Skip to main content

Table 4 Prognostic performance with the addition of Reti-CVD to the QRISK3 in the UK biobank dataset

From: Validation of a deep-learning-based retinal biomarker (Reti-CVD) in the prediction of cardiovascular disease: data from UK Biobank

 

Non-statin cohort (n = 42,473)

 

Stage 1 hypertension cohort (n = 11,966)

 

Middle-aged Cohort (n = 38,941)

 

Models

C statistic (95% CI)

p value

C statistic (95% CI)

p value

C statistic (95% CI)

p value

Reti-CVD

0.624 (0.615–0.633)

NA

0.591 (0.575–0.608)

NA

0.611 (0.600 -0.622 )

NA

Reti-CVD plus age, gender

0.699 (0.690–0.709)

NA

0.663 (0.644–0.681)

NA

0.681 (0.669–0.692)

NA

QRISK3

0.682 (0.672–0.692)

NA

0.639 (0.620–0.658)

NA

0.650 (0.638–0.662)

NA

Reti-CVD plus QRISK3

0.696 (0.686–0.706)

NA

0.652 (0.633–0.671)

NA

0.674 (0.661–0.686)

NA

 Δ Reti-CVD plus QRISK3 versus QRISK3

0.014 (0.010–0.017)

< 0.001

0.013 (0.007–0.019)

< 0.001

0.023 (0.018–0.029)

< 0.001

NRI

 Continuous NRI (95% CI)

0.133 (0.088–0.173)

< 0.001

0.094 (0.008–0.174)

0.033

0.248 (0.190–0.301)

< 0.001

  1. Data are C statistic (95% CI), unless stated otherwise. The Reti-CVD plus QRISK3 model is a logistic model fit on the UK Biobank. NRI for Reti-CVD plus QRISK3 versus QRISK3 models was provided
  2. NRI net reclassification index, CI confidence interval, CVD cardiovascular disease, Reti-CVD deep-learning-based retinal CVD biomarker