Page 314 - Personalised medicine of fluoropyrimidines using DPYD pharmacogenetics Carin Lunenburg

Chapter 12
Discussion
It is well recognized that DPYD genotyping is useful in preventing severe fluoropyrimidine- induced toxicity by applying initial dose reductions in patients who carry a specific DPYD variant.5 However, not all toxicity can be predicted and prevented by the current four DPYD variants. Indeed, still ~20% of the patients experience toxicity, thus the search for genetic variants predictive for severe fluoropyrimidine-induced toxicity continues. We executed a GWAS in order to identify novel genetic variants possibly associated with the onset of severe fluoropyrimidine-induced toxicity.
To perform this GWAS on toxicity of fluoropyrimidines, over 1,100 patients were genotyped. Severe toxicity includes the National Cancer Institute CTC-AE grades 3─5. Scoring clinical toxicities can be difficult sometimes, as it can be open to interpretation. As severe toxicity has the most clinical impact, we chose to maximize contrast to the toxicity endpoint, and we excluded patients with grade 2 toxicity in the primary analysis. We were unable to identify genome-wide significant SNPs, yet we identified six SNPs suggestive (p-value of 5x10-6) of association with severe fluoropyrimidine-induced toxicity. Possibly the number of patients in our study is too small to reach genome-wide significance. However, we repeated the analysis including the patients with grade 2 toxicity, increasing the number of patients while reducing the contrast between toxicities. Yet, this did not result in a different outcome. The suggestive variants need to be re-tested in an independent cohort of patients who were treated with a fluoropyrimidine drug.
This GWAS was executed using DNA from patients participating in the Alpe DPD study. A formal comparison of GWAS analysed patients with the entire Alpe DPD cohort shows the cohorts were comparable. The range of number of treatment cycles was statistically different, with fewer cycles in the GWAS cohort. Possibly this is due to the exclusion of patients with grade 2 toxicity in the GWAS cohort, as grade 2 toxicity, if not developing into severe toxicity, may possibly arise from longer periods of fluoropyrimidine-treatment. We have no reason to believe that selection bias was introduced by leaving out patients with grade 2 toxicity. We believe the GWAS cohort is representative for patients in daily clinical care, as in the Alpe DPD study there were only limited restrictions on the inclusion criteria and the burden for patients to participate was very low.
With a large amount of genotyping data, we were able to compare ethnicity strings in the MDS plots to self-reported ethnicity from the Alpe DPD study. When adding two principal components, including ethnicity, to the statistical analysis, no differences were visible. Therefore, ethnicity was not of influence on the outcome of this GWAS and no patients were excluded based on self-reported ethnic origin.
Data on the functionality of the six SNPs suggestive of association with severe fluoropyrimidine-induced toxicity is limited. To the best of our knowledge, these six SNPs were not previously identified by other GWAS or other studies, or previously described in relation to the fluoropyrimidine pathway. Genome-wide significant SNPs (rs4702484, rs8101143, rs576523, rs361433) and suggestive SNPs (rs16857540, rs2465403, rs10876844, rs10784749, rs17626122, rs7325568, rs4243761, rs10488226, rs6740660, rs1567482 and rs6706693) identified in previously executed GWAS,22,24,25 were not identified in this GWAS, possibly due to the differences in the design or endpoints of the study. For example, the
312