Detecting mutation rate change in Barrett’s Esophagus after treatment with NSAIDs

B29

BACKGROUND: To prevent or delay progression to cancer, it is vital to assess the effect of exposures such as alcohol, tobacco smoke, or non-steroidal anti-inflammatory drugs (NSAIDs) on the clonal evolution that drives neoplastic progression. NSAIDs have been shown to reduce the risk of esophageal adenocarcinoma (EA) in Barrett’s Esophagus (BE) patients (Vaughan et al., Lancet Oncol. 2005;6(12):945-52). However, the mechanism through which NSAIDs delay progression to EA is not well understood.
 >Four parameters determine the rate of evolution of neoplastic clones: mutation rate, stem cell population size and generation time, and clonal expansion rate. We hypothesize that NSAID use slows clonal evolution by reducing mutation rate. To test this hypothesis in-vivo, Illumina SNP arrays can be used to detect loss of heterozygosity (LOH) events in frozen longitudinal biopsies from BE patients from the Seattle Barrett’s Esophagus cohort. Using coalescent theory, the observed LOH events, known sampling times and known onset of NSAIDs treatment we can estimate the rate of LOH per year, the stem cell population size, and changes in these parameters after treatment.
 >METHODS: We performed in-silico analysis of the power to detect mutation rate changes, which is critical for selecting the number of time points and biopsies per time point to assay per patient. We modified SerialSimCoal (Excoffier et.al. J. Hered., 91, 506-509) to simulate a scenario in which 4 biopsies were taken from patients every 2 years for 10 years, with NSAID use beginning in year 5, and asked whether we could reliably detect mutation rate changes of various intensities (1.1, 2, 5, 10, and 100-fold). We varied stem cell population sizes and base mutation rates over 3 orders of magnitude on DNA sequences of 8,000 loci. We used BEAST (Drummond et.al. Genetics 161, 1307-1320) to estimate population sizes and mutation rates before and after year 5.
 >RESULTS: For 91% (41/45) of parameter combinations, we had >80% power to detect a 1.1- to 100- fold decrease in mutation rate (paired Wilcoxon test, p<0.05). In the four worst cases, a high stem cell population size and low base mutation rate, yielded broad confidence intervals for the estimated mutation rates and resulted in >70% power to detect 1.1- to 10- fold decrease in mutation rate.
 >CONCLUSION: These results allow for the design of longitudinal studies to be carried out in BE cohorts to determine the in-vivo effects of NSAIDs and other exposures. Our results suggest that assaying 24 longitudinal samples will give >80% power to detect as small as a 1.1-fold change in mutation rate associated with an intervention. We are currently adapting these methods to use SNP arrays to measure the rate of loss of heterozygosity events, number of stem cells and time since initiation in neoplastic progression.