SNPs and Haplotypes in DPYD and Outcome of Capecitabine–Letter

In their recently published article in Clinical Cancer Research, Deenen and colleagues reported a number of single-nucleotide polymorphisms (SNP) and haplotypes in DPYD that were associated with toxicity of capecitabine (1). In particular, the silent c.1236G>A mutation was strongly associated with toxicity and the authors claim that the functional effect of this mutation is unknown. However, last year, it was shown that the c.1236G>A mutation is in cis with a deep intronic mutation c.1129-5923C>G, creating a cryptic splice-donor site and affecting pre-mRNA splicing (2). In addition, the c.1129-5923C>G splice mutation proved to be in complete linkage with the previously identified B3 haplotype, which was significantly enriched in patients with severe 5-fluorouracil toxicity (2, 3). A peculiar finding, therefore, was the observation by Deenen and colleagues that no dose adaptation was required in carriers for this mutation (1).

Thus, intronic mutations can have a significant functional impact and a haplotype-based analysis might lead to the detection of important haplotypes without the need for analyzing complete intronic sequences (3). Unfortunately, we feel that the haplotype analyses carried out in a relatively large genomic region with low linkage disequilibrium (LD) by Deenen and colleagues should be interpreted with caution. Accurate haplotype inference is difficult in samples of unrelated individuals and all available methods, including the one applied by Deenen and colleagues (4), only carry out well in genomic regions with high LD (5). In this respect, the 6 SNPs included in their haplotype inferences are distributed across several haplotype blocks (3), that is, areas of high LD, for which haplotype inference should have been carried out separately. Our concerns are, for example, illustrated by the observation that the arbitrary haplotype pair groupings HP5 and HP6, which the authors report to be associated with grade 3–4 diarrhea, represent an enrichment of carriers of the c.496G, c.1236A, and c.2194A variants, which they also reported to be significantly associated with grade 3–4 diarrhea. On the basis of the estimated haplotype frequencies, approximately 74% of the HP5 genotypes should be carriers of these variants whereas only 23% of the non-HP5 genotypes are expected to be carriers. In the HP6 group, the expected carrier frequency is 53% and 20% in the HP6 and non-HP6 genotypes, respectively. Furthermore, it seems that these haplotype pair groupings do not just simply mirror an effect of risk variants but also have a distorting effect on the analyses. For example, based on the estimated haplotype frequencies and the number of observed HP5 or HP6 genotypes in different patient groups, 27% of all risk variant carriers (considering only c.496G, c.1236A, and c.2194A variants) are expected to be observed in the patient group experiencing severe toxicity, which is significantly lower than the observed data (38%). This implies that identical haplotype pair groups show a different haplotype composition between the 2 patient groups. Thus, it seems unlikely that one can draw any unbiased conclusions on the effect of particular SNPs or haplotypes based on these haplotype pair groupings.