- © 2004 by American Society of Clinical Oncology
Pharmacogenetic Risk Factors for Osteonecrosis of the Hip Among Children With Leukemia
- Mary V. Relling,
- Wenjian Yang,
- Soma Das,
- Edwin H. Cook,
- Gary L. Rosner,
- Michael Neel,
- Scott Howard,
- Raul Ribeiro,
- John T. Sandlund,
- Ching-Hon Pui and
- Sue C. Kaste
- From the Department of Pharmaceutical Sciences, Hematology-Oncology, Radiological Sciences, St Jude Children's Research Hospital; University of Tennessee, Colleges of Medicine and Pharmacy, Memphis, TN; Department of Human Genetics, University of Chicago, Chicago, IL; Department of Biostatistics, M.D. Anderson Cancer Center, Houston, TX.
- Address reprint requests to Mary V. Relling, PharmD, Department of Pharmaceutical Sciences, St Jude Children's Research Hospital, 332 N Lauderdale, Memphis, TN 38105-2794; e-mail: mary.relling{at}stjude.org
Abstract
Purpose One of the adverse effects of therapy for acute lymphoblastic leukemia (ALL) is osteonecrosis of the hip. Putative risk factors for osteonecrosis have included being female, white race, and older age. Our goal was to define possible genetic risk factors for osteonecrosis among children treated for newly diagnosed ALL.
Methods Using a candidate gene approach, we determined the genotypes for 16 common polymorphisms in genes likely to affect the pharmacokinetics or pharmacodynamics of antileukemic medications in 64 children with ALL. Therapy included glucocorticoids and antifolates. Magnetic resonance imaging of both hips was used to diagnose osteonecrosis, and was performed at similar times from the start of ALL therapy (P = .61) in the 25 patients with and the 39 patients without osteonecrosis (median, 447 days and 443 days, respectively).
Results In addition to age older than 10 years (odds ratio [OR], 24.2; P = .0001) and white race (OR, 11.1; P = .037), host factors for osteonecrosis included the vitamin D receptor FokI start site CC genotype (OR, 4.5; P = .045), and the thymidylate synthase low activity 2/2 enhancer repeat genotype (OR, 7.4; P = .049).
Conclusion Because folate-related and vitamin D–receptor genetic variants have been associated with bone and vasculature morbidity, these pharmacogenetic associations likely reflect the interaction of antileukemic medications with germline sensitivity to drug actions, and might identify ALL patients at highest risk to develop osteonecrosis.
INTRODUCTION
Glucocorticoids have played an increasingly important role in the therapy of childhood acute lymphoblastic leukemia (ALL). Treatment of ALL consists of daily or weekly therapy for 2.5 to 3.0 years. Intensive use of glucocorticoids (prednisone or dexamethasone) given orally two to three times per day during 4 to 6 weeks of remission induction and for 5 to 7 days per month as pulse therapy has been credited with improved cure rates in several trials.1,2
Osteonecrosis, a debilitating complication of treatment with glucocorticoids, can occur in 10% to 15% of ALL patients, especially those older than 10 years.3,4 Osteonecrosis of the hip can lead to collapse of the proximal femoral epiphysis and destruction of the articular surface of the hip, with subsequent pain and debilitating arthritis.5 Treatment is based on severity and ranges from rest and relief from bearing weight to surgical procedures for core decompression or total hip replacement.6,7
Risk factors for osteonecrosis remain poorly defined. Although osteonecrosis has clearly been linked to the use of glucocorticoids in ALL and in other settings,3,4,8 some ALL regimens include relatively high doses of glucocorticoids but have a low incidence of osteonecrosis. Thus, other elements of therapy may interact with glucocorticoids to contribute to the risk of osteonecrosis. Although several host-related risk factors for osteonecrosis have been identified among children with leukemia (being female, older age, and white race),3,4,8-10 genetic risk factors have not been studied. Genetic polymorphisms in the vitamin D receptor have been linked to the risk of osteopenia, but have never been studied as risk factors for osteonecrosis.
We performed this exploratory analysis to assess whether common, functional polymorphisms in target genes are associated with the risk of osteonecrosis in children treated for ALL who had been screened for this complication by magnetic resonance imaging of the hip.
METHODS
Patients
All patients were treated on one of two consecutive St Jude Children's Research Hospital treatment protocols for newly diagnosed childhood ALL: Total XIIIB or XIV, both of which were approved by the local institutional review board. Informed consent was obtained from the parent, guardian, or patient (as appropriate). Both treatment regimens included prednisone as part of induction treatment, dexamethasone as part of continuation therapy, and extensive use of the antifolate methotrexate; the details of therapy have been published previously.11,12 A total of 64 children on Total XIIIB or XIV (of a total of 299 children enrolled onto the two studies) had bilateral hip magnetic resonance imaging (MRI) examinations performed and interpreted as of November 2001 and are included herein. The occurrence of osteonecrosis was probably over-represented in this group, as patients were more likely to have had an MRI of the hips performed if osteonecrosis was suspected by the treating physician; a subset of children (those > 10 years of age on Total XIV; n = 12) had an MRI of the hips performed after reinduction, regardless of the presence or absence of symptoms. The radiologist (S.C.K.) was blinded to the presence or nature of symptomatology at the time of image reading. There was no other anatomic site that was simultaneously assessed in all patients. Because it is possible that children who did not have an MRI of the hips performed may have had undiagnosed osteonecrosis, such nonimaged patients were not considered acceptable negative controls. Genomic DNA and patient consent for pharmacogenetic testing was obtained as part of the primary ALL treatment protocol.
Phenotyping
MRI examinations of both hips were performed using a Siemens 1.5 Tesla magnet (Siemens Corp, Erlangen, Germany). Noncontrast coronal T1- and short τ inversion recovery (STIR) images of the hips were obtained in all patients. Additional sequences were obtained in selected patients to clarify a finding on the coronal images, and included sagittal and/or axial gradient echo sequences. All MRI studies were reviewed by a single pediatric radiologist (S.C.K.) experienced in pediatric orthopedic and oncologic imaging. Changes consistent with osteonecrosis of the capital femoral epiphysis comprised geographic regions of decreased signal as seen on T1-weighted images that also demonstrated increased signal intensity on STIR sequences.13 Osteonecrosis was coded as being present or absent. All clinical toxicity was assessed and graded prospectively according to the National Cancer Institute Common Toxicity Criteria version 2 (http://ctep.cancer.gov/reporting/CTC-3.html). We subsequently confirmed that all symptomatic cases of osteonecrosis patients on Total XIIIB and XIV protocols were confirmed with MRI findings.
Genotyping
DNA was extracted from normal blood cells. Genotyping was performed for 16 polymorphic loci in genes likely to affect the pharmacokinetics or pharmacodynamics of the anticancer drugs given to these children, using the indicated methods. CYP3A4*1B,14 CYP3A5*3,14 TPMT inactivating mutations,15 the UGT1A1 promoter repeat polymorphism,16 the thymidylate synthase (TYMS) enhancer repeat,17 the GSTT1 deletion,18 the GSTM1 deletion polymorphism,19 RFC 80G > A and MTHFR 677C > T,12 the glucocorticoid receptor NR3C1 1220A > G20 polymorphism, and the MDR1 exon 21 2677G > T/A and MDR1 exon 26 3435C > T, VDR intron 8 G > A, VDR start site Fok I, and GSTP1 313A > G polymorphisms were genotyped as described.21 The MTHFR 1298A > C polymorphism22 was genotyped by single base extension (SBE) and separation by denaturing high performance liquid chromatography (DHPLC).23 Polymerase chain reaction and extension primers were: 5′-TGGGGAGCTGAAGGACTACTA-3′ and 5-CACTTTGTGACCATTCCGGTT-3′and 5′-GGAGGAGCTGACCAGTGAAG-3′ (extension). For the SBE reactions, polymerase chain reaction amplified products were treated with shrimp alkaline phosphatase and exonuclease I by incubating at 37°C for 45 minutes; 1 μmol/L of extension primer, 250 μmol/L each ddNTP, and 1.25 U thermosequenase (Amersham Pharmacia Biotech, Uppsala, Sweden) was added to a total 10 μL volume, and reactions were cycled at 96°C for 30 seconds, 55°C for 30 seconds, and 60°C for 30 seconds for 60 cycles. Separation of the SBE products was performed on a WAVE 3500HT DHPLC system (Transgenomic Inc, Omaha, NE) at 70°C after denaturation of the samples.
For all the polymorphisms genotyped, genotype distributions were found to be consistent with Hardy-Weinberg equilibrium within racial groups.
Statistical Analysis
This analysis to identify possible genetic predictors for osteonecrosis was exploratory. The primary end point of osteonecrosis status was treated as a dichotomized variable (osteonecrosis present v absent). Most of the 16 loci had two possible alleles; MDR1 exon21 and the UGT1A1 promoter TA repeat both had three alleles. The rare third alleles (A for MDR1 exon 21, n = 2; (TA)5 for UGT1A1, n = 1) were excluded in univariate analyses due to the small number of patients and the fact that the functional consequences of these rare alleles are not clear.
After adjusting for race, genotype frequencies were compared using the Mantel-Haenzel test in the current cohort of 64 children versus the remainder of the entire group of children with ALL (n = 235) enrolled on the primary treatment protocols (Total XIIIB and XIV) who did not have hip MRIs available.
Univariate analysis was carried out using logistic regression for each individual genotype as a predictor of osteonecrosis. For loci with heterozygous and both homozygous genotypes, the genotypes were treated as ordered categoric variables and tested for a linear trend associated with osteonecrosis (ie, it was assumed that heterosis was unlikely). We also tested the interaction between race and each of the genotypes by logistic regression.
For multivariate analysis, we applied recursive partitioning24 to explore the interaction between multiple predictive variables and to evaluate the risk of osteonecrosis associated with specific subgroups. All variables, including demographic factors, treatment arm, and genotypes, were considered in building the classification tree.24 Genotypes were treated as ordered categorical variables as described above. At each step, a cutoff for a predictive variable was chosen to split patients into two subgroups that optimally increased the odds ratio for the osteonecrosis phenotype between subgroups. This process was repeated for each subgroup until no variables were found to further improve the risk classification, or the number of patients within each subgroup became too small (n ≤ 15) to be further divided. Guided by the results of the recursive partitioning analysis, we applied multiple logistic regression25 to obtain the estimates of odds ratios and their CIs, and we repeated the logistic regression after adjusting for the previously reported predictors of race and sex. P values were computed based on the likelihood ratio test.
RESULTS
A total of 64 children with ALL were evaluated by MRI of the hips (Table 1). MRI evaluations of both hips were performed at similar time points in therapy (P = .77) for the 25 patients with, and the 39 patients without osteonecrosis, at a median of 1.23 years (range, 0.46 to 4.90 years) and 1.23 years (range, 0.13 to 6.4 years) from start of ALL therapy, respectively. Approximately half of the patients with osteonecrosis had grade 1 (n = 8) or grade 2 (n = 5) osteonecrosis, and half had grade 3 (n = 9) or 4 (n = 3) osteonecrosis. Only two of the 25 patients who were, versus three of the 39 children who were not diagnosed with osteonecrosis were also diagnosed with venous thrombosis while on therapy, frequencies that did not differ (Fisher's exact, P = .99).
Univariate Analysis of Variables in Cases With Osteonecrosis Versus Controls
Univariate analysis (Table 1) indicated that patients who developed osteonecrosis were older at the time of start of ALL therapy (median, 13.5 years; range, 7.2 to 18.8 years) than those who had not (median, 8.6 years; range, 2.7 to 16.9 years) (P = .001). In this analysis, sex was not significantly different between those who did versus did not develop osteonecrosis (Table 1). Other possible predictors of osteonecrosis (Table 1) included white race (P = .105), thymidylate synthase (TYMS) enhancer 2-repeat allele (P = .027), and the vitamin D receptor (VDR) Fok I start site C allele (P = .058). There were no significant interactions between race and any genotypes under investigation.
Recursive partitioning analysis (Fig 1) indicated that age was the most significant predictor of osteonecrosis. Age was actually included in the model as a continuous variable, and the model determined that the best cutoff is 10 years of age, which is in accordance with prior studies.3,4 For children older than 10 years, all seven patients with the TYMS 2/2 genotype developed osteonecrosis compared with 17 (53%) of 32 with the higher activity genotypes. Among those with the lower risk thymidylate synthase genotypes, the VDR Fok I CC genotype was more common in those with osteonecrosis.
Classification tree (A), indicating the variables found by recursive partitioning analysis (see Methods) to be associated with risk of osteonecrosis (AVN). The percent of patients observed to have osteonecrosis in each of the terminal nodes of the tree is depicted in the inset (B). Pts, patients; MRI, magnetic resonance imaging.
Multiple logistic regression analysis using age, TYMS genotypes, and VDR genotypes further confirmed the significance of the selected predictors. We obtained odds ratios of 15.6 (P = .0002, 95% CI, 3.6 to 68.6) for age older than 10 years versus age ≤ 10 years, 7.2 (P = .044; 95% CI, 1.05 to 48.9) for TYMS 2/2 genotype versus other TYMS genotypes, and 3.7 (P = .056; 95% CI, 0.97 to 14.1) for VDR Fok I CC genotype versus CT and TT combined. Because race and sex have been identified as risk factors in other studies,3,4 we repeated the logistic regression including race and sex, and found that predictors were similar for age, TYMS genotype, and VDR Fok I CC genotype, with a higher risk for whites versus other races (Table 2). There were no significant differences in the associations of genotypes based on severity of osteonecrosis (grade 1 or 2 v grade 3 or 4; P > .098, Fisher's exact test). Overall, adjusting for race, genotype frequencies for TYMS and VDR FokI polymorphisms did not differ in this subset of 64 children compared with the remainder of the entire group of children with ALL (n = 235) enrolled on the primary treatment protocols (P = .82 and .80, respectively).
Multiple Logistic Regression for Predictors of Osteonecrosis
In this selected patient group, the four risk factors of age greater than 10 years, white race, the TYMS 2/2 genotype, and the VDR Fok I CC genotype together have a sensitivity for predicting the development of osteonecrosis of 96% (24 of 25 patients who developed osteonecrosis had these high risk characteristics) and a specificity of 82% (32 of 39 patients who did not develop osteonecrosis were predicted not to be at high risk). Of the 12 patients who were screened with MRI of the hips regardless of symptoms, seven developed osteonecrosis. Among these, three (43%) of seven patients with osteonecrosis, and none (0%) of five without osteonecrosis carried the low-activity TYMS 2/2 genotype; three (43%) of seven patients who did develop osteonecrosis versus none of 5 (0%) who did not develop osteonecrosis carried the vitamin D receptor Fok I start site CC genotype, consistent with the findings in the entire cohort of 64 patients.
DISCUSSION
In this exploratory analysis of patients who received comparable therapy with glucocorticoids and were screened with MRI of the hips, two genetic polymorphisms were associated with the risk of osteonecrosis. Interestingly, the only polymorphisms associated with osteonecrosis were in genes (TYMS and VDR) whose products affect the pharmacodynamics and not directly the pharmacokinetics of antileukemic medications. The strongest association was with the enhancer repeat polymorphism in TYMS. The TYMS 2/2 genotype is associated with low TYMS expression, rendering cells more susceptible to toxic and anticancer effects of TYMS inhibitors,26-28 which include methotrexate. We previously reported that, among children with ALL who all received identical doses of glucocorticoids, bone mineral density was lower in patients who received a higher proportion of their ALL therapy as “antimetabolites” (methotrexate plus mercaptopurine).29 Thus, this exploratory pharmacogenetic analysis may shed light on additional therapy-related risk factors for osteonecrosis (ie, the finding that patients who are expected to express less TYMS are more susceptible to osteonecrosis suggests that antifolate therapy may contribute to the risk of osteonecrosis). Homocysteine is integral to folate and thymidine homeostasis, and hyperhomocysteinemia is often associated with low folate and has been linked to thrombosis and vascular pathology30 and with the risk of osteonecrosis.31 Although in this group we did not measure homocysteine levels, nor did we observe a relationship between the occurrence of clinically evident thrombosis and osteonecrosis, subclinical thrombotic events cannot be excluded. Thus, it is possible that one or more defects in the folate/pyrimidine pathway interacted with the antifolate and glucocorticoid therapy to predispose to osteonecrosis, and more profoundly affected patients with low TYMS activity.
The only other polymorphism associated with risk of osteonecrosis was the vitamin D receptor. We studied VDR polymorphisms because of their prior association with osteopenia in noncancer settings, although the mechanisms underlying the genesis of osteopenia and osteonecrosis are generally regarded as unrelated.32 In addition, as a regulator of cytochrome P450 3A expression,33,34 VDR might play a role in the CYP3A metabolism and p-glycoprotein35 excretion of glucocorticoids. The functional consequences of the VDR start site polymorphism are controversial.36,37 The C allele changes the position of the 5′ ATG codon and thereby results in a protein that is truncated by two amino acids compared to the protein translated from the T allele.38 The truncated protein from the C-allele has been linked to a higher affinity for 1,25-dihydroxyvitamin D and to more potent interactions with the transcription factor IID than with the nontruncated T-allele receptor protein.36,37 However, some clinical epidemiologic studies have found more bone pathology in patients who are homozygous for the T than for the C allele.39,40 This study represents the first exploratory analysis of VDR polymorphisms in the risk of osteonecrosis. Whether the C allele confers greater sensitivity to the toxic effects of steroids directly, or by affecting the responsiveness of downstream targets of the vitamin D receptor to its ligands (which in turn regulate CYP3A expression33,34 and could thus affect pharmacokinetics of glucocorticoids) is not clear. The complex interdependent regulation of CYP3A, p-glycoprotein, and vitamin D receptor, along with the effects of p-glycoprotein and CYP3A on extent of drug absorption and metabolism, make a priori predictions of the functional consequences of genetic polymorphism for these related loci very difficult.
We confirmed that age is a very strong predictor of osteonecrosis.3,4 Because younger age is associated with faster systemic clearance of most antileukemic agents,41 it is possible that age serves partly as a surrogate for drug clearance, with older children having higher systemic exposure to the medications that cause osteonecrosis.
With only 64 patients and 16 polymorphisms studied, there is a risk of type I error, that the relationships we observed were due to chance, or of type II error, that we missed some important pharmacogenetic determinants. However, as opposed to a genome-wide approach, we have interrogated a relatively small number of loci that all have been associated with clinical or preclinical functional consequence; this target gene approach is less likely than many others to result in false-positive findings. It should also be acknowledged that these pharmacogenetic findings were observed in a patient group in whom the occurrence of osteonecrosis was higher than that likely had the entire cohort of children with ALL been screened; such prospective screening is currently underway at St Jude. The two pharmacogenetic polymorphisms associated with osteonecrosis have biologic plausibility, and if confirmed in other patient cohorts, could provide the foundation for future dosage individualization based on simple genetic tests.
Authors' Disclosures of Potential Conflicts of Interest
The following authors or their immediate family members have indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. Research Funding: Mary V. Relling, National Institutes of Health; Ching-Hon Pui, American Cancer Society. Other Remuneration: Mary V. Relling, Prometheus. For a detailed description of these categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration form and the “Disclosures of Potential Conflicts of Interest” section of Information for Contributors found in the front of every issue.
Acknowledgments
We thank Pam McGill, Nancy Duran, Peixian Chen for technical assistance; our clinical and research faculty and staff; and the patients and their families for participating.
Footnotes
-
Supported by NCI CA51001, CA78224, CA21765, and the NIH/NIGMS Pharmacogenetics Research Network and Database (GM61393, GM61374 http://pharmgkb.org/</index.html) from the National Institutes of Health; by a Center of Excellence grant from the State of Tennessee; by an F.M. Kirby Clinical Research Professorship from the American Cancer Society; and by American Lebanese Syrian Associated Charities (ALSAC).
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received November 5, 2003.
- Accepted July 26, 2004.
