Impact of cytidine deaminase polymorphisms on toxicity after gemcitabine: the question in still ongoing

We have read with interest the article by Ciccolini et al reporting data from an observational study where they retrospectively examined adverse events from three cohorts of patients treated with gemcitabine, alone or combined with other drugs, and evaluated the activity of cytidine deaminase (CDA) in serum. For that purpose, they used a test that was validated in a mouse model, showing that CDA deficiency increased gemcitabine plasma levels, with subsequent lethal toxicities. In addition, they performed genotyping analyses for three polymorphisms of the CDA gene. They demonstrated that patients with toxicities had a significantly lower CDA activity than patients without toxicities, even as no genotype-phenotype relationships were found, and concluded that determination of CDA activity is the most reliable test to identify patients at risk of developing severe toxicities with gemcitabine. The critical role of CDA in gemcitabine activity/toxicity has been demonstrated by several preclinical and clinical studies, showing that upregulation of CDA played a role in gemcitabine resistance, while impaired activity resulted in increased toxicities. Furthermore, three single nucleotide polymorphisms in the coding region of CDA (CDA 79A>C, 208G>A, and CDA 435C>T), have been correlated with CDA activity and toxicity/outcome in gemcitabine-treated patients. Although we agree with Ciccolini et al that their phenotypic test is simple and cost-effective, one can argue whether this test is really representative for the CDA activity in the specific catabolism of gemcitabine, given that it uses cytidine as substrate and measures ammonium to detect the activity. There is a difference in substrate specificity between cytidine and the deoxynucleosides deoxycytidine and gemcitabine,6 while ammonium can have other sources. Furthermore, it is not clear when samples are taken (out of the gemcitabine infusion), and therapy (eg, gemcitabine, but also capecitabine) may influence activity. We also believe that a DNA sample is easier to collect (eg, using buccal swabs) and store than a serum sample. Similarly, the results from a genotyping analysis can easily be used to divide patients into categories corresponding to the different genotypes. In contrast, the continuous data obtained from an enzymatic activity assay need a more accurate definition of the threshold limit to categorize patients with high and low activity. Complex statistical tools, such as receiver operating characteristic curves, should be used to determine whether this test could discriminate adequately between patients with different risks of developing severe toxicities. Therefore, assessing germline polymorphisms as predictive markers of toxicity is appealing, and single nucleotide polymorphism analysis of the drug metabolizing enzymes thiopurine S-methyltransferase and uridine diphosphate glucuronyltranferase has already been included in US Food and Drug Administration–approved tests to predict toxicity of 6-mercaptopurine and irinotecan, respectively.