uncertainty must be measured, disclosed, and considered before acting. A clinician should ask:

• Which ancestries were included in score development?
• Which groups were used for independent validation?
• Was the score recalibrated for this patient population?
• How does discrimination and calibration differ across groups?
• Is there a safer established clinical model if performance is uncertain?

Social determinants also affect absolute risk and access to prevention. A DNA score does not measure food access, environmental exposures, chronic stress, discrimination, healthcare access, or neighborhood conditions. A model that ignores those factors may look biologically precise while missing major drivers of health.

Equity also involves downstream care. Offering a test without ensuring access to confirmatory evaluation, imaging, medications, or prevention can create information without benefit. Clinical programs should evaluate who is offered testing, who receives results, who completes follow-up, and whether outcomes improve across groups.

Limitations and Common Errors

A PRS is highly dependent on context. Removing the number from its model and population can make it meaningless.

Treating percentile as probability

A 95th-percentile result means the score ranks higher than 95% of the reference group. It does not mean a 95% chance of disease. This is the most common interpretation error.

Assuming one score is interchangeable with another

Hundreds of PRS models may exist for one disease. They differ in variant count, statistical method, phenotype definition, ancestry, and validation. A result cannot be reproduced merely by ordering “a PRS test.” The exact model must be specified.

Ignoring calibration

A model can rank people correctly yet overestimate or underestimate actual risk. This is the difference between discrimination and calibration. Both are needed for clinical risk prediction.

Using a score outside its intended population

A model built for adults may not apply to children. A score for incident disease may not apply after diagnosis. A model validated in a population screening cohort may not work in a specialty clinic with much higher baseline risk.

Letting low polygenic risk override high-impact information

A low PRS should not cancel a pathogenic single-gene variant, a striking family history, symptoms, or established clinical risk factors. Rare monogenic disease and common polygenic susceptibility can coexist.

Assuming genes are destiny

Most common-disease PRSs explain only part of population variation. People with high scores may never develop disease; people with low scores may. Environment, behavior, healthcare, random biological events, and unmeasured genetics all contribute.

Acting without an evidence-based pathway

A high score may tempt people to request imaging, medication, or invasive screening earlier than recommended. Extra testing can create false positives, overdiagnosis, radiation exposure, complications, and cost. The action should be based on demonstrated net benefit, not fear.

Overinterpreting consumer reports

Direct-to-consumer services may calculate scores from genotyping arrays and imputed variants rather than directly measuring every marker. Results can change when the company updates its algorithm. Some products provide wellness or trait scores with limited clinical validation. A result that would change medical care should be reviewed and, when needed, confirmed through an appropriate clinical pathway.

Forgetting that the score is versioned

PRS methods evolve. Variant weights, genome builds, imputation panels, and reference datasets can change. Keep the report date and version. A new result should not be compared numerically with an old one unless the models are the same.

Psychological effects also deserve attention. A high score can cause fatalism; a low score can encourage complacency. Communication should emphasize that the result modifies risk and that many preventive actions remain useful across the distribution.

Choosing and Using a PRS Test

Before ordering, begin with the clinical decision. A test is more useful when the clinician can state what would change at a high, average, or low result.

Ask the laboratory or program:

1. What exact disease outcome does the score predict?
2. Which PRS model and version are used?
3. Was the score independently validated?
4. How well does it perform for my genetic ancestry and clinical group?
5. Is the result a percentile, relative effect, or calibrated absolute risk?
6. What clinical factors are combined with the score?
7. Which professional guideline supports the proposed action?
8. Does the laboratory confirm genotypes used for high-stakes decisions?
9. How are data stored, shared, or reanalyzed?
10. What happens if the model is updated?

After receiving a result, review it with a clinician familiar with the disease and, when appropriate, a genetic counselor. The review should include current health measurements, personal history, family history, and any indication for diagnostic or single-gene testing. A PRS is not a substitute for evaluating a family pattern of early cancers, extreme cholesterol, cardiomyopathy, sudden death, or another possible Mendelian disorder.

A practical result discussion separates three levels:

• Analytical result: the measured variants and calculated score.
• Risk interpretation: the person’s rank and estimated disease risk in a defined population.
• Management plan: the evidence-based screening, prevention, or monitoring step, if any.

If no management plan changes, the test may still contribute to research or personal understanding, but that is different from clinical utility.

Keep the complete report. Record the reference population, percentile, absolute-risk model, and date. Do not rely on a screenshot that says only “high risk.” Share the result with clinicians only when it is relevant, and ask that they avoid converting a probabilistic score into a disease diagnosis in the medical record.

Reassessment may be reasonable if the score is incorporated into a newer validated model, ancestry calibration improves, or clinical guidelines establish a new pathway. Routine repeat DNA collection is usually unnecessary because inherited markers do not change; recalculation from existing genotype data may be possible. However, a laboratory may need a new test if the original array did not measure or impute the required variants accurately.

Regardless of PRS category, respond promptly to symptoms and follow established screening recommendations unless a qualified clinician documents a supported alternative. High genetic susceptibility can strengthen the case for prevention, but low susceptibility never makes a person immune.

The most responsible use of a polygenic risk score is modest and specific. It can refine risk for selected common diseases, particularly when integrated with established predictors. It cannot replace diagnosis, clinical judgment, or the broad social and biological context that shapes health.

References

• The clinical application of polygenic risk scores: A points to consider statement of the American College of Medical Genetics and Genomics (ACMG) 2023 (Position Statement)
• The clinical utility of polygenic risk scores in genomic medicine practices: a systematic review 2022 (Systematic Review)
• Challenges in clinical translation of polygenic risk score analyses: A systematic review 2025 (Systematic Review)
• Selection, optimization and validation of ten chronic disease polygenic risk scores for clinical implementation in diverse US populations 2024
• Ancestry Calibration of Polygenic Risk Scores Improves Risk Stratification and Effect Estimation in African American Adults 2025
• Clinical implementation of polygenic risk scores 2026 (Review)

Disclaimer

This information is educational and does not provide a diagnosis or a personalized screening or treatment recommendation. PRS results require interpretation in the context of the exact model, ancestry performance, personal and family history, and established clinical risk factors. Do not delay evaluation of symptoms or change screening or medication solely because of a high or low polygenic score.