Table 2 External validation of PRS models in European populations using data from UK Biobank and CIMBA.

From: Polygenic risk modeling for prediction of epithelial ovarian cancer risk

Model (data set)

SNPs

UK Biobank

CIMBA BRCA1 carriersa

CIMBA BRCA2 carriersa

  

AUC

OR

95% CI

AUC

HR

95% CI

AUC

HR

95% CI

 

(a) PRS models based on OCAC data

       

Lasso (OCAC)

1403

0.587

1.37

1.27–1.48

0.573

1.27

1.21–1.34

0.627

1.48

1.33–1.63

Elastic net (OCAC)

10,797

0.588

1.36

1.26–1.47

0.583

1.32

1.26–1.39

0.617

1.47

1.33–1.63

Stepwise (OCAC)

22

0.588

1.35

1.26–1.46

0.563

1.21

1.16–1.26

0.605

1.39

1.26–1.54

Select and shrink (OCAC)

27,240

0.588

1.38

1.28–1.48

0.592

1.36

1.29–1.43

0.624

1.49

1.35–1.64

(b) PRS models based on meta-analysis of OCAC and CIMBA data

    

Stepwise (OCAC-CIMBA)b

36

0.595

1.39

1.29–1.50

NA

NA

NA

NA

NA

NA

 

Select and shrink (OCAC-CIMBA)

18,007

0.596

1.42

1.32–1.54

NA

NA

NA

NA

NA

NA

 
  1. AUC area under the receiver operating characteristic curve, OR odds ratio, HR hazards ratio.
  2. aEstimates are from unadjusted models.
  3. bResults in CIMBA are overfitted as the CIMBA data was used for model development.