We calculated a genetic score for type 2 diabetes in every member of a family of four. We published those scores as percentiles in one table. In the next table, which translated scores into risk comparisons, the entire type 2 diabetes row was blank. The explanation sits in the footnote: those scores fell below the thresholds for which the source study reported risk comparisons. The blanks did not mean zero risk. Tables 5 and 6 of our study.
This was part of a 2022 paper I wrote with Karyn Megy, Antonio Metastasio and Edmund Lehmann. We selected published polygenic risk scores for 15 conditions and applied them to whole-genome data from the family. Our question was practical: how do you take a score developed in a research study and use it to describe an individual’s genetic risk?
A polygenic risk score combines the contributions of many DNA variants, weighted according to their associations with a particular condition. To put an individual’s score in context, we compared it with scores calculated for participants in the 1000 Genomes Project. The resulting percentile tells you where the score sits within that comparison group. It does not tell you the percentage chance that the person will develop the disease. Our calculation method.
That next step required returning to the original studies. The studies did not all report the same comparisons. Some described the highest-scoring fifth of participants compared with everyone else. Others supplied comparisons further down the distribution. Our glaucoma entries included a comparison of the top half with the rest; our colorectal cancer entries used the top quarter. These were different reporting boundaries, not interchangeable definitions of illness. Table 6 and the accompanying results.
We could therefore calculate a score without having a reported risk comparison to assign under our method. The distinction matters when reading the table. An empty cell says something about the information available through that reporting rule. It does not establish that a person is free of the condition, or will remain so. Equally, a filled cell contains a comparison drawn from a group in a published study, not a direct measurement of that individual’s future.
Our own study was a proof of concept involving four related people of Iberian origin. It did not establish that using these reports improves health outcomes. We also acknowledged that the scores came from studies developed and validated in Northern European populations, which could affect their application to this family. Those limits belong alongside the results. Study limitations.
I would make the table easier to read by replacing each blank with an explicit explanation: no risk comparison assigned under the study’s reporting rule. That is a proposal for clearer reporting, not something we tested. A reader should be able to see whether a genetic score was calculated and what evidence supports the statement made from it. The footnote should not have to do all that work.
Disclosure: Cambridge Precision Medicine funded the study. All four authors, including me, declared an association with the company at the time. Published declarations.
Further reading
Manuel Corpas, Karyn Megy, Antonio Metastasio and Edmund Lehmann (2022). Implementation of individualised polygenic risk score analysis: a test case of a family of four. BMC Medical Genomics. Published case study.
The 1000 Genomes Project Consortium (2015). A global reference for human genetic variation. Nature. Published population reference study.
Leave a Reply