Significant matrix effect in fludarabine quantification in plasma: implications for CAR-T cell therapy monitoring

ABSTRACT

Background

Event-free survival (EFS) in patients with B-cell acute lymphoblastic leukemia (B-ALL) significantly improves with adequate exposure (as Area Under the Curve: AUC_0-inf or AUC) to fludarabine (FLU) during lymphodepletion prior to chimeric antigen receptor (CAR) T-cell infusion. A cumulative AUC of FLU exceeding 14 mgxh/L is considered optimal, making therapeutic drug monitoring (TDM) crucial for personalized dosing and prolonged CAR T-cell persistence.

Research design and methods

We developed and validated an European Medicines Agency (EMA)-compliant high-performance liquid chromatography-tandem mass spectrometry method (HPLC-MS/MS) for FLU TDM, based on similar and published methods, to support clinicians during the patient conditioning phase. All EMA acceptance criteria were met.

Results

Analysis of real plasma samples revealed unexpectedly high FLU concentrations and, consequently, an excessively high AUC. Analysis of plasma samples from a child receiving FLU therapy highlighted the necessity for accurate drug response verification in real matrices. Preparation of quality controls in FLU-free plasma consistently failed without preliminary dilution. Analysis of internal standard range variation enabled more precise and accurate analyses.

Conclusions

This experience highlights the importance of complementing standard validation tests with analyses in each patient’s plasma, collected immediately before FLU infusion, to ensure reliable TDM to support CAR T-cell therapy.

Plain Language Summary

The success of a special treatment called CAR-T for people with leukemia depends on a few things. One of them is the amount of a drug called Fludarabine they received a few days beforehand. Patients with leukemia need several doses of a drug called Fludarabine. The right amount to give them depends on how much of the drug is already in their blood. To find this out, doctors must take several blood samples, which are then analyzed using special and often complicated lab methods. Getting these methods to work correctly requires skilled people and special equipment. Sometimes, however, hard-to-solve problems come up. These problems must be fixed very quickly and with only a small amount of blood to work with. This article shares a story about how some researchers did their best to handle one of these difficult situations. They believe they found a good solution, even with all the challenges and limited resources they faced.

1. Introduction

Therapeutic drug monitoring (TDM) is of great importance in assisting physicians to tailor therapy to their patients. In recent years, many authors have highlighted the close relationship between a favorable outcome in pediatric patients with B-ALL treated with CAR-T infusion and prior correct cumulative exposure to FLU, given as part of the lymphodepletion regimen [1–4]. The authors agree that the minimum cumulative threshold AUC of FLU, given for four consecutive days at a dose of 30 mg/m²/day infused over 30 min, should be 14 mgxh/L: at lower levels, underexposed patients have a significant increase in cumulative incidence of relapse and reduction in EFS [3]. The definition of the toxicity threshold is still controversial, as Thavayogarajah and colleagues calculated a median AUC of 42.97 mg ×h/L in their 19 Acute Myeloid Leukemia (AML) patients, which is twice as high as the suggested value for an optimal toxicity profile [5,6]. However, none of their patients experienced acute toxicity, at least within 30 days after transplantation.

Based on these considerations, we developed and validated a custom HPLC-MS/MS analytical method according to current European Medicines Agency (EMA) guidelines [7]. Many authors before us have published articles that we considered to be good starting points, evaluating their strengths and weaknesses. For example, the use of a deuterated analog of FLU as an internal standard (IS) ([¹³C,¹⁵N₃]-FLU) is useful to reduce sample handling, matrix effects, and run times, but its cost makes its routine use prohibitive [8]. The alternative use of cladribrine or clofarabine overcomes this limitation but is useless for patients receiving combined treatment with FLU and cladribrine/clofarabine [9,10].

2-Chloroidenosine is also used as IS, but in the cited papers its retention time is different from that of FLU, making it difficult to correct for possible matrix effects, especially in the absence of a standard addition method [11,12]. We found chromatographic conditions that seemed to overcome this problem: retention times for FLU and IS were the same, but quality controls prepared in FLU-free plasma from transplant patients always failed if no prior and appropriate dilution was applied.

2. Methods

2.1. Instrumental

A TSQ Quantum Access was coupled to an Accela pump and an Accela auto-sampler, managed by Excalibur 2.07 and LCquan 2.5.6 software. Chromatographic separations were performed on a HyPURITY 5 µm C18 (150 ×4.6 mm) in gradient elution mode. Mobile phases were CH₃CN with 0.1% NaOH (0.1 M) (mobile phase A) and H₂O acidified with 0.1% HCOOH (mobile phase B). From 0 to 2.0 min, mobile phases were 70:30 (A:B): at 2.1 min until 3.1 min it was 100% A. From 3.2 min to 4.5 min were 70:30 (A:B) again.

Ten microliters of samples were injected into the column and eluted at a flow rate of 1 mL/min. The retention time for FLU and 2-chloroadenosine (IS) was 1.7 min. The ion pairs m/z 286 > 134 and 302 > 134 for FLU and IS were used for quantification in multiple reaction monitoring (MRM) mode (Figure 1).

Figure 1.

Typical chromatogram of FLU (0.8 µg/mL) and IS.

FLU and IS eluted at the same retention time: 1.7 min.

2.2. Preparation of calibrators and quality controls

Two different FLU stock solutions (0.5 mg/mL) were prepared in methanol acidified with HCl (10 mM): six plasma calibration standards were obtained by serial dilution in matrix (3–0.1 µg/mL). Three quality controls (high, medium, and low) were prepared in the same manner: 2.4, 1.2 and 0.3 µg/mL, respectively. FLU-free plasma was obtained from volunteers at the Department of Transfusion Medicine.

2.3. Extraction procedures

Fifty microliters of plasma – for each sample to be analyzed – were added to well-labeled Eppendorf tubes (1.5 mL): samples were then diluted with 25 µL of IS_WS (160 µg/mL in CH₃CN:H₂O = 70:30) and 25 µL of TCA (20%). The samples were then vortexed for a minimum of 15 s.

After centrifugation, 50 µL of the upper supernatant was diluted with 100 µL of mobile phase (CH₃CN:H₂O = 70:30), transferred to the glass auto-sampler vials and analyzed.

2.4. Calibration curve and calculation

The analytical run consisted of a zero sample (plasma blank with IS), six calibrators, and three quality controls (low, medium, and high).

The calibration curve was generated by least squares linear regression (weighted 1/x) of the peak area ratios (analyte/IS) against the nominal calibrator concentrations. The accuracy of the back-calculated concentrations of each calibration standard should be within ±20% of the nominal concentration at the lowest limit of quantification (LLOQ: 0.1 µg/mL) and within ±15% at all the other levels. At least 75% of the six calibration standard levels should meet the above criteria.

Plasma concentrations of patient samples were back calculated using the equation obtained for the calibration curve.

2.5. Patient

As only fully anonymized patient samples were used that were not obtained specifically for use in this study through an interaction or intervention with living individuals, neither informed consent nor an Institutional Review Board (IRB) review was required.

The only patient was a child hospitalized in the Pediatric Hematology and Oncology department of the Fondazione IRCCS Policlinico San Matteo in Pavia, awaiting a CAR-T cell transplant. At the time of his admission, his parents signed the consent for the therapy and the disclosure of scientific data.

We analyzed six consecutive plasma samples from this patient who received FLU at a dose of 30 mg/m² infused over 30 min to calculate the AUC using analytically validated procedures. A two-compartment PK model was used to describe the data.

3. Results and conclusions

Patient plasma concentrations and associated AUC were unexpectedly and unacceptably higher than expected.

We observed that the internal standard (IS) area of all these samples was consistently and significantly lower than that of the calibrators and quality control samples (Table 1), resulting in excessively high back-calculated FLU concentrations and FLU exposure on the first day of therapy (AUC: 20.5 mgxh/L).

Table 1.

Summarizing IS area (a.U.).

Calibrators	1,860,407	1,679,446	1,665,024	1,674,903	1,559,542	1,563,368
QcL, QcM, QcH	1,596,358	1,644,042	1,771,682	/	/	/
Patient plasma samples	78,576	74,772	86,990	77,626	70,524	66,934

We hypothesized a strong matrix effect due to drugs administered to the patients besides FLU, which could not be corrected by IS, even though its retention time was the same as that of FLU.

This issue was resolved by diluting the patient’s FLU-free plasma with plasma from healthy volunteers. The initial patient sample, collected just before the first FLU administration, was diluted at five consecutive ratios: 1:4, 1:5, 1:6, 1:7, and 1:8. After adding the internal standard (IS) to each diluted sample, its corresponding area was analyzed and is reported in Table 2.

Table 2.

IS area (a.U.) after dilution.

Dil. 1:4	Dil. 1:5	Dil. 1:6	Dil. 1:7	Dil. 1:8
892,849	848,897	1,038,381	1,312,779	1,260,694
837,886	1,057,634	1,130,846	1,209,456	1,302,666
884,526	966,851	1,100,483	1,336,583	1,328,381

Contrary to expectations, the internal standard’s area increases upon dilution, with its values becoming comparable to those of the QCs reported in Table 1. This is a behavior characteristic of a matrix effect. The optimal results for comparability were achieved with a 1:8 dilution factor.

We also confirmed our hypotheses by preparing and analyzing quality controls (low: QcL and high: QcH) in the patient’s “zero-time sample” collected just before the start of the infusion.

Quality controls were analyzed either without dilution according to validated analytical procedures or after appropriate dilution with blank plasma (from a healthy volunteer). The dilutions were 1:2, 1:4, and 1:8 (data not reported).

We obtained the best results in the latter scenario (Table 3), in terms of accuracy between calculated mean and nominal concentrations.

Table 3.

Quality controls accuracy in patient’s FLU- free plasma sample.

Qc	L		H
Nominal Conc. (µg/mL)	0.3		2.4
	Undiluted plasma	Diluted plasma (1:8)	Undiluted plasma	Diluted plasma (1:8)
Back-calculated FLU concentrations in samples prepared by adding FLU to patient’s drug – free plasma sample. (µg/mL)	1.505	0.387	15.085	2.218
	1.73	0.243	15.410	2.467
	1.599	0.340	15.754	2.561
Average	1.611	0.323	15.416	2.415
Accuracy (%)	437.1	7.7	542.3	0.6

Following a 1:8 dilution with drug-free plasma, accuracy at the QcL and QcH levels plummeted from 437.1% to 7.8% and 542.3% to 0.6%, respectively, a step not required during standard method validation.

In conclusion, a good validation of any analytical method is mandatory before approaching any TDM; the calculation of cumulative exposure to FLU is a new example recently introduced to improve EFS and reduce relapses in B-ALL patients undergoing CAR-T treatment.

We would like to emphasize the importance of the matrix effect analysis, not only in the matrices suggested by the EMA guidelines but also in patient plasma whenever possible. The presence of other drugs and/or excipients can affect the analyte and its IS in unpredictable ways, even if their retention time is the same.

Appropriate dilution with drug-free plasma prior to analysis is often a winning choice.

Article highlights

• One of the causes of variable outcomes after CAR T-cell therapy is lymphodepletion due to the variable and insufficient fludarabine exposure

• Fludarabine therapeutic drug monitoring (TDM), based on a reliable therapeutic monitoring, is mandatory

• Unpredictable matrix effects could affect the analysis even when the analyte and the IS elute at the same retention time

• Dilution of real samples before analysis, with the same drug-free matrix, is often a good choice

• FLU level results in real plasma samples should always be critically analyzed, even when they are obtained with analytical methods validated according to the most recent guidelines

• Although it is straightforward to find the best solution for a well-defined problem when time and resources are plentiful, an unpredictable event sometimes demands an immediate response. In these situations, we must quickly support physicians – often within a matter of hours – to ensure the best possible outcome for the patient

Funding Statement

This paper was not funded.

Disclosure statment

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.

Author contributions

Simona De Gregori: conceptualization – methodology – validation – data curation – writing original draft – visualization.

Mara Capone: investigation – data curation – visualization.

Marco Zecca: writing original draft – supervision.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Ethical Declaration

Because only fully anonymized patient samples were used and were not obtained specifically for this study through an interaction or intervention with living individuals, neither informed consent nor an Institutional Review Board (IRB) review was required. The single patient was a child from the Pediatric Hematology and Oncology department at Fondazione IRCCS Policlinico San Matteo in Pavia, who received all necessary treatment while awaiting a CAR-T cell transplant, in accordance with Good Clinical Practice guidelines. At the time of the child’s admission, the parents signed a consent form for the therapy and for the disclosure of scientific data.

We declare that our study follows the principles of the Declaration of Helsinki.