General Departmental Seminar Series
Using genomic data to optimize the choice of chemotherapy
regimen for metastatic colorectal cancer.
Dave Vanness, PhD,
Dept. of Population Health Sciences,
University of Wisconsin-Madison
December 17, 2004, 12 - 1 pm in room G5/113 Clinical Sciences Center (600 Highland Ave.)
Perhaps the greatest promise of pharmacogenomics is the potential to optimize treatment choices based on an individual's genetic profile. A recently-published randomized, controlled trial recommended that oxaliplatin and infused fluorouracil plus leucovorin (FOLFOX) become the standard therapy for patients with untreated metastatic colorectal cancer [Goldberg et al. 2004]. For 575 patients from that study, single-nucleotide polymorphisms (SNPs) were identified on four candidate genes hypothesized to affect tumor response, progression, survival, toxicity and consequently, quality-adjusted longevity (QALY). If genetic heterogeneity underlies some of the differences in outcomes among the treatment options, it may be possible to maximize individual anticipated utility by conditioning the treatment choice on each individual's genetic profile. We use classification and regression tree methods to identify potential treatment-SNP interactions for inclusion in Markov Chain Monte Carlo generalized linear models relating progression, survival, toxicity and quality-adjusted survival to individual characteristics and assigned treatment. Samples from the Markov chains are used in Monte Carlo simulations to generate posterior distributions of QALY, conditional on individual characteristics, for counterfactual treatment options. Bayesian decision theory is then applied to the treatment choice under a variety of individual objective functions that consider not only expected (mean) QALY, but other features of the posterior distribution. Preliminary results suggest that the posterior probability that a specific regimen is optimal for an individual depends on the presence of certain SNPs, particularly if that individual places higher value on the tails of the QALY distributions.
Return to seminar list