Ethnic-Specific Predictors of Neurotoxicity Among Pediatric Acute Lymphoblastic Leukemia Patients Following High-Dose Methotrexate

Abstract

High-dose methotrexate (HD-MTX; 5,000 mg/m²) is an important component of curative therapy in many treatment regimens for high-risk pediatric acute lymphoblastic leukemia (ALL). However, methotrexate therapy can result in dose-limiting neurotoxicity which may disproportionately affect Latino children. Thus, we evaluated risk factors for neurotoxicity in an ethnically diverse population of 351 patients (58.1% Latino) who received 1,183 HD-MTX infusions. Overall, thirty-five patients (10%) experienced neurotoxicity, 71% of whom were Latino. After adjusting for clinical risk factors, we found that serum creatinine elevations ≥50% of baseline were associated with a 3-fold increased odds (OR = 3.32, 95% CI: 0.98-11.21, p=0.05) for neurotoxicity when compared to creatinine elevation <25%. Notably, predictors of neurotoxicity differed by ethnicity. Specifically, Latino children experienced a nearly six-fold increase in neurotoxicity odds (OR = 5.80, 95% CI: 1.39-24.17, p=0.02) with serum creatinine elevation ≥50%, compared to creatinine elevation <25%. Thus, serum creatinine elevation ≥50% may be associated with an increased risk for neurotoxicity among Latino children with ALL and may identify potential candidates for therapeutic or supportive care interventions.

Precis:

Serum creatinine elevations ≥50% of baseline are associated with an increased risk for methotrexate-related neurotoxicity among pediatric acute lymphoblastic leukemia patients of self-reported Latino ethnicity.

1 ∣. Introduction

Methotrexate is an integral component of curative chemotherapy for pediatric acute lymphoblastic leukemia (ALL). When given in high-doses (i.e., 5,000 mg/m²) methotrexate provides added protection against central nervous system (CNS) relapse and significantly improves survival in B-ALL patients with high-risk disease (1-4). However, high-dose methotrexate (HD-MTX) can result in dose-limiting neurotoxicity, with a reported incidence among pediatric ALL patients ranging between 3 and 14% (1,5,6). Although acute symptoms are typically transient, neurotoxicity can be both critical and enduring, often resulting in methotrexate treatment modifications. These modifications may include either dose omissions or complete elimination of methotrexate, which potentially diminish treatment efficacy (5,7,8). Clinical predictors of methotrexate neurotoxicity are poorly defined, which is a major barrier to the development and delivery of neurotoxicity prevention and supportive care interventions.

The risk of methotrexate-related neurotoxicity appears to correlate with older age (>10 years), comparatively increased exposure to HD-MTX, and liver dysfunction during the induction or consolidation phases of therapy (5,6,9). Additionally, there is growing evidence that methotrexate differentially affects Latino patients with ALL. For example, previous studies have reported a higher frequency of both delayed methotrexate clearance and methotrexate-related neurotoxicity among Latino children compared to non-Latino white children (2,10). Because methotrexate-related toxicity may adversely impact the delivery of curative therapy and compromise treatment efficacy, ethnic differences in methotrexate toxicities may contribute to the well-characterized disparities in ALL relapse and survival experienced by Latino children with ALL (8,11,12). Recognizing these important disparities, the objectives of this study were to (i) evaluate clinical predictors of methotrexate-related neurotoxicity following HD-MTX and (ii) assess whether the prognostic potential of these factors differed between racial and ethnic groups.

2 ∣. Methods

2.1. Study Population

We performed an updated and expanded retrospective chart review of a previously published cohort of patients diagnosed with B- or T-lineage ALL at ≤21 years of age who received at least one dose of HD-MTX at Texas Children’s Cancer Center (Houston, TX) from October 2010 through December 2017 (2). All patients were treated on or according to one of the following Children’s Oncology Group (COG) protocols: AALL0031 (NCT00022737), AALL01P2 (NCT00049569), AALL0232 (NCT00075725), AALL02P2 (NCT00096135), AALL0331 (NCT00103285), AALL0433 (NCT00381680), AALL0434 (NCT00408005), AALL08P1 (NCT00866307), AALL1122 (NCT01460160), AALL1131 (NCT02883049), AALL1231 (NCT02112916). Specific treatment information for each protocol is accessible on www.clinicaltrials.gov and have been well documented elsewhere (3,4,13-21). Eligible patients received at least one HD-MTX (5,000 mg/m²) dose as an inpatient intravenous (IV) infusion. Ten percent of the dose was administered as a bolus over 30 minutes followed by the remaining 90% of the dose over 23.5 hours. Baseline creatinine levels were obtained for all patients immediately prior to each HD-MTX infusion. Beginning at 24 hours from the start of the infusion, methotrexate and creatinine levels were measured at regular protocol-specific intervals until threshold levels for methotrexate clearance were met. Standard supportive care with hydration, alkalinization, leucovorin rescue, and antiemetics was provided per the treatment protocol. This study was approved by the Baylor College of Medicine (Houston, TX) Institutional Review Board.

2.2. Data Collection

Variables related to drug clearance, creatinine, supportive care, and sociodemographic factors were systematically abstracted from the electronic health record. Racial and ethnic variables were collected using self-report forms based on standard government definitions. For the purposes of this analysis, ethnicity was classified as Latino or non-Latino while race was categorized as white, Black (or African American), or “other categories” (which included American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, or other/not reported).

Methotrexate levels were serially monitored starting at 24 hours after the initiation of the methotrexate according to standard practice, which at a minimum was performed at hour 24, 42, 48 and then every 12 hours until “clear” (as subsequently described). Time to methotrexate clearance was calculated as the number of hours between starting the methotrexate infusion (hour 0) and the time at which methotrexate met a clearance parameter. For example, patients with a methotrexate concentration ≤0.4 μM at hour 48 met parameters for on-time clearance. Patients with a methotrexate concentration >0.4 μM at hour 48 were delayed, and time to clearance was defined as the time at which serum methotrexate reached <0.1 μM. Serum creatinine elevations were calculated as the percent increase in serum creatinine from the pre-infusion baseline level. Maximum creatinine elevations ≥25% and ≥50% of baseline trigger additional clinical intervention per protocol and were used as indicators of clinical change in our analysis. Glucarpidase, a rescue medicine used for severe methotrexate-induced nephrotoxicity or delayed clearance, was administered for severely elevated methotrexate level or creatinine based on reported guidelines for use (22). Some patients received methotrexate according to an institutional Intensive Monitoring Protocol (IMP) (23,24), where multiple levels in the first 24 hours were obtained and the methotrexate infusion rate was adjusted according to the IMP. Body mass index (BMI) Z-scores were calculated from the height and weight at the time of methotrexate infusion using sex- and age-specific growth charts (25).

Methotrexate neurotoxic events were defined as neurologic episodes (e.g., seizures, stroke-like symptoms, or aphasia) occurring within 21 days of HD-MTX infusion which resulted in subsequent methotrexate treatment modifications. All events were confirmed after an independent review of suspected cases by at least two clinical investigators. No discrepancies in case ascertainment between the reviewers were reported.

2.3. Statistical Analysis

All statistical analyses were conducted in Stata version 17 (Stata-Corp, College Station, TX). Appropriate descriptive statistics for continuous (e.g., mean, standard deviation (SD)) and categorical (e.g., frequency and proportion of total) variables were calculated to describe the study population. Differences in baseline covariates by neurotoxicity status were evaluated using t-tests for continuous variables and Chi-square test for categorical variables. Clinical response across the study population were summarized using appropriate statistics. Differences in therapeutic response across doses were evaluated using ANOVA or chi-square test, where appropriate. Mixed effects logistic regression was used to estimate the crude and adjusted odds ratio (OR) and corresponding 95% confidence interval (CI) for the association between clinical factors and neurotoxicity following each course of HD-MTX, accounting for repeated observations within individuals. All clinical and sociodemographic variables of interest were evaluated in association with neurotoxicity in unadjusted models. Variables evaluated include methotrexate clearance and concentration at hour 48 and hour 24, time to clearance, creatinine elevation, glucarpidase administration, intensive monitoring, order of the methotrexate dose, race, ethnicity, age at diagnosis, obesity status, leukemia subtype, and sex. Any variable with p<0.20 was then considered for the multivariable analysis. Final, adjusted multivariable models included statistically significant (p<0.05) variables. In addition, based on published literature, the a priori decision was made to account for age at diagnosis, race, ethnicity, time to methotrexate clearance, and IMP in all multivariable models. Because ethnic disparities in methotrexate-related neurotoxicity have been reported, we also performed secondary analyses stratifying on ethnicity and used the likelihood ratio test to assess whether effect estimates between clinical predictors and neurotoxicity differ significantly between ethnic groups. Prior to model building, we performed regression diagnostics (e.g., VIFs < 1.5) to confirm model assumptions were satisfied. Statistical significance was set at p<0.05.

3 ∣. Results

A total of 1,183 doses of methotrexate were administered to 351 eligible patients. Tables 1 and 2 present demographic and clinical characteristics of the study population. Overall, the study population was 58% Latino, 62% male, with a median age of 9 years at diagnosis (range: 1- 21 years). In total, 10% (n=35) of patients experienced a neurotoxic event. Twelve percent (n=25) of Latino children experienced a neurotoxic event while 7% (n=10) of non-Latino patients experienced a neurotoxicity. Sixty-three percent (n=22) of all primary events occurred following the first HD-MTX dose (Figure 1). Seventy-one percent of all patients who experienced neurotoxicity were Latino (n=25), while 57% of those without a neurotoxic event were Latino (p=0.09), which is representative of the patient population in our clinic. One patient experienced a second neurotoxic event upon methotrexate rechallenge for a total of 36 events included in this study. Rechallenge with IV methotrexate after a neurotoxic event was common (but ultimately left to the individual provider) in patients who fully recovered from a previous neurotoxic event and for whom additional doses of IV methotrexate were clinically indicated.

Table 1.

Demographic characteristics of pediatric acute lymphoblastic leukemia patients (n=351)

	Neurotoxicity Status
	Cases (n=35)	Controls (n=316)	Total	p-valuec
Ethnicity, n(%)
Non-Latino	10 (28.6)	137 (43.4)	147 (41.9)
Latino	25 (71.4)	179 (56.6)	204 (58.1)	0.09
Race, n(%)
Asian	0 (0.0)	23 (7.3)	23 (6.6)
Black	3 (8.6)	31 (9.8)	34 (9.7)
Native American	2 (5.7)	10 (3.2)	12 (3.4)
Unknown	0 (0.0)	6 (1.9)	6 (1.7)
White	30 (85.7)	246 (77.8)	276 (78.6)	0.39
Sex, n(%)
Female	16 (45.7)	119 (37.7)	135 (38.5)
Male	19 (54.3)	197 (62.3)	216 (61.5)	0.35
Leukemia Subtype, n(%)
B-ALLa	27 (77.1)	257 (81.3)	284 (81.9)
T-ALLa	8 (22.9)	59 (18.7)	67 (19.1)	0.55
Obesity Status, n(%)
Under or Normal Weight	24 (68.6)	227 (71.8)	251 (71.5)
Obese	11 (31.4)	89 (28.2)	100 (28.5)	0.69
Age at Diagnosis, year (SD)b	12.4 (2.0)	8.5 (4.9)	8.9 (4.9)	<0.01

^aALL, acute lymphoblastic leukemia

^bSD, standard deviation

^cchi-square test

Table 2.

Clinical characteristics of 1183 methotrexate dose administrations among 351 pediatric acute lymphoblastic leukemia patients

	Dose Order
	1	2	3	4	Overall	p-valueb
Creatine Level at Baseline, mean (SD)a	0.41 (0.2)	0.40 (0.2)	0.39 (0.2)	0.40 (0.1)	0.40 (0.2)	0.28
Creatinine Elevation, n(%)
<1.25	246 (71.1)	231 (75.9)	191 (69.7)	212 (81.8)	880 (74.4)
1.25 - 1.49	62 (17.9)	47 (15.5)	57 (20.8)	32 (12.4)	198 (16.7)
1.5+	38 (11.0)	26 (8.6)	26 (9.5)	15 (5.8)	105 (8.9)	0.03
Clearance at Hour 48, n(%)
Cleared	209 (60.8)	200 (65.8)	173 (63.6)	159 (61.4)	741 (62.9)
Delayed	135 (39.2)	104 (34.2)	99 (36.4)	100 (38.6)	438 (37.1)	0.58
Hours to Clearance, mean (SD)a	72.2 (49.1)	66.6 (36.0)	64.3 (28.9)	63.7 (25.2)	67.1 (37.2)	0.02
Neurotoxicity, n(%)
No Neurotoxic Event	324 (93.6)	299 (98.4)	269 (98.2)	255 (98.5)	1,147 (97.0)
Neurotoxic Event	22 (6.4)	5 (1.6)	5 (1.8)	4 (1.5)	36 (3.0)	<0.01

^aSD, standard deviation

^bANOVA for continuous and chi-square for categorical

Figure 1. Primary Neurotoxic Events.

Percent of children experiencing a neurotoxic event both overall and occurring after the first four HD-MTX infusions.

Most cases presented with stroke-like symptoms (n=26), followed by seizure (n=4), altered mental status or aphasia (n=3), vision changes (n=1), or muscle weakness and/or numbness (n=2). The average time from methotrexate infusion to neurotoxicity symptoms was 10 days (range: 3-20 days). Thirty-three (92%) of the events resulted in a modification to the intrathecal (IT) methotrexate and thirty-two (89%) of the events resulted in a change to the IV methotrexate therapy. Of those requiring modifications to IT therapy, 20 had IT therapy held or delayed, 22 had IT methotrexate replaced by cytarabine and or hydrocortisone, and 9 had both. Of those requiring modifications to IV therapy, HD-MTX was reversed with Capizzi dosing among 20 patients, omitted in subsequent doses for 10 patients, with 2 patients having both modifications. For one individual with neurotoxicity, IT and IV methotrexate schedules were also modified to not occur on the same day.

Unadjusted associations between clinical/demographic variables and neurotoxicity are presented in Supplemental Tables 1 and 2. In adjusted models (Table 3), we found patients diagnosed at older ages (OR=1.23, 95% CI: 1.10-1.38, p<0.01) were significantly more likely to experience neurotoxicity. We also observed a suggestive, but not statistically significant, association among patients with a ≥50% creatinine elevation from baseline having a 3-fold increased odds (OR=3.32, 95% CI: 0.98-11.21, p=0.05) for neurotoxicity when compared to those with a creatinine elevation <25%.

Table 3.

Multivariable^* mixed effects logistic regression for clinical predictors of neurotoxicity

	Odds Ratio	95% CIa Lower Bound	95% CIa Upper Bound	p-value
Creatinine Elevation
<1.25	Ref.
1.25 - 1.49	1.78	0.69	4.59	0.23
1.5+	3.32	0.98	11.21	0.05
Race
White	Ref.
Black	1.32	0.29	5.98	0.72
Other	0.61	0.13	3.00	0.55
Ethnicity
Non-Latino	Ref.
Latino	2.15	0.84	5.47	0.11
Time to Clearance b	0.94	0.85	1.05	0.27
Age at Diagnosis (years)	1.23	1.10	1.38	<0.01

^aCI, confidence interval

^bModeled in 12-hour increments

^*Model adjusted for patient monitoring schedule, and methotrexate dose order

The association between clinical factors and neurotoxicity significantly differed between ethnic groups (test for interaction p-value=0.01). A creatinine increase ≥50% from baseline preceded 62% (n=16) of neurotoxic events in Latino patients whereas, among non-Latin children, a serum creatinine elevation ≥50% from baseline preceded just 20% (n=2) of neurotoxic events. In multivariable models stratified on ethnicity (Table 4), we again observed the adjusted association between serum creatinine elevations ≥50% from baseline and neurotoxicity was most pronounced in Latino patients. Specifically, Latinos with a creatinine elevation ≥50% experienced an almost 6-fold increased odds (OR=5.80, 95% CI: 1.39-24.17 p=0.02) for neurotoxicity, compared to Latinos with a creatinine elevation <25%. Conversely, creatinine elevations ≥50% did not reliably precede neurotoxicity in non-Latino patients (p=0.74). In these children, longer time to methotrexate clearance was suggestively associated with subsequent neurotoxicity (OR=1.19, 95% CI: 0.98-1.46, p=0.08), a trend which was not observed in Latino patients (OR=0.88, 95% CI: 0.78-1.01, p=0.08). Overall, Latino patients were slightly more likely than non-Latinos to experience creatinine elevations, although this difference did not reach statistical significance on initial univariable analysis (Supplemental Table 3). The distributions of time to methotrexate clearance also did not differ significantly between Latino and non-Latino patients.

Table 4.

Multivariable^* mixed effects logistic regression for clinical predictors of neurotoxicity stratified by ethnicity

	Non-Latino (n=147)			Latino (n=204)
	ORa	95% CIb	p-value	ORa	95% CIb	p-value
Creatinine Elevation
<1.25	Ref.			Ref.
1.25 - 1.49	1.70	(0.31, 9.38)	0.54	2.10	(0.64, 6.82)	0.22
1.5+	0.62	(0.38, 10.30)	0.74	5.80	(1.39, 24.17)	0.02
Time to Clearance c	1.19	(0.98, 1.46)	0.08	0.88	(0.78, 1.01)	0.08
Age at Diagnosis (years)	1.13	(0.98, 1.30)	0.08	1.29	(1.09, 1.52)	<0.01

^aOR, odds ratio

^bCI, confidence interval

^cModeled in 12-hour increments

^*Model adjusted for race, dose order, and monitoring schedule

4 ∣. Discussion

Consistent with previous reports (2,26,27), Latino children with ALL were slightly more likely than non-Latinos to experience neurotoxicity following HD-MTX, although this difference was not statistically significant in the current study (p=0.09). HD-MTX is a critical component of curative chemotherapy for many pediatric ALL patients with high-risk disease. However, neurotoxicity is a frequent complication of HD-MTX which often leads to modifications in methotrexate chemotherapy. Emerging data suggest methotrexate-related neurotoxicity disproportionately affects Latino patients with ALL and may contribute to poorer overall and event-free survival in affected individuals. Specifically, it has been demonstrated that children who experience neurotoxicity tend to receive, on average, 2-3 fewer doses of curative methotrexate therapy (10). The downstream effects of the withheld treatment are unknown but may adversely impact patient survival and outcomes (10). Given this, it follows that ethnic disparities in neurotoxicity may contribute to well-recognized disparities in ALL outcomes. This study identified serum creatinine elevations and time to methotrexate clearance as potentially important clinical factors associated with neurotoxicity following courses of HD-MTX. Importantly, the strengths of the associations between each of these factors and neurotoxicity risk appeared to differ between Latinos and non-Latinos.

The proportion of patients experiencing neurotoxicity in the current study (10%) is consistent with rates reported in previous studies (3-14%) (5,6,10). However, most published estimates reflect the frequency of neurotoxicity following any intrathecal or intravenous methotrexate over all phases of therapy. This is the first study to evaluate neurotoxicity specifically occurring after treatment with HD-MTX. Methotrexate is primarily cleared through the kidneys via glomerular filtration and active tubular secretion (28). Several prior studies suggest that renal clearance of methotrexate differs between Latinos and non-Latinos, with Latinos being more likely to require glucarpidase for renal toxicity (27) and experience delayed methotrexate clearance (2), suggesting ethnic disparities in renal dysfunction among childhood ALL patients. This may provide the biological rational underlying our finding that creatinine elevation, a biomarker of renal function, is strongly linked to subsequent methotrexate-related neurotoxicity. Polymorphisms in several genes have been implicated in methotrexate metabolism, transport, and toxicity (29-32). For example, a pharmacogenomic variant commonly studied for its contribution methotrexate toxicity is MTHFR 677C>T (rs1801133). This T-allele of this polymorphism has been linked to reduced MTHFR activity and altered susceptibility to methotrexate-related toxicities in various populations (33-36). Although meta-analyses on this variant have not yielded a significant finding, interest in the variant remains, with ongoing research examining the impact of co-existing haplotypes or other modifier alleles within other genes that together (perhaps with MTHFR) increase the risk of toxicity (37). Notably, the frequency of the T-allele appears to differ considerably between ancestral populations, occurring in approximately 18% of HapMap reference populations of African genetic ancestry to 53% of individuals of predominantly European genetic ancestry and 66% of individuals of Mexican ancestry (38). Ancestral differences in the frequency of key pharmacogenomic variants may in part contribute to the ethnic differences in clinical outcomes observed in the current study. However, additional research is warranted to fully characterize the mechanisms responsible for possible disparities observed in creatinine elevation, delayed methotrexate clearance, and subsequent toxicities.

This study leveraged a large, ethnically diverse cohort to explore clinical factors associated with methotrexate-related neurotoxicity following HD-MTX in historically understudied populations. Despite these strengths, there are several limitations to the study. First, the retrospective single-institution nature of this study made it impossible to prospectively target clinical parameters and outcomes among geographically diverse populations. Similarly, our ascertainment of neurotoxicity was limited to information available in the electronic medical record. In many cases, the level of detail provided was not sufficient to clearly grade the severity of neurotoxic events or determine if or when symptoms completely resolved. Finally, we accounted for many potential confounding factors in the multivariable analysis, but as an observational study, the potential for residual confounding or confounding due to unmeasured factors remains.

Our study adds to the growing body of evidence indicating that Latino children are at higher risk for toxicity due to HD-MTX. Furthermore, our findings indicate that a specific risk factor, a ≥50% increase in serum creatinine, is associated with a greater risk of neurotoxicity in Latino compared to non-Latino children. Further studies are needed to identify optimal neuroprotective interventions to use in at-risk patients. Still, this discovery should prompt clinicians to be more vigilant with their Latino patients with observation, monitoring and hospitalization. In addition, increases in creatinine should be addressed per protocol or institutional standard quickly and accurately. Some possibilities for future neuroprotective interventions include broader application of IMP or increasing the use of alkalinized hydration, which are likely to reduce renal toxicity and aid in methotrexate clearance (39,40). Additionally, effective monitoring and neurotoxicity mitigation strategies may help also reduce unnecessary morbidity. Our institutional experience suggests individualized HD-MTX infusion rates based on intensive monitoring of serum methotrexate levels drawn at 2 hours and 6-8 hours into the infusion may be a safe and effective alternative to traditional HD-MTX administration protocols (23,24). Given the potential impact of neurotoxicity on ALL treatment efficacy (6,10), effective neuroprotective interventions may help to narrow ethnic disparities in relapse and survival in children with ALL.