Which subgroups carry which cancer risk
Status. Research use. Population-level analysis. Not a medical device, and it produces no output about any individual.
The setting
Whether someone develops a particular cancer is shaped by their genetics, how they live, and the circumstances they live in. Those three things are recorded in three different systems that were never designed to be read together, so the question of which subgroups carry which risk is rarely asked at population scale.
What we did
Within CODIGE, a project funded under the AGE-IT extended partnership (PNRR M4C2, project code PE00000015), we developed one of the algorithms in the delivery. It links genotype, lifestyle and socioeconomic variables to cancer subgroup incidence at population level.
That claim is deliberately narrow. It says what we built. It does not say the project was ours, and it invites a reader to check PE00000015, which identifies AGE-IT publicly.
What exists now
A delivered algorithm within a completed project. There is no published evaluation of it.
What it does not do
- No individual risk score. No screening. No eligibility. No output that resolves to one person.
- This is the same boundary as the cohort study above, and for the same reason: a population-level association is a research finding, while individual stratification is a clinical claim carrying an entirely different regulatory and data-protection weight.
What has not been established
- Whether the associations replicate outside the cohort they were built on.
- Whether any of them changes a decision anywhere.
- There is no published evaluation.
Governance
- Genotype combined with socioeconomic variables on identifiable people is Article 9 special-category processing, and at population scale it is squarely Article 35(3)(b) territory, which requires a data protection impact assessment.
- The legal basis sat with whoever held the data, not with us. Whoever that is should be the party describing it.
Where it goes next
The associations were built on one cohort and have not been tested anywhere else, which is the first thing a reader should ask about work of this kind.
What we are looking for
A second cohort to replicate the associations on, and co-authors to publish the result either way. A failed replication is a publishable finding here, and saying so up front is part of the ask.
Get in touch →