Retinal Artificial Intelligence Reclassifies SCORE2-Based Cardiovascular Risk Across the ESC/EAS Statin-Indication Boundary: A UK Biobank Prospective Cohort Study
Main finding · ASCVDResult — go to the resultSCORE2-adjusted HR 2.32Authors' conclusion
Adding the retinal AI risk score to SCORE2 produced modest bidirectional risk reclassification across the ESC/EAS statin-indication boundary; these findings warrant prospective validation. Abstract word count: 250 words
Study context
- Population
- Statin-naive UK Biobank participants aged 40–69 without diabetes
- Participants
- 40,672 statin-naive UK Biobank participants
- Exposure
- Adding the retinal artificial intelligence (AI) risk score to SCORE2
- Outcome
- ASCVD
- Follow-up
- 10.0-year median
Methodological detailsUK Biobank, 10-fold cross-validation
- UK Biobank
- 10-fold cross-validation
Main finding
As reported in the abstract
SCORE2-adjusted HR
2.32
P < 0.001
- Atherosclerotic cardiovascular disease (ASCVD) events
- 914
Observational data: associations, not effects.
Further findings
| Outcome | SCORE2-adjusted HR | P value |
|---|---|---|
| ASCVD in the statin-indicated subgroup, by retinal AI risk groupsComparator groups not stated · subgroup | 2.16 | P < 0.001 |
Reclassification across the ESC/EAS statin-indication boundary (retinal AI risk score added to SCORE2)
- Statin-eligibility NRI
- 0.012, p<0.001
- Continuous NRI
- 0.445, p<0.001
- Without C-index or net-benefit gain
- Reclassification was statistically significant but modest
Observed 10-year risk: statin-not-indicated participants in high or very high retinal AI risk groups vs all statin-indicated, untreated participants
- Numerically similar observed 10-year risk
- P = 0.758
Effect modification by clinically detectable retinal disorder
- No evidence of effect modification
- Interaction p=0.93
Incremental value by sex
- Incremental value was greater in women than men
- Sex interaction p<0.001
Original publication
Summarised from the published abstract; every number is checked against it