The impact of direct-to-consumer personal genomic testing on perceived risk of breast, prostate, colorectal, and lung cancer: findings from the PGen study

Background: Direct access to genomic information has the potential to transform cancer risk counseling. We measured the impact of direct-to-consumer genomic risk information on changes to perceived risk (Delta PR) of breast, prostate, colorectal and lung cancer among personal genomic testing (PGT) customers. We hypothesized that Delta PR would reflect directionality of risk estimates, attenuate with time, and be modified by participant characteristics. Methods: Pathway Genomics and 23andMe customers were surveyed prior to receiving PGT results, and 2 weeks and 6 months post-results. For each cancer, PR was measured on a 5-point ordinal scale from "much lower than average" to "much higher than average." PGT results, based on genotyping of common genetic variants, were dichotomized as elevated or average risk. The relationship between risk estimate and Delta PR was evaluated with linear regression; generalized estimating equations modeled this relationship over time. Results: With the exception of lung cancer (for which Delta PR was positive regardless of result), elevated risk results were significantly associated with positive Delta PR, and average risk results with negative Delta PR (e.g., prostate cancer, 2 weeks: least squares-adjusted Delta PR = 0.77 for elevated risk versus -0.21 for average risk; p-value(difference) < 0.0001) among 1154 participants. Large changes were rare: for each cancer, <4 % of participants overall reported a Delta PR of +/- 3 or more units. Effect modification by age, cancer family history, and baseline interest was observed for breast, colorectal, and lung cancer, respectively. A pattern of decreasing impact on Delta PR over time was consistently observed, but this trend was significant only in the case of colorectal cancer. Conclusions: We have quantified the effect on consumer risk perception of returning genetic-based cancer risk information directly to consumers without clinician mediation. Provided via PGT, this information has a measurable but modest effect on perceived cancer risk, and one that is in some cases modified by consumers' non-genetic risk context. Our observations of modest marginal effect sizes, infrequent extreme changes in perceived risk, and a pattern of diminishing impact with time, suggest that the ability of PGT to effect changes to cancer screening and prevention behaviors may be limited by relatively small changes to perceived risk.