Urinary coproporphyrin I/(I + III) ratio as a surrogate for MRP2 or other transporter activities involved in methotrexate clearance

Abstract

Aims

The urinary coproporphyrin I/(I + III) ratio may be a surrogate for MRP2 activity. We conducted a prospective study in patients receiving methotrexate (MTX) to examine the relationship between this ratio and the pharmacokinetics of a MRP2 substrate.

Methods

Three urine samples were collected from 81 patients for UCP I/(I + III) ratio determination: one before (P1), one at the end of MTX infusion (P2), and one on the day of hospital discharge (P3). Three polymorphisms of ABCC2 were analysed and their relationships with basal UCP I/(I + III) ratio values assessed. All associated drugs were recorded and a drug interaction score (DIS) was assigned. Population pharmacokinetic analysis was conducted to assess whether MTX clearance (MTXCL) was associated with the basal UCP I/(I + III) ratio, its variation during MTX infusion, the DIS or other common covariates.

Results

The basal UCP I/(I + III) ratio was not associated with ABCC2 polymorphisms and did not differ according to the DIS. Significant changes in the ratio were observed over time, with an increase between P1 and P2 and a decrease at P3 (P < 0.001). No association was found between basal UCP I/(I + III) ratio and MTXCL. The final model indicates that MTXCL was dependent on the change in the ratio between P1 and P3, DIS and creatinine clearance.

Conclusion

The basal UCP I/(I + III) ratio is not predictive of MTXCL. However, it is sensitive to the presence of MTX, so it is plausible that it reflects a function modified in response to the drug.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• High dose methotrexate (HD-MTX) can cause severe toxicity in patients who do not eliminate the drug well.

• Elimination of MTX from humans is primarily renal and involves various membrane transporters including organic anion transporters (OAT1, OAT3), multi-drug resistance proteins (MRP2, MRP4) and breast cancer resistance protein (BCRP). Polymorphisms in the SLCO1B1/OATP1B1 have been found to be associated with MTX clearance suggesting that the liver may also contribute.

• All these transporters are under genetic control and may be inhibited by co-administered drugs. However, the contribution of each transporter to MTX elimination is unknown and difficult to assess in vivo.

• The ratio between urinary coproporphyrins (the UCP I/(I + III) ratio) is used to diagnose patients with Dubin–Johnson's syndrome, caused by mutations in the ABCC2/MRP2 gene. The value of this ratio in healthy subjects is also dependent on ABCC2 polymorphisms, thus making it a potential surrogate marker for MRP2 function.

WHAT THIS STUDY ADDS

• Contrary to our hypothesis, the basal UCP I/(I + III) ratio cannot be used to predict MTX_CL. However, we report that HD-MTX infusion causes changes in the UCP I/(I + III) ratio over time, suggesting that MTX may initially inhibit, then up-regulate the putative functions represented by this ratio. MTX clearance is also modulated while these changes are taking place, and therefore the UCP I/(I + III) ratio may be a useful and innovative tool for investigating the pharmacokinetics of MTX or other transporter's substrates.

• Our study suggests that drug transporters may be modulated by their own substrates, possibly leading to dose or time-dependent elimination.

Introduction

There is substantial inter and intra-individual variability in the pharmacokinetics of methotrexate (MTX) administered as high dose regimen. This can cause significant clinical problems even if susceptibility factors, such as glomerular filtration rate (GFR), age or known drug interactions, are taken into account [1–3].

The main mechanism of MTX elimination from humans after intravenous infusion is renal excretion, with biliary secretion contributing less than 30% [4,5]. MTX and its major metabolite, 7-OH-MTX, are eliminated by glomerular filtration and tubular secretion which involves various membrane transporters on proximal tubular cells: OAT1/SLC22A6 and OAT3/SLC22A8 at the basal pole, and MRP2/ABCC2, MRP4/ABCC4 and BCRP/ABCG2 at the apical pole (the nomenclature used here for transporters is in accordance with the Guide to Receptors and Channels (GRAC) [6]). Little is known about the extent and the determinants of variability of transporter function in man but both drug interactions and genetic polymorphisms are now recognized as being important. Transporter-mediated drug interactions involve members of the OAT [7,8] and OATP [9] families, and also MRP2 [10], MRP4 [10] and BCRP [11]. As a result, a number of drugs are contraindicated or not recommended during the administration of MTX. They include probenecid, NSAIDs, β-lactamins and gemfibrozil (for review, see Leveque et al. [12]). The contribution of genetic polymorphism to the variability of MTX clearance (MTX_CL) has also been studied. Recent studies consistently indicate that polymorphisms in OATP1B1/SLCO1B1 have substantial effects [13–16], and there is evidence that polymorphisms in RCF/SLC19A2 [17] and MRP2/ABCC2 may also contribute [18–20].

The importance of MRP2 for the in vivo distribution of MTX has been clearly established by pharmacokinetic studies in rats and mice lacking Mrp2 [21]. However, transposition of the results obtained in animals to the human beings is limited by interspecies differences in transporter functions and tissue expression patterns [22]. The differences involve both the substrate specificities of particular transporters and the relative fraction of clearance depending on each of them. For example, MTX_CL depends mainly on biliary elimination, through Mrp2, in laboratory animals, but is primarily renal in humans with the contribution of MRP2 being unknown [23].Work in vitro suggests that MRP2 is also a target central to drug interactions [24,25], although confirmatory studies in vivo in man are still lacking for some of these interactions. Similarly, the consequences of ABCC2/MRP2 polymorphisms for MTX_CL remain unclear. Rau et al. [18] observed higher plasma concentrations and lower clearance in girls carrying the −24T allele of ABCC2, whereas Simon et al. found that MTX_CL was 30% higher for patients carrying the T allele than for controls [19]. We reported previously that the urinary coproporphyrin I / coproporphyrin I + III ratio-UCP I/(I + III) ratio, a marker commonly used to diagnose patients with Dubin–Johnson syndrome (i.e. with ABCC2 mutations), is dependent on ABCC2 polymorphisms and may serve as a biomarker of MRP2 activity in humans [26]. We report a study aimed at analyzing the inter and intra-individual variation of the UCP I/(I + III) ratio and assessing whether its basal value may be a determinant of MTX_CL. The existence of such an association would provide insights into the role of MRP2 in MTX elimination in humans and the UCP I/(I + III) ratio may serve as an innovative tool for predicting the capacity of a subject to eliminate MRP2's substrates.

Methods

Recruitment of patients

The COMETH study (clinicaltrial.gov number NCT00822432) was conducted in patients hospitalized in the haematology department of University Hospital of Tours and in the haematology and neurology departments of Pitié-Salpêtrière University Hospital in Paris. All patients were more than 18 years of age and were prescribed chemotherapy with high dose (HD)-MTX (>1 g m⁻²) for a lymphoid malignancy. The non-inclusion criteria were severe renal impairment (creatinine clearance <30 ml min⁻¹. 1.73 m^–2), severe hepatic insufficiency (prothrombin time and/or factor V < 50%), elevated liver enzyme activities (ALT and AST>3 times normal), chronic respiratory insufficiency, pregnancy, breastfeeding, concomitant medication with phenytoin, probenecid, trimethoprime, phenylbutazone, NSAIDs or salicylates, and vaccination against yellow fever. Patients who had received MTX within the 3 months before enrolment were not included. The COMETH protocol was approved by the ethics committee of Pitié-Salpêtrière University Hospital (09/05/2007) and all subjects gave written informed consent.

Study protocol

Patients were monitored during the first HD-MTX infusion, administered over 3 or 24 h, depending on the chemotherapy regimen. Pre-hydration and urine alkalinization with intravenous sodium bicarbonate to maintain urine pH>7 were applied. Blood samples for plasma MTX concentration determinations were collected at 24 h, 48 h and 72 h, and then every 24 h until concentrations fell below the non-toxic threshold of 0.2 μmol l⁻¹. An additional blood sample was collected at the end of the infusion from patients receiving 3 h infusions (3 h). The concentration measured at 72 h was used to assign patients to the ‘good elimination’ group (72 h < 0.2 μm) or ‘delayed elimination’ group (72 h ≥ 0.2 μm). If the 72 h sample was not available, the concentration was simulated using estimated pharmacokinetic parameters for the individual (see PK analysis). Biological (blood cell counts, serum creatinine, liver enzymes) and clinical variables reflecting mucocutaneous, digestive or other toxicities were recorded on day 1 and then once a week after MTX administration.

UCP I/(I + III) ratio: basal value and effect of associated factors

Three different urine samples were used to explore this variable.

The first was that collected before administration of MTX, and before the start of alkaline pre-hydration. The basal value of the UCP I/(I + III) ratio, referred to as ‘P1’, was determined in each patient. To investigate whether it specifically reflects MRP2 function, we studied whether the basal value was dependent on known variant alleles of ABCC2. Patients were genotyped for the three single nucleotide polymorphisms (SNPs) in ABCC2 with the largest effects on the pharmacokinetics of MRP2 substrates (rs717620-c.-24C/T, rs2273697-c.1249G/A and rs3740066-c.3972C/T) [27–31]. We also assessed whether the basal UCP I/(I + III) ratio was dependent on co-administered drugs, and particularly those that may be MRP2 substrates or inhibitors. Thus, all medications (regular and occasional treatments, including those associated with chemotherapy) taken at the time of the P1 sampling were precisely recorded. Each drug was assigned a ‘drug interaction score’ (DIS), based on data in the literature [23,32–47]. A drug known to be involved in an interaction with MTX was assigned a score of 2 (DIS = 2), a drug possibly interacting with transporters (known substrate of MRP2 or of other transporters involved in the elimination of MTX) was scored 1 (DIS = 1) and a drug with no theoretical interaction with transporters was scored zero (DIS = 0).

As MTX itself may interfere with MRP2, being a substrate of this transporter, MTX treatment may affect the UCP I/(I + III) ratio. Thus, a second urine sample (referred to as ‘P2’) was collected at the end of the infusion (when MTX concentrations are at their highest), i.e. at 3 h or at 24 h (3 or 24 h, respectively, after the beginning of the infusion), according to the chemotherapy regimen. A third sample (‘P3’) was collected when the patient left the hospital to determine whether the UCP I/(I + III) ratio returned to its basal value after complete elimination of MTX. We also calculated the change in the UCP I/(I + III) ratio relative to the basal value at these two different times, i.e. at the end of MTX infusion (DP2 = P2/P1) and at the time of hospital discharge (DP3 = P3/P1).

Covariates for MTX population pharmacokinetic modelling

Covariates were selected by analyzing factors that may influence MTX elimination from the body. Some were classical covariates, such as bodyweight, height, body surface area, age, gender, serum creatinine, creatinine clearance, serum urea, total protein, albumin, AST, ALT, PAL, GGT, LDH and bilirubin.

We evaluated whether the basal UCP I/(I + III) ratio, and its variation during the HD-MTX treatment (i.e. DP2 and DP3 values) were associated with MTX pharmacokinetics.

We also studied the effect of co-administered drugs. Each drug was considered individually by testing the effect of its DIS as a covariate in the pharmacokinetic model, i.e. by testing the relationships between pharmacokinetic parameters and each of ‘at least one drug of score 2’ (SCO2, Yes/No) and ‘at least a drug of score 1 but no score 2’ (SCO1, Yes/No).

Assessment of toxicity

Toxicity evaluations were performed by the medical staff at the end of MTX infusion and until the end of the hospitalization. Toxicities were scored according to recommendations of the Common Terminology Criteria for Adverse Events v.4.03 (CTCAE).

We assessed whether the occurrence of severe toxicity (grade 3–4) was dependent on MTX_CL, delayed elimination or DIS. We also tested whether the basal UCP I/(I + III) ratio was associated with toxicity.

Determination of urinary coproporphyrin concentrations and of the UCP I/(I + III) ratio

Urinary coproporphyrin concentrations were determined by HPLC, using a method described previously [48]. We quantified the two isomers of coproporphyrin from standard curves established with coproporphyrin calibrators (Recipe, Munich, Germany). The UCP I/(I + III) ratio was then obtained by dividing the peak height for isomer I by the sum of the peak heights for the isomers I and III. The value is reported as a percentage.

ABCC2 genotyping

DNA was extracted from peripheral blood leukocytes with Puregene Blood Core Kits (Qiagen, Hilden, Germany) in Paris and Flexigene DNA kits (Qiagen, Hilden, Germany) in Tours, both according to the manufacturers' recommendations. ‘TaqMan® Drug Metabolism Genotyping Assays’ (Applied Biosystems, Foster City, CA, USA) kits and the Sequence Detection System 7900HT (Applied Biosystems) were used for genotyping by allelic discrimination.

The three SNPs were tested for Hardy–Weinberg equilibrium.

Determination of MTX concentrations

MTX plasma concentrations were determined by EMIT on an Xpand Dimension® analyzer (Siemens, Deerfield, IL, USA) in Paris, and by Fluorescence Polarization Immunoassay (Abbott FPIA-2) on a TDx® device (Abbott Park, IL, USA) in Tours. The EMIT assay has a mean precision of 5.6%, a mean accuracy of 4% and a limit of quantification (LOQ) of 0.03 μm (kit specification sheet). For the FPIA-2 assay, precision is in the range of 5.4 to 14% and accuracy 0 to 3.51%; the LOQ was 0.05 μm (kit specification sheet). Both tests may interfere with metabolite 4-amino-4-deoxy-N-methylpteroic acid (DAMPA), but this metabolite is normally present at very low concentrations. Internal and external quality control assessments were carefully performed throughout the study period within each laboratory. No cross-validation was performed. Comparisons between these assays were already available in the literature and show discordance of below 20% [49]. However, to account for the consequences of possible analytical variability, we included the assay type as a covariate in the POP-PK model.

Population pharmacokinetic analysis

Pharmacokinetic analysis was performed using the non-linear mixed effect modeling program nonmem (version VI; Globomax LLC, Hanover, USA) with the interface Wings for nonmem (http://wfn.sourceforge.net/). The first order conditional estimation was used for parameter estimations. A two compartment model was used and four pharmacokinetic parameters were estimated: CL total clearance, V₁ volume of central compartment, V₂ volume of peripheral compartment and Q inter-compartment clearance. Various error models were tested (proportional, additive, exponential) to describe interindividual and residual variabilities.

The influence of continuous covariates was investigated using the following modes:

where TVP is the typical value of a pharmacokinetic parameter, θ_i is the mean value of the pharmacokinetic parameter in a patient with a covariate value (COV) equal to the median value of covariate in the study population (MED_COV), and θ_(i+1) is the effect of the covariate on the pharmacokinetic parameter.

The influence of categorical covariates was tested using the following equation:

Covariates were selected using a forward and backward selection process. During forward selection, covariates were selected if they were biologically relevant, if they decreased the objective function value (OFV) of the basic model by at least 3.84 and if they decreased the inter-individual variability of pharmacokinetic parameters. All significant covariates were then incorporated into an intermediate model. Each covariate was independently removed from the intermediate model to confirm its relevance. An increase in OFV>6.63 (P < 0.01) was required to confirm that the covariate was significant and thus its retention in the model. The resulting model was called the ‘final’ model.

Diagnostics of the final model were established using plots of observed concentrations vs. predicted and vs. individual predicted concentrations, and plots of individual weighted residuals vs. time and vs. predicted concentrations. A total of 1000 Monte Carlo simulations were performed to calculate normalized prediction distribution errors (NPDE) using an add-on package for R [50,51]. Finally, a bootstrap approach was used to investigate the robustness of the final model [52]. One thousand bootstrap data sets were generated and subjected to population pharmacokinetic modelling. The mean values of each set of parameters were then compared with those obtained with the final model.

Once the final population pharmacokinetic model was available, individual MTX concentrations that had not been actually measured at 72 h were estimated for each patient using the post hoc function of nonmem.

Statistical analysis

R v2.14.1 (R Foundation for Statistical Computing, Vienna, Austria) and SAS v9.2 (SAS Institute, Cary, North Carolina, USA) software were used for statistical analysis.

Scatter plots were used to evaluate the association between the basal UCP I/(I + III) ratio and individual MTX clearance. Spearman's linear coefficient of correlation was estimated with its 95% confidence interval.

The values of the UCP I/(I + III) ratio at different sampling times (P1, P2 and P3) were compared by anova.

A non-parametric Kruskal–Wallis rank sum test was used to compare the basal value of the UCP I/(I + III) ratio in the three ABCC2 genotype groups.

Wilcoxon tests were used to compare the basal UCP I/(I + III) ratio according to the presence of a putatively interacting drug. Comparisons were made between patients with at least one score 2 drug and no score 2 drug, and between patients with at least one score 1 drug (without score 2 drug) and only score 0 drugs.

Logistic models were used to study associations between the basal UCP I/(I + III) ratio and delayed elimination, between DIS and delayed elimination, between DIS and toxicity and between delayed elimination and toxicity.

Results

Data description

Patients and treatment

Eighty-five patients were included in the COMETH study between May 2008 and July 2011. Four patients were excluded, one by the investigator's decision before HD-MTX infusion, two because no P1 urine sample was collected and one because of HD-MTX treatment 2 weeks before inclusion. Demographic, biological and HD-MTX treatment data for the 81 patients included in the analysis are summarized in Table 1.

Table 1.

Mean demographic, biological and treatment characteristics of the patients studied (n = 81)

Parameter	n	Median value	Min-Max
Gender (M/F)	46/35
Age (years)		60.6	18.8–84.6
Body weight (kg)		75	43–110
Height (cm)		170	147–195
Body surface area (m2)		1.86	1.34–2.29
Diagnosis
Primary cerebral B-cell NHL	39
Large B-cell NHL	27
Burkitt-like NHL	5
ALL	8
Others*	2
Serum creatinine (μmol l−1)		74	37–160
Creatinine clearance† (ml min−1.1.73 m−2)		91.6	36.1–257.0
Serum urea (mmol l−1)		6	2.0–15.0
Total protein (g l−1)		63	45–85
Albumin (g l−1)		39	21–46
AST (IU l−1)		21	8–56
ALT (IU l−1)		28	8–188
PAL (IU l−1)		67	24–162
GGT (IU l−1)		43	10–406
LDH (IU l−1)		452	129–1402
Total bilirubin (μmol l−1)		8	2–29
Duration of MTX infusion
3 to 6 h	73
24 h	8
Dose of MTX (mg m−2)		3000	976–8000

^*Nasal NK NHL, follicular NHL.

^†Calculated from the Cockroft & Gault formula.

Co-administered drugs

All co-administered drugs and their ‘drug interaction scores’ are listed in Table 2. Forty of the 81 patients were receiving at least one score 2 drug, including 39 receiving benzimidazole proton pump inhibitors, 32 were receiving at least one score 1 drug but no score 2 drug and nine were receiving only score 0 drugs.

Table 2.

List of treatments received on the day of MTX infusion used to attribute the drug interaction score (DIS)

Score	Drugs	Number of patients	Bibliographic reference for DIS assignement
2	Benzimidazoles	39	[38], [42], [43], [74],
	Omeprazole
	Esomeprazole
	Lanzoprazole
	Pantoprazole
	β-lactamins	3	[36]
	Amoxicillin
	Cotrimoxazole	2	[32]
1	Angiotensin II receptor blockers	12	[47]
	Valsartan
	Irbesartan
	Losartan
	Olmesartan
	Telmisartan
	Anticancer drugs
	Doxorubicin	7	[46]
	Imatinib	3	[33]
	Vincristine, vindesine	28	[45]
	Antiviral drugs
	Valaciclovir	13	[44]
	Tenofovir	1	[37]
	Corticoids	2	[23]
	Dexamethasone
	Hydrocortisone
	Others
	Statins	9	[39]
	Colchicine	1	[35]
	Levothyroxine	3	[40]
	Furosemide	3	[34]
	Spironolactone	1	[41]

ABCC2 polymorphisms

ABCC2 genotypes for the three polymorphisms studied were available for 77 patients (four missing due to DNA sample quality being insufficient). All polymorphisms were in Hardy–Weinberg equilibrium. Minor allele frequencies were consistent with data from the HapMap project [53] (rs717620 – 14.9%, rs2273697 – 22.1% and rs3740066 – 36.4%; see Table 3).

Table 3.

Relationships between the basal value of the UCP I/(I + III) ratio and ABCC2 genotype

SNP	Genotype	Number of patient (%)	MAF (%)	Basal value of the UCP I/(I + III) ratio (%)	P value*
				median [min-max]
c.-24C/T	C/C	55 (71)		21.5 [9.6–41.8]
	C/T	21 (27)	14.9	22.9 [9.0–42.5]	0.83
	T/T	1 (2)		19.1 [NA]
c.1249G/A	G/G	46 (60)		24.0 [9.0–42.5]
	G/A	28 (36)	22.1	20.7 [10.1–35.1]	0.33
	A/A	3 (4)		19.2 [9.6–25.5]
c.3972C/T	C/C	33 (43)		22.4 [10.1–41.8]
	C/T	32 (42)	36.4	21.3 [9.6–41.8]	0.57
	T/T	12 (15)		26.6 [9.0–42.5]

^*Kruskall–Wallis rank sum test.

UCP I/(I + III) ratio

Basal values of the UCP I/(I + III) ratio, i.e. those for the P1 urine sample, were between 9.0% and 42.5% (median 22.5%) with a mean (± SD) value of 22.5 ± 7.9% (Figure 1). The value was not dependent on ABCC2 polymorphism (Table 3) and did not vary according to co-administered drugs (P = 0.08).

Figure 1.

Distribution of the basal values of the UCP I/(I + III) ratio (P1) for the 81 patients

The values of the UCP I/(I + III) ratio (mean ± SD) in urine sampled at the end of MTX infusion (P2) were significantly higher (33.1 ± 13,6%), and those at the end of hospitalization (P3) significantly lower (14.2 ± 8.6%) than basal values (22.5 ± 7.9%) (P < 0.001) (Figure 2). Patients receiving MTX as a 3 h infusion presented a higher UCP I/(I + III) ratio at P2 than those receiving MTX over 24 h (33.1 ± 13.5% vs. 22.3 ± 12.9%, P = 0.02). The MTX concentrations (mean ± SD) at the end of infusion (P2) were 355.5 ± 195.2 μmol l⁻¹ and 31.6 ± 22.1 μmol l⁻¹ for 3 h and 24 h infusions, respectively. A significant correlation was found between DP2 and the MTX concentration at the end of infusion (r² = 11.3%, P = 0.003).

Figure 2.

The UCP I/(I + III) ratio at the three sampling periods: P1, before MTX infusion; P2, at the end of MTX infusion; P3, on patient discharge

Population pharmacokinetic analysis

The results of 363 MTX assays were available for the 81 patients included. Plasma MTX concentrations were measured in two laboratories using two different methods, so we tested the influence of the method used on the residual variability, but this did not improve the model.

The relationship between CL and diverse variables was tested. CL was dependent only on creatinine clearance (P < 0.001), the difference of the UCP I/(I + III) ratio between P3 and P1 (DP3) (P < 0.001) and the presence or absence of at least one score 2 drug (SCO2) (P < 0.001) (Figure 3). CL was not related to the basal value of the UCP I/(I + III) ratio (Figure 4).

Figure 3.

Relationship between DP3 and MTX clearance and between the presence or absence of at least one score 2 drug (SCO2) on the day of MTX infusion and MTX clearance (P < 0.001). (SCO2 = 0 indicates no score 2 drug at the time of the MTX infusion, SCO2 = 1 indicates at least one score 2 drug at the time of the MTX infusion)

Figure 4.

Plot of MTX clearance vs. basal value of the UCP I/(I + III) ratio

The final equation describing the effect of covariates on CL was:

No additional covariates on V₁ or V₂ were found to improve the model significantly. The mean values of parameter estimates were: CL = 6.85 ± 1.76 l h⁻¹, V₁ = 24.6 ± 4.9 l, V₂ = 3.05 ± 0.91 l and Q = 0.13 l h⁻¹. MTX_CL did not differ between patients receiving MTX over 3 and 24 h (6.67 ± 1.72 l h⁻¹ vs. 7.80 ± 1.68 l h⁻¹, respectively, P = 0.08).

Diagnostic plots and NPDE distribution (global adjusted P value = 0.210) confirmed the satisfactory quality of this model (Figure 5). Table 4 summarizes the final parameter estimates and bootstrap results expressed as means and 95% confidence intervals of parameter estimates.

Figure 5.

Validation plots of the POP-PK model: (A) log observed concentrations vs. log individual predicted concentrations (DV vs. IPRED); (B) distribution of npde (normalized predictive distribution errors); (C) npde vs. log predicted concentrations; (D) npde vs. time

Table 4.

Final parameter estimates

Parameter	Final model		Bootstrap*
	Mean value	s.e.	Mean	95% CI
† (l h−1)	6.23 (SCO2 = 1)	1.48	–	–
	7.46 (SCO2 = 0)	1.81
θ1	7.05	0.47	7.07	6.07; 8.08
θ2	0.27	0.08	0.27	0.10; 0.45
θ3	0.16	0.07	0.17	0.03; 0.31
θ4	0.93	0.39	0.95	0.20; 1.70
V1 (l)	23.5	1.81	23.2	19.3; 27.0
V2 (l)	3.01	0.42	3.05	2.10; 4.01
Q (l h−1)	0.13	0.02	0.13	0.08; 0.18
Inter-individual variability (%)
ωCL	23.0	10.8	22.1	16.5; 27.7
ωV1	34.0	28.2	33.8	5.80; 61.7
ωV2	32.1	13.8	33.8	24.4; 39.0
Residual variability (%)
σ	41.7	12.7	41.6	37.5; 45.7

^*From 768 successful bootstrap replicates (1000 runs programmed).

^†SCO2 = 1 for patients receiving at least one score 2 drug during MTX infusion SCO2 = 0 for patients not receiving any score 2 drug during MTX infusion.

Simulated or observed MTX concentrations at 72 h were in the range 0.03 to 1.36 μmol l⁻¹ (mean = 0.15 μmol l⁻¹; SD = 0.19 μmol l⁻¹) such that 21 patients were classified as having a delayed elimination of MTX.

Clinical results

Fifty-one patients (62.9%) presented with at least one toxicity (hepatic, renal, hematological or mucocutaneous toxicity) and 27 (33.3%) with at least one severe toxicity (grade > 3–4).

We confirmed that the patients with delayed elimination were at higher risk of at least one toxicity [OR = 4.9, 95% CI: (1.3; 18.4), P = 0.02] and at least one severe toxicity (grade 3–4) [OR = 3.0, 95% CI: (1.08; 8.47), P = 0.04].

The occurrence of toxicity (all grades) was not dependent on basal values of the UCP I/(I + III) ratio, DP3 factor or drug interaction score (P = 0.63, P = 0.27, P = 0.10 respectively). Similarly, these variables did not explain delayed elimination of MTX (P = 0.74, P = 0.94, P = 0.19, respectively, by comparison with normal elimination).

Discussion

The range of basal values of the UCP I/(I + III) ratio (9.0 to 42.5%) found in our population was similar to that reported by previous studies, including ours in healthy volunteers [26,54,55]. However, the distribution was different, with more values below 30% than found in healthy subjects. Our previous study of 74 healthy volunteers suggested that the UCP I/(I + III) ratio was dependent on the c.3972C/T polymorphism of ABCC2 [26], but, by contrast, we found no such association between the ratio and three polymorphisms known to affect MRP2 function [29,56] in the present study. Impaired porphyrin metabolism has been described in patients with malignancies, leading in most cases to an increase in the urinary excretion of porphyrins, and in particular coproporphyrins [57,58]. This suggests, retrospectively, that a population of patients with haematological malignancies was perhaps not the most appropriate to test our hypothesis.

An interesting finding is that the UCP I/(I + III) ratio changed following MTX infusion. Indeed, at the end of the MTX infusion when MTX concentrations were at their maximum the ratio was higher than at baseline, and at P3 when the MTX was completely eliminated the ratio was lower than at baseline. This means that the UCP I/(I + III) ratio is sensitive to the effect of substances interfering with MRP2, MTX being a substrate for this transporter [59]. Our results also indicate that this effect is concentration dependent as the higher the concentrations of MTX, the greater the effect on DP2. By analogy with the increase of the ratio in patients presenting with Dubin–Johnson syndrome, an increase of the UCP I/(I + III) ratio after MTX infusion may reflect inhibition of MRP2. Conversely, the lower UCP I/(I + III) ratio at P3 may reflect an increased activity of MRP2, perhaps as a result of an up-regulation of the transporter to eliminate MTX. Thus, MTX seems to modify a phenomenon, as yet unidentified, that is reflected by the UCP I/(I + III) ratio.

Consistent with previous studies, we observed a wide inter-subject variability in HD-MTX pharmacokinetics [1,3,19,38,60]. The data were best described by a two compartment model and mean values of our final parameter estimates were in agreement with those reported previously [1,2,60]. We found no relationship between MTX clearance and the basal value of the UCP I/(I + III) ratio, and this finding does not support our hypothesis. However, MTX_CL was associated with creatinine clearance, the presence or absence of at least one score 2 drug at the time of MTX infusion and the change of the UCP I/(I + III) ratio between P1 and P3 (DP3). Patients who had the largest decrease in the UCP I/(I + III) ratio between P1 and P3 had a significantly higher MTX_CL than those with a stable or slightly increased ratio (r = −0.346, P < 0.001). We also found that MTX_CL was dependent on the presence of at least one score 2 drug, including in particular benzimidazole proton pump inhibitors (PPIs). Interaction between MTX and PPIs has been described [38,43] and attributed to the inhibition of BCRP/ABCG2 by benzimidazoles [11,61]. However, our study is uninformative about which transporters are predominantly affected by the interaction.

After HD-MTX infusion in humans, MTX and its major metabolite 7-OH-MTX, are mainly eliminated by the kidney [2,5]. MTX secretion depends on its coordinated transport from blood to urine, through tubular cells. OAT1 and OAT3 are the two transporters most involved in MTX uptake on the basal side of tubular proximal cells and are thus central to MTX elimination. They are involved in most of the clinically relevant drug–drug interactions, notably those involving non-steroidal anti-inflammatory drugs (NSAIDS), probenecid or β-lactamins. Recent studies indicate that the hepatocellular uptake transporter OATP1B1 may also contribute to MTX elimination. Indeed, genome-wide and candidate gene approaches both confirmed that variants of SLCO1B1 are significantly associated with MTX kinetics [13–16]. Efflux transporters, such as MRP2, expressed on the apical side of both tubular cells and hepatocytes, may also limit MTX elimination as they are present in the two organs involved in this process. In vitro transport studies using membrane vesicles [59,62] and those using polarized cells overexpressing MRP2 [10,63,64] revealed that MTX is a substrate for MRP2. Several in vivo studies in rats lacking Mrp2 (EHBR or TR- rats) [34,65] and in Mrp2 knock-out mice [21,66–68] confirmed the major role of this transporter in MTX elimination. A mutation in ABCC2 was discovered in a patient presenting with severe MTX overexposure [69]. Also, two studies have shown a relationship between pharmacokinetic parameters of MTX and the c.-24C/T polymorphism of ABCC2 [18,19] and, recently, the c.3972C/T SNP, as part of a haplotype, was found to be associated with increased MTX plasma concentrations [20]. Thus, overall, our results and published data support the idea that MRP2 is central to the behaviour of MTX in man.

In the absence of urinary MTX concentration data, it is difficult to determine the relative contributions of secretion, and filtration and reabsorption to overall MTX clearance. Nevertheless, from the relationship CL_renal = CL_filtration + CL_secretion – CL_reabsorption (where CL_filtration = f_u (unbound fraction) x GFR, and assuming that CL_reabsortion is negligible for MTX), it is possible to estimate the theoretical effects of inhibiting tubular secretion on MTX clearance. We therefore did so using our results and pharmacokinetic data available in the literature. Using values for renal clearance (CL_renal) of MTX of about 90 ml min⁻¹ [2], the unbound fraction of MTX in plasma of 0.48 [70–72] and the GFR of 120 ml min⁻¹, the theoretical tubular secretion rate (CL_Secretion) is about 32 ml min⁻¹. This is equivalent to only about a third of the renal clearance and a quarter of the total clearance (130 ml min⁻¹ in our study) of MTX. This implies that transporter-mediated drug interactions, or any other factor able to modify transporter function, can indeed be major contributors to MTX pharmacokinetic variability.

Although the liver is responsible for only 10 to 30% of MTX elimination in humans [4,5], several studies have shown that polymorphisms in the SLCO1B1 gene, encoding the basolateral hepatocellular transporter OATP1B1, are significantly associated with reduced MTX clearance [13–16]. This led to suggestions that hepatic uptake via OATP1A1 may be the limiting step for MTX excretion However, although these genome-wide, or in some cases more restrictive, association studies have allowed the identification of novel pathways, they can still miss important parts of the problem. For example, Lopez-Lopez et al. found that a SNP in ABCC2, which was not included in the previous GWAS studies, was more significant than the SNPs previously identified in SLCO1B1 [20]. Our study was not sufficiently powered to study the influence of ABCC2 SNPs on MTX_CL and we did not attempt to perform any such analysis. However, by using an indirect method, we were able to show that MRP2 probably contributes to MTX elimination in humans.

Interestingly, our observation that MTX clearance was associated with changes in the UCP I/(I + III) ratio also suggests that MTX may influence its own elimination by modulating transporter activity. This is consistent with, and may provide an explanation for, the non-linearity of MTX pharmacokinetics when administered at very different doses. However, confirmation is required. In our study, although MTX concentrations affected the UCP I/(I + III) ratio at the end of infusion, MTX_CL did not differ between patients receiving 3 h or 24 h infusions. The difference in concentrations, albeit significant, may have been too small to lead to non-linear behaviour in this particular setting. The small number of patients in our study receiving a 24 h infusion (n = 8) may also explain this result. Indeed, one of the most relevant factors in methotrexate pharmacokinetic variability is the dosing regimen. The initial observation of a dose-dependent elimination of MTX was by Relling et al. [73] and was recently confirmed by Trevino et al. [15] who found that the treatment regimen accounted for 17.9% of the variation in MTX clearance. In their GWAS study, Ramsey et al. found that MTX_CL was lower for short (4 h) than long (24 h) durations of infusion, and that the contribution of the treatment regimen was larger than that of SLCO1B1 polymorphisms [16]. Our results may provide clues about the pharmacological basis for this phenomenon. The relationship we found between MTX concentrations and the UCP I/(I + III) ratio at the end of infusion is consistent with the observation that the higher concentrations associated with 4 h than 24 h infusions cause stronger ‘inhibition’ of transporter function, resulting in lower clearance. Indeed, the apparent effect of the duration of infusion may be mediated by associated differences in the concentration rather than, or in addition to, the duration itself. It may be informative to study this phenomenon in the context of the very different doses administered when MTX is used as an anticancer drug or as an immunomodulating agent.

Time-dependent MTX elimination has been suggested in a recent communication [74].

If MTX modulates its own renal transport, other environmental factors may also do so. MTX toxicity is unpredictable, occurring during some courses and not others. It would therefore be interesting to record not only drugs co-administered with HD-MTX, but also any other substance that may modify transporter activity. For example, flavonoids, increasingly used in food supplements as natural antioxidants, such as quercetin and lutolein (present in fruits, vegetables and in many drinks), and components of green tea, are inhibitors of MRP2 [40,75] and should be studied as potentially interfering substances.

Competing Interests

All authors have completed the ICMJE uniform disclosure form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.