A significant portion of patients treated for pediatric acute lymphoblastic leukemia (ALL) relapse. We hypothesized that common polymorphisms with moderate effect sizes and large attributive risks could explain an important fraction of ALL relapses. Methylenetetrahydrofolate reductase (MTHFR) is central to folate metabolism and has two common functional polymorphisms (C677T and A1298G). Methotrexate (MTX), which interrupts folate metabolism, is a mainstay of pediatric ALL therapy. MTX inhibits the synthesis of dTMP needed for DNA replication by blocking the conversion of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate by MTHFR. We hypothesized that a deactivating MTHFR allele would increase ALL relapse risk by potentially increasing 5,10-methylenetetrahydrofolate and dTMP, enhancing DNA synthesis and thus opposing MTX. To test this hypothesis, we genotyped 520 patients on the Children's Cancer Study Group ALL study, CCG-1891. The MTHFR C677T variant allele was statistically significantly associated with relapse (χ2 = 4.38, P = 0.036). This association remained significant (hazard ratio = 1.82, P = 0.008), controlling for important covariates, and was more predictive of relapse than other predictors, including day 7 bone marrow response. The MTHFR C677T variant allele was not associated with an increased risk of toxicity or infection. The MTHFR A1298G polymorphism was not associated with altered risks of relapse, toxicity, or infection. Haplotype analysis showed six common haplotypes that did not provide additional information predictive for relapse. These data provide evidence that the MTHFR C677T polymorphism is a common genetic variant conferring a moderate relative risk and a high attributable risk for relapse in pediatric ALL patients.
- Leukemias and lymphomas
- Pediatric cancers
- Genetics of Risk and Outcome
Whereas current treatment protocols cure ∼80% of pediatric patients with acute lymphoblastic leukemia (ALL), a significant proportion of children experience a relapse of their disease (1). Genetic variation in the enzymes responsible for chemotherapy metabolism may play a role in determining relapse and toxicity risks (2, 3) . We hypothesized that a portion of ALL relapses could be explained by genetic polymorphisms that occur with relatively high frequency and have moderate effect sizes but substantial attributive risks.
Methylenetetrahydrofolate reductase (MTHFR) catalyzes the conversion of 5,10-methylenetetrahydrofolate to 5- methyltetrahydrofolate in the folic acid cycle, and is interrupted by methotrexate (MTX), a critical chemotherapy agent in pediatric ALL therapy. The MTHFR C677T and A1298G polymorphisms are nonsynonymous amino acid changes that have been associated with a decreased activity of MTHFR and increased levels of homocysteine (4–6) . The frequencies of the MTHFR C677T and A1298G variant alleles vary by ethnicity: The MTHFR C677T variant allele is present in 34% of Caucasians and 14% of African Americans (7), whereas the A1298G variant allele is present in 27% to 36% of Western Europeans with little allele frequency data in African Americans (8).
Other investigators have reported that MTHFR variants are associated with clinical and disease outcomes, including a decreased risk of developing ALL both in children (9) and adults (10). One prior study has shown an increased risk of ALL relapse in patients with the MTHFR C677T variant allele (11). The MTHFR C677T variant allele has also been associated with increased toxicity with MTX (12–15) . Although informative, all of these studies have utilized relatively small and heterogeneous sample sets and have not concurrently analyzed relapse and toxicity risks. Therefore, we sought to test the impact of MTHFR polymorphisms on clinically relevant relapse and toxicity end points in a large homogenous sample set.
Based on the available functional data and prior studies of ALL susceptibility and MTHFR genotype, we hypothesized that patients with MTHFR variants could have a higher risk of relapse as increased levels of 5,10-methylenetetrahydrofolate might provide additional substrate to the thymidylate synthase–mediated conversion of dUMP to dTMP, thus counteracting the effect of MTX on folate metabolism and DNA synthesis. Based on prior reports, we also hypothesized that patients with MTHFR variant alleles could experience increased toxicity.
This report evaluates the role of the MTHFR C677T and A1298G polymorphisms on relapse and toxicity risk in 520 pediatric ALL patients, all of whom received the same dose and schedule of MTX according to Children's Cancer Group (CCG) protocol CCG-1891. The CCG-1891 trial randomized 1,204 patients to one of three treatment arms, all of which received identical MTX dosing throughout therapy (16).
Materials and Methods
Study Design and Patient Population. We undertook a case-control study of patients treated on CCG-1891, a national intermediate risk ALL study run by CCG from 1989 to 1992 in 137 participating institutions. Intermediate risk was defined as age between 1 and 10 years at diagnosis, and initial WBC count of <5 × 104/μL. The design and results of CCG-1891 have been previously described and typify the Berlin-Frankfurt-Munster approach to ALL therapy (16). Of 1,204 eligible patients, 223 relapsed. All treatment arms were prescribed the identical schedule and dose of age-adjusted I.T. and weekly p.o. 20 mg/m2 MTX during interim maintenance and maintenance therapy. Boys received 38 months of therapy; girls, 26 months.
Sample and Data Collection. Bone marrow and peripheral blood slides were obtained from the Cooperative Human Tissue Network on eligible patients as previously reported (17). At study enrollment, parents or guardians of patients gave informed consent for the use of archived samples in future biological studies. The Institutional Review Boards of the University of Pennsylvania and The Children's Hospital of Philadelphia approved this research.
Data on patient age at diagnosis, gender, ethnicity, course length, toxicity, and infectious complications were obtained from CCG Group Operations Office (Arcadia, CA). These data were reported to the CCG Group Operations Office on case report forms prospectively during the trial by participating centers. Relapse data are reported to CCG on an ongoing basis. The CCG-1891 trial prospectively captured and graded central nervous system toxicity, diarrhea, hyperbilirubinemia, peripheral neuropathy, and transaminitis according to the CCG toxicity criteria. Concurrently, data on the following infectious complications were collected: fever and neutropenia, other infections, interstitial pneumonitis, and sepsis. Data on prescribed MTX dose were not available for analysis.
Laboratory Analysis. DNA was extracted from slides with either a NaOH/heat extraction followed by Qiagen DNA Blood Mini kit (Qiagen, Valencia, CA) clean up or with the Puregene DNA Tissue Extraction kit (Gentra Systems, Minneapolis, MN). All DNA extractions were done in a dedicated DNA extraction area. Gloves and razor blades were changed between each extraction and the work area was wiped down with 70% ethanol between slide extractions. DNA extractions by the Puregene DNA Tissue kit from archived slides were optimized for maximal DNA yields as previously described (18).
MTHFR genotyping was initially done with a PCR-based RFLP assay but subsequently done with a Pyrosequencing assay due to concerns of incomplete digestion with RFLP (19). Genotyping was done by a technician blinded to the outcome status of the sample and negative controls were included in each PCR and Pyrosequencing run. All PCR reactions were set up in a dedicated PCR area with dedicated PCR pipettes and reagents. All RFLP digests and Pyrograms were read by two reviewers blinded to the outcome status of the sample. All genotype calls were rechecked by at least one reviewer before analysis.
MTHFR primer sequences were as follows: For M677, MTHFRForBiotin: 5′-/5BioTEG/AGG CTG ACC TGA AGC ACT TG-3′; MTHFR677Rev: 5′-TCA CAA AGC GGA AGA ATG; MTHFR677SNP1: 5′-TGC GTG ATG ATG AAA TCG-3′. For M1298, M1298-1F-BT: 5′-/5Bio/GGA GCT GCT GAA GAT GTG G-3′; M1298-8R-DS: 5′-TTT GGT TCT CCC GAG AGG TA-3′; M1298-SeqP1: 5′-CAA AGA CTT CAA AGA CAC T-3′ (nucleotide dispensation order available from the corresponding author).
Statistical Analysis. The primary outcome of interest was relapse. The secondary outcomes of interest were development of grade III or IV toxicity and the occurrence of infectious complications. For analysis of relapse risk, cases were defined as patients who experienced leukemia relapse at any site. Controls were defined as patients who remained in continuous remission defined by <5% leukemia blasts on bone marrow aspirate. For analysis of toxicity, cases were patients who experienced toxicity and controls were patients who did not experience toxicity.
All 520 patients who had available slides that yielded DNA adequate for genotyping were included in all analyses. This sample set included 124 patients experiencing relapse and 396 patients in continuous remission. Statistical analyses were done in STATA 8.0 (Stata Corp., College Park, TX) and haplotypes were constructed with Phase 2.0.2 (20, 21) .
Univariate analysis of MTHFR genotype and relapse was done with a χ2 test. Kaplan Meier analysis was used to estimate relapse-free survival (RFS) and the log-rank test was used to compare time to relapse (22). Attributive risk was estimated for unstratified case-control data (23). Cox proportional hazards analysis was used to model the risk of relapse given MTHFR genotype with the clinically important covariates (24). Differences in haplotype frequencies between patients with relapse and those in continuous remission were tested by Phase 2.0.2 (20, 21) .
Analysis of MTHFR genotype and toxicity risk was done as a case-control analysis. Cases were patients who developed the toxicity of interest and controls were patients who did not develop the toxicity. The case-control analysis of toxicity compared the variant allele frequency between cases and controls. Likewise, the analysis of MTHFR genotype and infectious complications compared the variant allele frequency between patients who developed infectious complications and those who did not. Each analysis included all patients with samples adequate for genotyping.
Table 1 presents characteristics of all patients on CCG-1891 and those included in this study. The 520 (43%) patients with samples adequate for genotyping were representative of the entire study with regard to gender, age at diagnosis, ethnicity, initial white blood count, leukemia phenotype, treatment regimen, day 7 bone marrow status, and relapse site.
Table 2 shows that genotype frequencies are consistent with published reports. On univariate analysis, patients with the MTHFR C677T variant allele had a significantly increased relapse risk (χ2 = 4.38, P = 0.036). Kaplan Meier analysis stratified on allele showed a significant decrease in RFS for the variant allele (78.99% versus 69.94, P = 0.0486). RFS by genotype showed that homozygous wild-type patients had a 78.9% RFS, heterozygous patients had a 71.1% RFS, and homozygous variant patients had a 66.6% RFS at 10 years ( Fig. 1 ). Comparison of RFS between homozygous wild type (CC) and variant (TT) patients was significant by the log rank test (P = 0.0477). The attributive risk of relapse among patients with the MTHFR C677T variant allele was 34.7% (95% CI, −0.192-58.4%) and the population attributive risk was 21.3%.
Cox proportional hazards modeling also showed an increased risk to the MTHFR C677T variant allele. The Cox model containing gender, ethnicity, age, treatment arm, initial white blood count, day 7 bone marrow response, leukemia phenotype, and MTHFR C677T variant allele showed an increased hazard ratio (HR) relapse in patients with the MTHFR C677T variant allele compared with the wild-type allele (HR, 1.82; 95% CI, 1.16-2.84; P = 0.008; Table 3 ). Female gender, Caucasian ethnicity, standard treatment (arm A), combined M1 or M2 day 7 bone marrow responses, and B-cell leukemia phenotype were taken as reference groups.
The MTHFR C677T variant allele did not significantly modify the length of chemotherapy courses. The mean duration of a maintenance course was 81.8 days for the wild-type allele and 82.4 days for the variant allele (P = 0.36) and the mean duration of maintenance therapy was 739 days for the wild-type allele and 760.5 days for the variant allele (P = 0.34). Moreover, the MTHFR C677T variant allele did not alter toxicity or infectious complication risk ( Table 4 ).
The MTHFR A1298G variant allele was not associated with a significantly altered risk of relapse, chemotherapy course length, toxicity, or infectious complications.
Haplotype analysis yielded nine observed haplotypes from a total of 16 possible haplotypes. Table 5 shows that two haplotypes (CACG and TACG) accounted for ∼57% of the observed haplotypes and three haplotypes had a frequency of <5% (TGCG, TATG, TGTG). The 95% confidence intervals around the RFS estimates overlapped for all observed haplotypes. The permutation test for a significant difference in haplotype frequencies in patients with relapse and those in continuous remission approached, but did not reach, statistical significance (P = 0.09).
This study shows that the MTHFR C677T variant allele is associated with an increased risk of relapse in standard-risk pediatric ALL patients who received MTX over an extended maintenance course of therapy. Consistent with our hypothesis, the effect size of the MTHFR C677T variant allele is moderate. Notably, the MTHFR C677T variant allele attributive risk is ∼34%, thus indicating that among patients with the variant allele, approximately one third of all relapses are linked to the variant allele (25). Moreover, the population attributive risk of 21% indicates that approximately one of four relapses is related to the MTHFR C677T variant allele (25).
Multivariate analysis shows that MTHFR C677T variant allele is more predictive of relapse than M3 day 7 bone marrow status, initial WBC count, and treatment regimen. From a clinical standpoint, the 12% difference in 10-year relapse-free survival between homozygous wild type and homozygous variant patients is a clinically meaningful difference given the already relatively good outcome of pediatric ALL.
The Cox proportional hazards analysis also shows that non-Caucasian ethnicity was significantly associated with relapse risk, thus raising concern for population stratification. However, several authors have shown that population stratification is unlikely to cause significant bias in studies with multiple admixed populations (26–28) . Furthermore, the impact of the MTHFR T677 variant allele remained significant after controlling for ethnicity, thus indicating an effect of the variant allele that is unlikely to be confounded by ethnicity.
This study has several strengths that support the validity and generalizability of these findings. First, patients were drawn from one national clinical trial of standard-risk ALL patients in which all patients received identical MTX therapy. Second, the sample size of this study permits meaningful evaluation of small to moderate effect sizes. Third, genotyped patients represent patients from the entire study ( Table 1). Finally, all clinical data were collected prospectively and systematically through the CCG data reporting system, thus minimizing the possibility of reporting bias.
In addition, this study utilized haplotype data with inferred allelic phase information in the case-control setting. Phase 2.0.2 enables inference of allelic phase in non-family-based studies; of available haplotype construction algorithms, Phase 2.0.2 seems to give the most robust estimates of true haplotype (20). Whereas informative, the haplotype analysis did not provide additional predictive information on the risk of relapse beyond MTHFR C677T allele data.
This study also has limitations, including the limited number of loci used to construct haplotypes and more modest numbers of patients in each haplotype, the lack of information about actual MTX dose delivered, and the absence of serum homocysteine levels. The MTHFR C677T and A1298G polymorphisms do not capture all the genetic variation in the MTHFR gene. This limitation stems from the candidate polymorphism approach, which was used because of the small amounts of available DNA and very limited haplotype data available at the beginning of the study. Although this study is the largest single trial pediatric oncology pharmacogenetics study, the sample size is still inadequate to evaluate modest differences in the five haplotypes occurring with a >10% frequency. Finally, although simulation studies have shown that haplotype estimation with Phase 2.0.2 occurs with a very low error rate (20), the haplotype estimation error rate in this sample is not known. Thus, this study clearly shows that statistically meaningful evaluations of haplotypes will require a comprehensive knowledge of the candidate gene haplotype map as well as substantial sample sizes.
Because actual MTX dose delivered was not available, we used length of chemotherapy course as a surrogate marker. CCG-1891, like most pediatric ALL protocols, adjusted the dose of MTX and antimetabolites to reach a target absolute neutrophil count. Doses of MTX were held if the absolute neutrophil count dropped below 500, thus patients with substantial myelosuppression may have maintenance course lengths exceeding those of patients without toxicity. No significant difference in course length was observed, thus indicating that the MTHFR C677T variant allele was not associated with substantial myelosuppression. Concordantly, we did not observe that the MTHFR C677T variant allele was a risk factor for any of six prospectively collected treatment toxicities. Furthermore, the MTHFR C677T variant allele was not associated with the occurrence of four infectious complications ( Table 4). Because MTX dose was adjusted on CCG-1891 for toxicity, these results argue that toxicity-based MTX dose modification may not adequately optimize MTX dose.
Krajinovic et al. (11) recently described an increased risk of relapse and leukemia-related death associated with the MTHFR C677T variant allele in a sample set of 201 patients treated on various ALL protocols at Hôpital Sainte-Justine. Our results not only confirm these findings in a larger sample set but also extend them in several important ways. First, our analysis evaluated a single end point, relapse risk, rather than the combined end point of relapse and leukemia-related mortality. Thus, our results allow the estimation of the population attributive relapse risk of the MTHFR C677T polymorphism. In this study, the MTHFR C677T variant allele accounts for ∼20% of all relapses. Furthermore, among patients with the variant allele, approximately one third of relapses are due to the variant allele. Finally, this study combines an analysis of relapse and toxicity risks, thus providing a more comprehensive analysis of the impact of MTHFR polymorphisms on treatment response.
Several other reports have evaluated the MTHFR C677T variant allele and MTX toxicity. Ulrich et al. (12) first described an association between the MTHFR C677T variant genotype and an increased p.o. mucositis index in 220 patients receiving MTX in allogenic stem cell transplantation for chronic myelogenous leukemia. Chiusolo et al. (14) reported in a group of 61 patients that those heterozygous for the MTHFR C677T polymorphism had an increased risk of treatment-related toxicity in four different adult ALL protocols with different doses and schedules of MTX administration. Likewise, Toffoli et al. (15) reported that the MTHFR C677T homozygous variant genotype was associated with increased grade III and IV toxicity in a group of 43 relapsed or refractory ovarian cancer patients treated with low-dose p.o. MTX (2.5 mg) daily for 21 day cycles. Recently, Kishi et al. found no association between the MTHFR C677T variant allele and either seizure or thrombosis risk among 53 pediatric patients receiving high-dose MTX therapy for de novo ALL; other toxicity end points were not reported (29).
Although our report and others described a role for the MTHFR C677T polymorphism in determining MTX response, we did not observe an association between MTX toxicity and the MTHFR C677T variant allele. Several possible explanations exist for these discordant results. First, the MTX toxicity is likely to fit the model of a complex trait. That is, the impact of the MTHFR C677T variant allele may be modified by MTX dose and schedule, concurrent medications, folate status, diet, or other environmental factors. These factors differ substantially between adult patients undergoing myeloablative stem cell transplantation and pediatric patients receiving outpatient therapy for ALL, and may explain the disparity between the results of Ulrich et al. and this study. Second, the combining of several toxicities in the reports by Chiusolo and Toffoli may have led to false associations. Third, the relatively small number of cases in several of our toxicity categories limited the statistical power of our study to detect modest effects of the MTHFR C667T variant on these toxicities. Finally, the dose adjustment of MTX for absolute neutrophil count in CCG-1891 may have minimized the occurrence of other toxicities. Unfortunately, MTX dose data is limited in the CCG-1891 trial and could not be incorporated into this analysis.
In summary, this report shows an association between the MTHFR C667T variant allele and an increased risk of relapse in pediatric patients treated on CCG-1891. The magnitude of this association was consistent with a high prevalence (31.8%), modest effect size (HR, 1.82) with a substantial attributive risk (34.7%). Analysis using the MTHFR C677T allele was more informative than a haplotype-based approach. Additional studies, with adequate statistical power to detect modest effect sizes and careful measurements of prescribed MTX dose, absolute neutrophil count, and toxicity, are needed to confirm and refine these findings. If our outcome and toxicity results are validated in such a study, then MTX dose adjustment by genotype as well as absolute neutrophil count may be appropriate for optimal therapy of pediatric ALL.
Grant support: Doris Duke Charitable Foundation (R. Aplenc) and Leonard and Madilyn Abramson Endowed Chair.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank the Cooperative Human Tissue Network for help in obtaining samples, the Children's Cancer Group for support of the study, and Harland Sather for statistical support.
Note: R. Aplenc had full access to all data and had the final responsibility of submitting the manuscript for consideration.
- Received July 21, 2004.
- Revision received September 22, 2004.
- Accepted November 30, 2004.
- ©2005 American Association for Cancer Research.