Contribution of Common SNPs to Variability in Late Radiation Therapy Toxicity in Prostate Cancer

Abstract

Late toxicities following prostate cancer RT are dose-limiting and negatively affect quality of life. Inter-individual variability in the development of these toxicities is partly due to genetics. Several single nucleotide polymorphisms (SNPs) have been identified, but have modest effect sizes, so larger studies are needed to identify further associations. Also, it is not known whether the genetic contribution to late toxicities is mainly from common SNPs with modest effects or rarer, unknown variants that may have large effects. It is also not known whether SNPs contribute equally to different toxicities. We analyzed many genome-wide SNPs simultaneously in two prostate cancer cohorts from the RAPPER and RADIOGEN studies to estimate the variance in late RT toxicities explained by common SNPs. Participants received external beam RT and were followed prospectively for development of late toxicity toxicities between from 6 months to 5 years post-RT (measured as cumulative incidence) in RAPPER (LENT-SOM, RMH scales) and RADIOGEN (CTCAEv3.0). Selected binary endpoints were urinary frequency (increase ≥ 2x pre-RT or requiring treatment), hematuria (macroscopic or worse), and rectal bleeding (mild or worse). Logistic regression using restricted maximum likelihood estimated the variance in each endpoint explained by common (frequency > 1%) SNPs (methods in Yang et al, Nature Genetics 2010). A likelihood ratio test compared each model with genetic components to the reduced model without genetic components. Analyses were adjusted for age at RT, total biologically effective dose, hormone use (in RADIOGEN; all RAPPER participants received hormones), and pre-RT symptoms (in RAPPER; toxicity grade in RADIOGEN accounts for pre-RT symptoms). Analysis of rectal bleeding was also adjusted for diabetes and percent rectal volume receiving ≥65Gy. In RAPPER, 140/1707 patients (8%) reported increased urinary frequency, 45/1705 (3%) frank hematuria, and 199/1542 (13%) rectal bleeding. In RADIOGEN, 155/569 patients (27%) reported increased urinary frequency, 39/569 (7%) hematuria, and 131/569 (23%) rectal bleeding. The proportion of variance in toxicity explained by common SNPs was 0.35 (standard error, se 0.20, P=0.035) in RAPPER and 0.36 (se 0.58, P=0.28) in RADIOGEN for urinary frequency; 0.08 (se 0.21, P= 0.38) in RAPPER and 0.01 (se 0.54, P=0.5) in RADIOGEN for hematuria, and 0.11 (se 0.22, P=0.31) in RAPPER and 0.27 (se 0.55, P=0.30) in RADIOGEN for rectal bleeding. Common, known SNPs explain approximately 35% of the variance in late urinary frequency among men treated with RT for prostate cancer. Common SNPs explain less variance for late rectal bleeding and hematuria, suggesting that dose and other clinical factors are more important. Rare variants, not captured on SNP arrays, likely explain an additional proportion of variance in all toxicities.