A polygenic score combines information from many genetic variants associated with a trait or disease. It can help distinguish groups with different average risks, but it is not a diagnosis and does not incorporate every source of risk. The clinical question is whether a specific score improves decisions for a specific population and use case.
Ask what the reported number means
A percentile locates someone within a reference distribution. It does not directly state their probability of developing disease. Absolute risk also depends on the relevant baseline risk, time horizon and other predictors. A report should make clear whether it supplies a raw score, a relative association, or an externally validated estimate of absolute risk.
The Polygenic Risk Score Reporting Standards describe information needed to assess how scores are developed and validated. This includes the underlying data, the population studied and the performance evaluation. Reproducible reporting is a prerequisite for appraisal; it is not proof of clinical utility. Read the reporting standards.
Evaluate transferability and fairness
Martin and colleagues demonstrated how ancestry imbalances in genetic research can limit score performance across populations. A score’s behavior cannot be assumed to remain constant when transferred to a different population or setting. The issue is not resolved by assigning a person a broad racial label: the reference data, genetic structure and clinical environment all deserve consideration. Read the disparities analysis.
Ask whether validation included people similar to those who will receive the score, and whether risk estimates were calibrated in the intended setting. A model can rank individuals reasonably while systematically overestimating or underestimating their absolute risk. These performance properties should be assessed separately.
Connect performance to a decision
- What action would change because of the score?
- Does it add useful information beyond established clinical predictors?
- What are the consequences of false reassurance or unnecessary intervention?
- Is there evidence that the proposed use improves outcomes?
Those questions prevent statistical performance from being mistaken for patient benefit. Research progress may justify testing a new care pathway without establishing that the pathway is already standard practice. Clinical summaries should distinguish model development, external validation and prospective utility evidence, and should identify the remaining uncertainty for the actual decision being considered.