Genome‐wide methylation profiling of diagnostic tumor specimens identified DNA methylation markers associated with metastasis among men with untreated localized prostate cancer

BACKGROUND: We used a genome‐wide discovery approach to identify methylation markers associated with metastasis in men with localized prostate cancer (PCa), as better identification of those at high risk of metastasis can inform treatment decision‐making. METHODS: We identified men with localized PCa at Kaiser Permanente California (January 1, 1997–December 31, 2006) who did not receive curative treatment and followed them for 10 years to determine metastasis status. Cases were chart review‐confirmed metastasis, and controls were matched using density sampling. We extracted DNA from the cancerous areas in the archived diagnostic tissue blocks. We used Illumina's Infinium MethylationEPIC BeadChip for methylation interrogation. We used conditional logistic regression and Bonferroni's correction to identify methylation markers associated with metastasis. In a separate validation cohort (2007), we evaluated the added predictive utility of the methylation score beyond clinical risk score. RESULTS: Among 215 cases and 404 controls, 31 CpG sites were significantly associated with metastasis status. Adding the methylation score to the clinical risk score did not meaningfully improve the c‐statistic (0.80–0.81) in the validation cohort, though the score itself was statistically significant (p < 0.01). In the validation cohort, both clinical risk score alone and methylation marker score alone are well calibrated for predicted 10‐year metastasis risks. Adding the methylation score to the clinical risk score only marginally improved predictive risk calibration. CONCLUSION: Our findings do not support the use of these markers to improve clinical risk prediction. The methylation markers identified may inform novel hypothesis in the roles of these genetic regions in metastasis development.