R. Wellmann, K. Danahey, S. Hussain, M. J. Ratain, P. H. O'Donnell; The University of Chicago, Chicago, IL

BACKGROUND: Germline and tumor PGx factors both impact oncology drug responses, but germline PGx less commonly guides oncology prescribing. We hypothesize that the number of clinically actionable germline PGx associations in oncology is modest despite an abundance of research in the field.
METHODS: We analyzed 125 oncology drugs for author-reported ‘positive’ germline PGx associations in journals with impact factors ≥ 5. Studies were critically assessed for design quality, genotyping quality, clinically relevant outcomes, evidence of drug-gene (rather than gene-disease) effect, and statistical significance. Associations from studies meeting these criteria were deemed potentially clinically actionable.
RESULTS: 49/125 drugs (39.2%) had ≥ 1 published positive PGx association, representing 138 publications critically assessed. Median sample size was n=161. Progression-free survival was the most common outcome (41% of studies). Valid associations were detected for 18 drugs (14%), including 5 with FDA label PGx information or published PGx guidelines (capecitabine/fluorouracil, irinotecan, mercaptopurine/thioguanine). Only 10 associations were replicated. The most common flaw inhibiting potential actionability was lack of statistical significance after multiple test correction (63% of studies), followed by prognostic associations with no controls.
CONCLUSION: A modest, but not insignificant, number of oncology drugs have potentially actionable germline PGx associations, but many other positive associations lack actionability primarily due to analysis in underpowered studies. Clinically actionable associations were translated into decision supports, now being delivered to oncologists in a larger PGx implementation study, to determine their clinical utility.