Table 2 PRS compared for each outcome and their performance characteristics in the UK Biobank.

From: Assessing agreement between different polygenic risk scores in the UK Biobank

Disease (N)

PRS

nSNPs

OR (95% CI)

Crude-AUC (95% CI)

Multi-AUC (95% CI)

NRI (95% CI)

r

LD: R2 > 0.8 (% overlap)

Breast cancer (171,490)

A

313

3.41 (2.89, 4.03)

0.63 (0.63, 0.64)

0.64 (0.63, 0.64)

0.03 (0.00, 0.05)

0.65

225 (72%)

B

118,388

3.94 (3.36, 4.61)

0.64 (0.63, 0.64)

0.64 (0.63, 0.65)

Hypertension (317,581)

A

267

1.83 (1.70, 1.98)

0.57 (0.56, 0.57)

0.69 (0.69, 0.69)

0.16 (0.16, 0.17)

0.66

210 (79%)

B

884

2.18 (2.02, 2.35)

0.59 (0.59, 0.59)

0.70 (0.70, 0.70)

Dementia (335,689)

A

57

1.67 (1.30, 2.13)

0.55 (0.54, 0.56)

0.80 (0.79, 0.80)

0.13 (0.10, 0.16)

0.51

13 (23%)

B

39

2.30 (1.85, 2.87)

0.57 (0.56, 0.57)

0.80 (0.79, 0.81)

  1. N: number of participants whose PRS score was obtained. nSNPs: number of SNPs in PRS prior to genetic quality control. OR: odds ratio for top 1% versus middle quintile of PRS from multivariable logistic regression model adjusted for age, sex, genotyping array and first 5 PCs. AUC: area under receiver-operating curve. Crude-AUC: only continuous PRS was fitted in a regression model. Multi-AUC: continuous PRS was fitted, further adjusted for age and sex. NRI: continuous net reclassification index obtained from predicted risks by two multivariable logistic regression models that contain age, sex, continuous PRS for this disease, genotyping array and first 5 PCs. The model containing PRS-B is considered the “updated” model. \(r\): Pearson correlation coefficient between the two continuous PRS for this disease. LD: number (%) of SNPs in PRS-A which either appear in or are in linkage disequilibrium (R2 > 0.8) with SNPs in PRS-B. Breast cancer models are not adjusted for sex because its population is restricted to females. 95% CI for AUC and NRI calculated by bootstrapping.
Back to article page