Fig. 4: Validation of the PGS_PSA in two cancer prevention trials.

a–f, Performance of PGS_PSA was evaluated in the PCPT and the SELECT. a,b, Violin plots show the distribution of baseline log(PSA) within quantiles of PGS_PSA, comprising 1,058,173 and 1,071,278 variants in the PCPT (a) and the SELECT (b). Box plots extend from the 25th to the 75th percentiles, with a trend line connecting the median value within each age stratum. Two-sided P values were derived from linear regression models for the effect of a quantile increase in PGS_PSA on log(PSA). c,d, Crossbar plots show the effect estimates (β) and corresponding 95% CIs per s.d. increase in the standardized PGS_PSA on baseline log(PSA) in the PCPT (c) and the SELECT (d). Ancestry-stratified and pooled multi-ancestry estimates are presented. Two-sided P values based on linear regression models are annotated. e,f, Comparison of distributions for PSA and PSA^G, with the horizontal line at 4 ng ml⁻¹, a commonly used threshold for further diagnostic testing. Box plots show the median value, with lower and upper hinges corresponding to the 25th and 75th percentiles or first and third quartiles. Whiskers extend as a multiple of the interquartile range (IQR × 1.5). Outlying values beyond the end of the whiskers are plotted individually.