Allelic variations in CYP2B6 and CYP2C19 and survival of patients receiving cyclophosphamide prior to myeloablative hematopoietic stem cell transplantation


  • Conflict of interest: The authors have no financial disclosures or conflicts of interest

In hematopoietic stem cell transplantation (HSCT), patients receive cyclophosphamide (CPA) conditioning based on body weight, under the assumption that each individual will metabolize the drug with the same efficiency [1–3]. However, up to 20% of patients experience adverse outcomes related to CPA [4–7]. We hypothesized that due to their effects on CPA metabolism certain allelic variations of cytochrome P450 (CYP) 2B6 and 2C19 enzymes would be associated with higher regimen-related toxicity and worse outcomes in patients. We genotyped 359 patients who received allogeneic HSCT with high dose CPA conditioning. We investigated the effect of allelic variants of CYP2B6 and CYP2C19 on toxicity outcomes, nonrelapse mortality (NRM), relapse, progression free survival (PFS) and overall survival (OS). Overall, 65 (18%) patients experienced toxicity. There was no significant difference in toxicity events in patients with allelic variants of CYP2B6 or CYP2C19. However, compared to poor metabolizers, ultrarapid CYP2B6 metabolizers had worse PFS (P = 0.04). In addition, patients homozygous for CYP2C19 *2 allele had significantly worse PFS (P = 0.008) and OS (P = 0.004). Our results suggest that genotyping may help to predict survival in patients receiving high dose CPA, guide patient management and improve outcomes.

CPA requires biological activation to 4-hydroxycyclophosphamide (4-OH-CPA) and phosphoramide mustard to exert its effects. CYP2B6 and CYP2C19 are two of the primary enzymes involved in producing the active metabolites and determining their concentration and the rate of CPA clearance. At least 25 allelic variants in these enzymes have been found [2, 5, 6, 8–12]. Four CYP2B6 variants, *4, *5, *6, and *7, and two CYP2C19 variants, *2 and *3, carry higher allelic frequencies and have a greater impact on metabolism (Supporting Information, Table III) [13].

Adverse outcomes including veno-occlusive disease of the liver (VOD), cardiotoxicity, idiopathic pneumonia syndrome (IPS), morbidity, and mortality, may result from altered CPA metabolism due to allelic variations in CYP2B6 or CYP2C19 [14, 15]. The *4 allelic variant of CYP2B6 confers an ultrarapid metabolizer phenotype and is associated with increased enzyme activity, increased clearance of CPA, higher concentrations of active metabolites, and consequently increased rates of toxicity and mortality [2, 5, 14, 16–21]. By contrast, patients homozygous for *6 and *7 CYP2B6 alleles are poor metabolizers leading to a decrease in protein activity, slower CPA elimination, lower maximum concentrations of active metabolites and potentially higher rates of relapse and lower rates of toxicity [21–23]. The CYP2C19 *2 and *3 allelic variants prevent protein expression and also confer a poor metabolizer phenotype [10, 11, 24]. Although some studies have shown that allelic variants of CYP2B6 and CYP2C19 are associated with CPA-related toxicity and worse outcomes, the results are inconclusive [4, 5, 7]. We therefore studied the relationship between genotype and outcomes in patients from our institutions who received CPA prior to HSCT between 2002 and 2008.

Patient characteristics are summarized in Supporting Information Table IV. Most patients received matched related or unrelated HSCT using peripheral blood stem cells (PBSC) after a conditioning regimen of CPA/total body irradiation (TBI), (Supporting Information Table IV). Acute myelogenous leukemia (AML), acute lymphocytic leukemia (ALL), and myelodysplastic syndrome/myeloproliferative disorder (MDS/MPD) were the most common diseases treated.

The prevalence of *4, *5, *6, and *7 alleles of CYP2B6 was 6, 15, 41, and 5%; respectively, while the prevalence of CYP2C19*2 and CYP2C19*3 was 26% and <1%; respectively. Both were in Hardy Weinberg equilibrium and the majority of patients with known ethnicity were Caucasian (93%). Individual genotypes were assigned metabolizer phenotypes based on previously published data (Supporting Information Table V) [10, 21, 24, 25].

Overall, 18% of patients experienced toxicity: 12% VOD, 1% cardiotoxicity, and 5% IPS. Two patients experienced both VOD and IPS. There was no significant difference in the proportion of toxicity events when wild-type patients were compared to those with allelic variants of either CYP2B6 (P = 0.89) or CYP2C19 (P = 0.99) (Supporting Information Table V). When classified phenotypically the incidence of toxicity in CYP2B6 ultrarapid, extensive and poor metabolizers was 29, 17, and 22%; respectively (P = 0.30), while the incidence of toxicity in CYP2C19 extensive and poor metabolizers was 18 and 17%; respectively (P = 0.99). Furthermore, in exact logistic regression, the incidence of toxicity did not differ significantly when comparing CYP2B6 ultrarapid to extensive metabolizers, CYP2B6 ultrarapid to poor metabolizers or CYP2C19 extensive to poor metabolizers (Table I). However, as observed earlier [26], patients who received sirolimus for GVHD prophylaxis were significantly more likely to experience toxicity (OR = 1.91; P = 0.03, adjusted OR = 1.92; P = 0.03), specifically VOD (OR = 3.42; P = 0.003, data not shown). (Table I).

Table I. Univariable and Multivariable Exact Logistic Regression Models Including the CYP2B6 and CYP2C19 Phenotypes for all Toxicity Events (VOD, Cardiac and IPS)
OR95% CIP-valueOR95% CIP-value
  • a

    Full multivariable model not shown.

Age, ≤50 vs. >50 yrs1.350.72–2.600.411.390.75–2.590.30
Sex, M vs. F1.180.67–2.120.631.100.63–1.920.75
Conditioning regimen (CPA/TBI vs. others)1.600.53–6.520.551.260.40–3.940.70
Prognosis (poor vs. good)0.930.51–1.710.900.840.47–1.530.58
BMI (<30 vs. ≥30)0.870.47–1.660.740.920.50–1.670.77
GVHD prophylaxis (sirolimus vs. no sirolimus)1.911.07–3.460.031.921.06–3.490.03
Cell type (BM or BM/PBSC vs. PBSC)1.860.79––4.210.12
HLA matching (MUD or mismatched vs. MRD)1.570.87–2.880.141.180.64–2.190.60
CYP2B6 (extensive vs. poor)0.720.26–2.290.650.690.26–1.840.46
(Ultrarapid vs. poor)1.390.31–6.370.861.650.42–6.480.47
CYP2B6 (poor vs. extensive)a1.390.44–3.780.651.460.54–3.920.46
(Ultrarapid vs. extensive)1.940.59–5.610.302.410.84–6.930.10
CYP2C19 (poor vs. extensive)a0.940.47–1.820.990.860.45–1.640.64

The median follow-up time for patients surviving without relapse was 36 months (range, 12–86 months). Among extensive, poor, and ultrarapid CYP2B6 metabolizers there was no significant difference detected in NRM (P = 0.19), PFS (P = 0.22), OS (P = 0.24), or the cumulative incidence of relapse (P = 0.83), (Table II). However, a pairwise comparison between ultrarapid CYP2B6 metabolizers and poor metabolizers revealed that ultrarapid CYP2B6 metabolizers had significantly worse PFS (3 years: 31%) compared to poor metabolizers (70%), (Fig. 1a, Table II and Supporting Information Table VII, P = 0.04; adjusted HR = 1.83, P = 0.13). A similar difference was observed in OS (3 years: 31% for ultrarapid CYP2B6 metabolizers vs. 70% for poor metabolizers) but this difference did not reach statistical significance (Table II, P = 0.08).

Figure 1.

Progression-free survival and overall survival by CYP2B6 phenotype and CYP2C19 genotype. a. The % progression-free survival for poor (n = 27), extensive (n = 311), and ultrarapid (n = 21) CYP2B6 metabolizers is plotted for 86 months post-transplant. Compared to CYP2B6 poor metabolizers, ultrarapid metabolizers had worse progression-free survival (P = 0.04). b. The % progression-free survival for CYP2C19 *1/*1 (n = 249), *1/*2 (n = 82), *1/*3 (n = 1), and *2/*2 (n = 12) is plotted for 86 months posttransplant. Patients with *2/*2 genotype had worse progression-free survival than all other genotypes (P = 0.008). c. The % overall survival for CYP2C19 *1/*1 (n = 249), *1/*2 (n = 82), *1/*3 (n = 1), and *2/*2 (n = 12) is plotted for 86 months post-transplant. Patients with *2/*2 genotype had worse overall survival than all other genotypes (P = 0.004).

Table II. Summary of Outcomes for CYP2B6 Phenotypes
OutcomeCYP2B6P-value (all three)P-value (1) vs. (3)
Poor metabolizer1 estimate (95% CI)Extensive metabolizer2 estimate (95% CI)Ultrarapid metabolizer3 estimate (95% CI)
Cum. Inc. NRM 100 days11 (3, 26)11 (8, 15)14 (3, 33)0.190.08
Cum. Inc. Relapse 2 years19 (7, 35)25 (20, 30)24 (8, 45)0.830.65
PFS; 3 years70 (49, 84)48 (42, 54)31 (12, 52)0.220.04
OS; 3 years70 (48, 84)52 (46, 58)31 (13, 52)0.240.08

No significant differences in survival and relapse were seen when extensive CYP2C19 metabolizers were compared to poor metabolizers (Table III). However, CYP2C19 *2/*2 patients had significantly worse PFS (3 years: 25%) than all other CYP2C19 genotypes (53%) (Fig. 1b, Table III and Supporting Information Table VII, P = 0.008, adjusted HR = 2.38, P = 0.01). OS was also significantly worse in CYP2C19 *2/*2 patients (3 years: 25%) versus all others (55%), (Fig. 1c, Table III and Supporting Information Table VII P = 0.004, adjusted HR = 2.78, P = 0.004). No significant difference in NRM and relapse was seen (P = 0.10 and P = 0.42, respectively).

Table III. Summary of Outcomes for CYP2C19 Phenotypes
Poor metabolizer estimate (95% CI)Extensive metabolizer estimate (95% CI)P-valueGenotype *2/*2All others, i.e., *1/*1, *1/*2, *2/*3P-value; *2/*2 vs. all others
N95249 12331 
Cum. Inc. NRM 100 days14 (8, 21)10 (7, 15)0.9025 (5, 52)11 (8, 14)0.10
Cum. Inc. Relapse 2 years24 (16, 34)24 (19, 29)0.7733 (9, 61)24 (19, 28)0.42
PFS; 3 years49 (38, 59)49 (42, 55)0.7725 (9, 67)53 (42, 65)0.008
OS; 3 years52 (41, 62)53 (46, 59)0.6425 (9, 67)55 (45, 68)0.004

As expected, other factors that were associated with worse NRM, PFS, and OS in multivariable analysis were age >50 (P = 0.01, 0.004, 0.001, respectively) and mismatched vs. matched unrelated donor (P = 0.01, 0.02, 0.01, respectively), (Supporting Information Table VII).

Certain allelic variants of CYP2B6 and CYP2C19 were predictive of OS and PFS. CYP2C19*2 homozygous patients died relatively quickly after transplant. Nine out of 12 patients died (3 of relapse, 2 of graft vs. host disease, 2 of infection and 2 of toxicity), the majority between 1 and 6 months post-transplant. Since these findings are inconsistent with the hypothesis, this suggests the association between poor outcomes and homozygosity for CYP2C19*2 are unrelated to CPA metabolism by CYP2C19. Instead, the worse survival could be related to the administration and subsequent toxic effects of another drug metabolized by CYP2C19. Linkage of the CYP2C19 gene with other relevant genes such as CYP2C9 that impact survival should also be considered. A plausible explanation for the poor PFS in CYP2B6 ultrarapid metabolizers may be the increased toxicity; however, the results did not reach statistical significance. It should be noted that we studied a heterogenous group of individuals in which not all factors could be controlled for in multivariable analysis.

No significant relationships between toxicity, relapse or NRM were seen in patients with allelic variations of CYP2C19 and CYP2B6. The lack of a relationship between phenotype and toxicity suggests that cumulative exposure to the active metabolite contributes more to toxicity than the rate of CPA activation and maximum concentration achieved. The lack of significant differences in relapse in poor metabolizers is most likely due to the fact that patients receive, in addition to CPA, TBI and possibly other medications to ensure complete bone marrow ablation.

The discrepancy between our results and our initial hypothesis as well as with the results of other studies may be due to additional factors. Despite including 359 patients, the number of patients with CYP2B6 ultrarapid metabolizer phenotype and CYP2C19 *2 homozygotes was relatively low. In addition, the large doses (and higher concentrations) of CPA used in myeloablative regimens may have lead to an increased contribution of lower affinity/higher capacity CPA hydroxylase enzymes such as CYP2C9 and/or CYP3A4 to CPA activation [27] which, in turn, may have obscured the effects of genetic variations in CYP2B6 and CYP2C19 [2, 25]. Finally, although most articles support our phenotypic characterization of CYP2B6, the impact of CYP2B6 SNPs on protein expression, enzymatic activity and CPA pharmacokinetics is controversial.

Two important elements of our analysis should also be mentioned as additional possibilities for the discrepancies with other published studies. Medications such as thiotepa or busulfan can inhibit CPA metabolism and influence outcomes. Various concomitant medications used in other studies may explain discrepant findings, particularly if the effects of these drugs were not controlled for statistically as was done in this study [4, 7]. Furthermore, CYP2B6 has a pseudogene, CYP2B7, with a very similar sequence, particularly near the common SNPs. In our analysis we used nested PCR to ensure the pseudogene was not amplified. In addition, we confirmed some of our results with a direct sequencing method and illustrated our allelic frequencies were in Hardy–Weinberg equilibrium. Other studies do not mention the existence of the pseudogene and have not utilized such a robust methodology to avoid coamplification of the pseudogene providing another explanation for differences in results and the unexpected allelic frequencies reported in some articles. Indeed, reported allelic frequencies in some publications differ from those reported by others and seen in our patients and may result from coamplification of this pseudogene.

Genotyping may prove to be a useful adjunct in predicting outcomes in the 30,000–40,000 patients receiving CPA each year. If the presence of an allelic variant is predictive of adverse outcomes, genotyping patients prior to initiation of therapy may allow clinicians to individualize drug dosages, utilize continuous monitoring of drug levels during treatment or change the drug regimen entirely. Larger studies are necessary to elucidate the mechanisms behind the poor survival in ultrarapid CYP2B6 metabolizers and CYP2C19*2 homozygotes, as well as to determine if the trends observed for toxicity would reach statistical significance. Efforts are underway to perform a multicenter prospective study that will increase the number of patients with the less common but most predictive allelic variants, incorporate analysis of other metabolic enzymes and allow pharmacokinetic analysis and correlation with drug concentrations.


Patient specimens.

A total of 359 pretransplant DNA samples from patients who received high dose cyclophosphamide-based (3.6 g m−2 or 120 mg kg−1) myeloablative conditioning regimens for allogeneic HSCT between 2002 and 2008 were genotyped for polymorphisms in CYP2B6 and CYP2C19. All patients were genotyped for allelic variants of CYP2B6. Of the 359 samples, 344 had sufficient DNA remaining for CYP2C19 genotyping.

Genotyping results were correlated with patient data retrieved from the Dana-Farber Cancer Institute Bone Marrow Transplant data repository. This study was approved by the Institutional Review Board of the Dana-Farber/Harvard Cancer Center, and all patients had previously provided consent to have blood samples banked for research.


Genomic DNA was extracted from blood using the Qiagen EZ1 Kit and Qiagen BioRobot EZ1 workstation (Qiagen, Valencia, CA). Assays for CYP2B6 were designed to detect three SNPs, 516A>G, 785A>G, and 1459C>T, associated with four allelic variants: *4 (785A>G), *5 (1459C>T), *6 (785A>G, 516G>T), and *7 (785A>G, 516G>T, 1459C>T) and assays for CYP2C19 were designed to detect two SNPs, 681G>A and 636G>A, associated with two allelic variants: *2(681G>A) and *3(636G>A).

Polymerase chain reaction amplification.

PCR primers for CYP2B6 (Addendum, Table VII) were designed using Biotage primer design software (Biotage, Uppsala, Sweden) in conjunction with published sequences [21, 22]. PCR primers for CYP2C19 were obtained from the literature [28]. For 516A>G and 1459C>T, primers that differentiated the functional CYP2B6 gene and the nonfunctional pseudogene (CYP2B7) were designed from the intron regions that generated PCR products appropriate for pyrosequencing. For 785A>G, the primers that were CYP2B6-specific generated a PCR product that was too long for direct pyrosequencing; therefore, a nested reaction was used to generate a template for pyrosequencing. All PCR amplifications were performed with 1.5 mM MgCl2 in a total volume of 50 μL and 45 cycles with denaturation at 95°C for 30 sec, annealing at 59°C (60°C for nested 785A>G reaction) for 30 sec, and extension at 72°C for 45 sec.


Pyrosequencing was used for the detection of CYP2B6 and CYP2C19 SNPs in PCR products. The pyrosequencing primers for CYP2B6 (Addendum, Table VII) were designed using Biotage software, while CYP2C19 primers were obtained from Eriksson et al. [28]. Sequencing was performed using a Biotage PyroMark MD instrument and the reagents provided. For CYP2C19 the pyrosequencing protocol outlined by Eriksson et al. was followed [28] with minor modifications. To verify the accuracy of CYP2B6 genotyping and to exclude the presence of the CYP2B7 pseudogene, the sequences of eight specimens were verified using direct (Sanger dideoxynucleotide) DNA sequencing of the PCR products [21]. The CYP2C19 sequences were run in duplicate to verify the accuracy of SNP assignment.

Clinical outcomes.

The National Cancer Institute criteria were utilized to define VOD, cardiotoxicity and IPS. Patient history and occurrence of toxicity from 0 to 100 days after transplantation was confirmed by chart review. CPA-related cardiotoxicity was defined as myopericarditis ≥ Grade 3. Nonrelapse mortality (NRM), relapse, progression-free survival (PFS), and overall survival (OS) outcomes were also recorded.

Phenotype classification.

The effect of both genotype and phenotype (predicted from genotype) on clinical outcomes was assessed. Based on the literature CYP2B6 *6/*6 and *6/*7 were classified as poor metabolizers, CYP2B6 *1/*1 (i.e., wild type), *1/*5, *1/*6, *1/*7, and *5/*5 were classified as extensive metabolizers and CYP2B6 *1/*4, *4/*5, and *4/*6 were classified as ultrarapid metabolizers [21, 24, 25]. CYP2C19 *1/*2, *1/*3, and *2/*2 were classified as poor metabolizers and CYP2C19 *1/*1 (i.e., wild type) were classified as extensive metabolizers [10]. CYP2C19 *2 homozygotes that carried two null mutations were also treated as a separate group.


Fisher's exact test was performed for categorical comparisons. Exact logistic regression models were constructed and adjusted to test for toxicity differences between CYP phenotypes. Multivariable analysis was adjusted for sex, conditioning regimen, disease, body mass index (BMI) and graft vs. host disease (GVHD) prophylaxis. The Kaplan-Meier method was used to generate OS and PFS curves and groups were compared using the log-rank test. OS was defined as the time from the date of HSCT to the date of death from any cause censored at the date last known alive. PFS was defined as the time from the date of HSCT to the date of relapse/progression or death, censored at the date last known alive and relapse/progression free. Cox regression models were constructed and adjusted for baseline patient characteristics. Cumulative incidence curves were constructed for NRM and relapse and measured from the time of HSCT, treating NRM and relapse as a competing risk to each other and were compared using Gray's test [29]. NRM was defined as any death due to any post-transplant event or complication other than underlying disease. Competing risk regression models were constructed and adjusted for baseline patient characteristics [30].


The authors thank Dimity Zepf, Jesse Ladner, and Chloe Andac of the Center for Advanced Molecular Diagnostics, BWH, and members of the Ritz lab, DFCI, for their expert technical assistance. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Stacy E.F. Melanson*, Kristen Stevenson†, Haesook Kim†, Joseph H. Antin†, Michael H. Court‡, Vincent T. Ho†, Jerome Ritz†, Robert J. Soiffer†, Frank C. Kuo*, Janina A. Longtine*, Petr Jarolim*, * Department of Pathology, Brigham and Women's Hospital, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, † Division of Hematologic Malignancies, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, ‡ Department of Pharmacology and Experimental Therapeutics, Tufts University School of Medicine, Boston, Massachusetts.