Letter regarding Zhao et al. entitled “ DPYD gene polymorphisms are associated with risk and chemotherapy prognosis in pediatric patients with acute lymphoblastic leukemia”

Methodological shortcomings

In our view, the article by Zhao et al. has multiple methodological shortcomings. One shortcoming is that cases and controls were matched solely by age, and not by sex and race, which is also associated with ALL. ⁸ In addition, hospital-based controls were used instead of population-based controls which would be more appropriate, and would have allowed more controls per case to be included. Furthermore, the preferred statistical analysis would be conditional logistic regression, and not chi-square tests.

A major shortcoming is that in the association analysis between DPYD polymorphisms and treatment outcome, no adjustment for the strongest general prognostic factors in pediatric ALL, that is, immunophenotype, age, sex, race, initial white blood cell count, and genomic/cytogenetic features, was made. ⁹ An imbalance in these prognostic factors could have affected the results. In our view, even if the association between DPYD c.85T>C and treatment outcome would exist, it is unlikely that its effect size would outweigh that of the known prognostic factors, strongly questioning the clinical value of DPYD c.85T>C as a prognostic and/or predictive factor as claimed by the authors. In view of these methodological issues, we question the internal validity of the results and ask the authors to report baseline prognostic factors according to c.85T>C genotype (TT, TC, and CC) and report a multivariable analysis.

In addition, it is noted that the observed association between c.85T>C and risk of ALL seems to result solely from a higher frequency of the homozygous c.85T>C CC genotype among ALL cases (7.5% vs 1.0% in controls), while frequencies of TT and CT genotypes were comparable between groups. Similarly, only the 11 patients carrying the CC genotype had a lower complete response (CR) rate (64%), while patients carrying TT or CT had a CR rate similar to the population (88%). Based on these findings, it would be logical to analyze event-free survival (EFS) of CC versus TT + CT genotypes. Yet, the authors compared TT with CT + CC genotypes, without further justification. We would be interested to see EFS for each individual genotype (TT, TC, and CC).

Furthermore, it is noted that the range of follow-up is reported to be 0.5–36 months. Surprisingly, however, EFS data upto 60 months are provided. Moreover, the 36-month EFS times presented in Table 6 are not the same as depicted in Figure 2(a)–(d). Also, there is a discordance between the reported frequencies of the c.85T>C CC genotype in the text and Table 3 (1 vs 11 homozygous CC carriers among the cases, respectively).

Finally, the authors did not adjust p values for multiple testing. If correction would have been applied, the results might not have remained significant.

Selected treatment regimen

Zhao et al. reported that patients were treated with surgery and FOLFOX. First, it is unclear that what type of surgery patients underwent for their hematological disease. Second, FOLFOX is not a standard treatment for pediatric ALL according to Asian, US, or European clinical practice guidelines.^9,10 Moreover, we are not aware of any data supporting efficacy of FOLFOX in ALL. Standard treatment agents of pediatric ALL include glucocorticoids, vincristine, asparaginase, doxorubicin, cytarabine, and cyclophosphamide. ⁹

Since FOLFOX was used by Zhao et al. the external validity (extrapolability) can be strongly questioned in view of current clinical practice.

Toxicity associations

Besides the alleged association between DPYD c.85T>C and susceptibility to ALL and effectiveness of FOLFOX, the authors also conclude that DPYD c.85T>C is associated with toxicity.

Common 5-fluorouracil (5-FU)-associated toxicities are hematological toxicity, mucositits, stomatitis, diarrhea, and hand-foot syndrome and are typically associated with DPD-deficiency.^11,12 Nonetheless, no differences in the incidences of these toxicities according to DPYD genotypes were found. The authors did, however, find associations between DPYD c.85T>C and both “renal function damage” and “liver function damage” (not further specified). These terms appear to refer to reduced creatinine clearance and elevated liver function tests (e.g. aspartate transaminase/alanine transaminase (AST/ALT)) according to the reference provided for toxicity scoring (which, besides, is not standard, as the common terminology criteria for adverse events [CTC-AE] criteria are typically used for scoring toxicity in cancer).^13,14Reduced renal function and elevated liver function tests are unlikely results of 5-FU treatment.^11,12 Importantly, these adverse events could be related to the disease state of the respective patients rather than to treatment, since renal failure and elevated liver function tests can be complications of ALL and/or concomitant pharmacotherapy. ¹⁵ Therefore, it could be hypothesized that the 11 patients carrying the DPYD c.85T>C CC genotype presented more often with renal and liver-function related adverse events as a result of poorer disease state, which could in turn explain their inferior CR rate and EFS. This makes an association between DPYD c.85T>C and treatment-related toxicity highly unlikely.

Conclusion

In summary, we have major concerns regarding the internal and external validity of the findings by Zhao et al., and in view of the described methodological shortcomings, question whether DPYD c.85T>C is associated with risk of ALL and treatment outcome. We feel that the results from this study should be interpreted with great caution.