Validation of an UHPLC–MS/MS Assay for the Quantification of Isoniazid and Four Major Metabolites in Human Plasma

ABSTRACT

Isoniazid (INH) is a key component of tuberculosis treatment regimens but is also associated with hepatotoxicity. This toxicity is mediated in part by its metabolites, but some degrade rapidly in plasma. A rapid freeze/thaw process with a methanol protein precipitation extraction was developed to slow the rapid degradation of acetylisoniazid (AcINH) into isonicotinic acid (INA) and acetylhydrazine (AcHZ) in human plasma. An assay consisting of two UHPLC–MS/MS methods was developed and validated to determine concentrations of INH and four of its metabolites in human plasma: the “polar” method covering AcINH and INA, and the “nonpolar” method covering p‐tolualdehyde derivatized AcHZ, hydrazine (HZ), and INH. Polar and nonpolar method chromatographic separation was accomplished with biphenyl or C18 columns, respectively, mobile phases of 4 mM ammonium formate in water or acetonitrile, and method specific gradient conditions. The validated range for AcINH, INA, and INH was 10.0 to 5000 ng/mL and for AcHZ and HZ was 10.0 to 500 ng/mL from 0.1 mL of plasma, respectively. Though other groups have validated methods to quantitate some or all of these analytes, none have adequately characterized or addressed AcINH degradation. This validated assay will help facilitate a better understanding of INH metabolite disposition and relationships with toxicity.

1. Introduction

Nearly a quarter of the world is estimated to have latent tuberculosis infection (Cohen et al. 2019). Without treatment, around 5% to 10% of people will progress to active tuberculosis infection, and this risk is increased several‐fold in persons with HIV (World Health Organization 2025a). Isoniazid (INH) is a key component of multiple first‐line treatment options for both latent and active tuberculosis infection (World Health Organization 2025b). Despite the efficacy of INH in various treatment and prophylaxis regimens, it is associated with drug‐induced liver injury (DILI), limiting its clinical utility (Badrinath et al. 2026; Batool et al. 2026; Mahajan and Tyagi 2024; Metushi et al. 2016; Prakongsup and Udomsinprasert 2026).

Though the development of DILI with INH is multifactorial, it is believed to be mediated in part by the formation of reactive metabolites and immune‐mediated responses (Batool et al. 2026; Mahajan and Tyagi 2024; Prakongsup and Udomsinprasert 2026). INH metabolism is complex, resulting in several identified metabolites, including acetylisoniazid (AcINH), isonicotinic acid (INA), acetylhydrazine (AcHZ), and hydrazine (HZ) (Figure 1). These metabolites, particularly AcHZ and HZ, can ultimately be converted into reactive oxygen species, which can lead to hepatotoxicity (Badrinath et al. 2026; Blair et al. 1985; Gent et al. 1992; Metushi et al. 2016; Mitchell et al. 1976; Nelson et al. 1976; Sarich et al. 1996; Timbrell et al. 1980; World Health Organization 2025a), necessitating a more accurate and comprehensive characterization of multiple metabolites to better understand exposure–response relationships in people requiring INH therapy.

FIGURE 1.

Schematic of isoniazid metabolism. Acetylation reactions are catalyzed by N‐acetyltransferase type 2 (NAT2), and enzymatic hydrolysis reactions are catalyzed by amidases. Oxidation reactions are catalyzed by cytochrome P450 2E1 (CYP2E1). Chemical structures obtained from MolView.org.

Though other LC–MS/MS methods have been published for the determination of INH and/or its metabolites in human or rat plasma (Chen et al. 2005; Hee et al. 2015; Huang et al. 2009; Ky Anh et al. 2022; Ng et al. 2007; Song et al. 2017), the breakdown of the parent drug and metabolites may not have been adequately characterized. Instability of INH in storage at −20°C for 6 h (Ng et al. 2007), at 4°C or room temperature for up to 20 h (Huang et al. 2009), and of both INH and INA after three freeze/thaw cycles (Ky Anh et al. 2022) have been reported. Only three methods performed stability tests with INH and the relevant metabolites for their assays, though these were only performed with quality controls (QCs) utilizing similar concentrations between analytes (i.e., all analytes at quality low [QL] or high [QH] level, respectively). Stability tests utilizing solutions with similar concentrations of all analytes may not accurately reflect clinical samples from people on INH therapy, where dramatically different analyte concentrations within a single sample may exist. Song et al. quantitated only AcHZ and HZ and did not perform stability tests, while Chen et al. and Huang et al. only quantitated INH, so no analysis of breakdown between INH and its metabolites was performed. As none of these assays investigated the impacts of individual analyte degradation into other analytes of interest, potentially significant contributions from the breakdown of a high concentration analyte to lower concentration ones in the same sample may have been missed.

Here we detail the development and validation of an ultra‐high performance liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) assay consisting of two methods—the “polar method” and the “nonpolar method”—for the determination of multiple analytes utilizing stable labeled internal standards. The polar method was used for quantitation of INA and AcINH, and the nonpolar method was used for the quantitation of p‐tolualdehyde derivatized AcHZ, HZ, and INH. During assay development and validation, the instability of AcINH was evaluated, and a novel rapid freeze/thaw method was developed to preserve INA and AcHZ concentrations.

2. Experimental

2.1. Chemicals and Reagents

Reference standards for all compounds (INH, AcINH, INA, AcHZ, and HZ) and their deuterated or stable labeled internal standards (INH‐IS: D₄, AcINH‐IS: D₄, INA‐IS: D₄, AcHZ‐IS: D₃, HZ‐IS: ¹⁵N₂) were obtained from LGC Standards (Toronto, Canada), except for the HZ‐IS, which was obtained from Sigma‐Aldrich (St. Louis, Missouri). Acetonitrile (MeCN), methanol (MeOH), 2‐propanol (IPA), ammonium hydroxide (NH₄OH), and formic acid (FA) were all purchased from Fisher Scientific (Fairlawn, New Jersey). Organic solvents (MeCN, MeOH, and IPA) were of HPLC or Optima grade while the NH₄OH and FA were of Certified A.C.S. Grade. p‐Tolualdehyde was purchased from Sigma‐Aldrich (St. Louis, Missouri). Ultrapure water was sourced from an in‐house Barnstead GenPure Pro UV/UF water filtration system. Human K₂ EDTA plasma was sourced from BioIVT (Hicksville, New York).

2.2. Preparation of Calibrators and Quality Controls

Two separate preparation stock‐1 (PS1) solutions were made at 1.0 mg/mL in MeOH for each compound except for INA, which was prepared in water. One PS1 solution was utilized for calibrator preparation and the other for quality control (QC) sample preparation. Two sets of calibration solutions were prepared by diluting PS1 solutions in matrix to the ULOQ level and subsequent dilutions of the ULOQ for the remaining levels. Prepared calibrator solutions include spiking calibrators in MeOH containing AcINH, INA, and INH for each calibrator level, and calibrators in plasma containing AcHZ and HZ at each calibrator level (plasma percentage > 99%). Concentrations of each calibrator are presented in Table 1.

TABLE 1.

Calibrator and quality control concentrations for polar and nonpolar analytes.

Calibrators	INH, AcINH, INA (ng/mL)	AcHZ, HZ (ng/mL)	Quality control	INH, AcINH, INA (ng/mL)	AcHZ, HZ (ng/mL)
H	10.0	10.0	LLOQ	10.0	10.0
G	25.0	15.0
F	50.0	25.0	QL	30.0	30.0
E	100	50.0
D	500	75.0	QM	800	80.0
C	1000	100
B	2500	250	QH	4000	400
A	5000	500

Abbreviations: AcHZ, acetylhydrazine; AcINH, acetylisoniazid; HZ, hydrazine; INA, isonicotinic acid; INH, isoniazid; LLOQ, lower limit of quantitation; QH, high level QC; QL, low level quality control (QC); QM, mid‐level QC.

QC spiking solutions were prepared for each QC level in methanol. QC solutions in plasma were prepared, aliquoted, and rapidly frozen on a pre‐chilled freezing block on dry ice at the lower limit of quantitation (LLOQ), QL (3× LLOQ), QM, and QH (80% upper limit of quantitation (ULOQ)) (Table 1). Two dilution QC (dilQC) solutions were prepared in plasma at five times the QH concentration (one containing AcINH and INA, and one containing INH, AcHZ, and HZ). A third dilQC was prepared at 10 times the QH concentration containing AcHZ and HZ. All aliquots were immediately placed into a pre‐chilled freezing block on dry ice.

A separate plasma sample containing only AcINH at the assay ULOQ was prepared to monitor maximum likely AcINH degradation in plasma through sample handling and extraction.

Internal standard PS1 solutions were also prepared at approximately 1 mg/mL for each IS in MeOH, except for INA‐IS, which was prepared at approximately 0.5 mg/mL in 50:50 MeOH:water. A spiking‐IS solution was prepared at 2500 ng/mL for INH‐IS and AcINH‐IS, 1250 ng/mL for INA‐IS, and at 1000 ng/mL for AcHZ‐IS and HZ‐IS by diluting the IS PS1 solutions into MeOH.

Methanol and water‐based solutions were stored in 13 × 100 glass tubes with PTFE‐lined caps at −20°C as standard laboratory practice. The plasma‐based solutions were stored in 2 mL plastic cryovials stored at −80°C.

2.3. Sample Preparation

Prior to extraction, all QC and study samples were removed from storage at −80°C and kept frozen on dry ice. Pre‐prepared plasma and spiking calibrator and IS solutions were removed from storage (−80°C and −20°C for plasma and spiking solutions, respectively) and allowed to come to room temperature prior to extraction. A total of 20 μL MeOH was added to blank sample tubes, and 10 μL MeOH was added to blank with internal standard (blank‐IS), QC, and study sample tubes. Next, 10 μL of spiking‐IS solution was added to all tubes, excluding those for blank samples. To the blank and blank‐IS tubes, 100 μL of blank plasma was added. To the calibrator tubes, 10 μL of spiking calibrator and 100 μL of plasma calibrator were added. To facilitate protein precipitation, 400 μL MeOH was added to all tubes, and blank, blank‐IS, and calibrator samples were briefly vortex mixed.

QC and study samples were placed one at a time into a pre‐heated (50°C) dry heat block until thawed (about 3 to 7 min). Each sample was then briefly vortex‐mixed, and 100 μL was transferred to the appropriate microcentrifuge tube. The thawed sample tube was then immediately placed into a freezing block pre‐chilled on dry ice to allow it to rapidly freeze. Each extraction tube was then closed and briefly vortex mixed to facilitate protein precipitation.

Once all plasma samples were precipitated, tubes were spun down in a microcentrifuge at 13,000 × g for 5 min. Samples were then removed from the microcentrifuge, and each sample supernatant was split into two 200 μL aliquots. One set of aliquots was used for the polar method and one for the nonpolar method. The aliquots for the polar method were dried under nitrogen in a nitrogen evaporative dryer and reconstituted in 300 μL of mobile phase A (4 mM ammonium formate in water). The aliquots for the nonpolar method were derivatized by adding 50 μL of 12% p‐tolualdehyde and placed in an ultrasonic bath for 30 min and were then diluted with 250 μL of mobile phase A. All samples were transferred to autosampler vials with low volume inserts, and 3 μL of sample was injected onto the system for each method. Generally, the polar method was run immediately after extraction, and the nonpolar method was run the following day after the polar method completion.

2.4. Instrumentation

UHPLC–MS/MS analysis of human plasma samples was achieved using a Thermo Scientific Vanquish Horizon UHPLC system with a Thermo Scientific Accucore Biphenyl column (2.1 × 100 mm, 2.6 μm) for the polar method and a Waters Acquity UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm) for the nonpolar method. Detection was achieved using a TSQ Altis Plus tandem mass spectrometer equipped with an OptaMax NG ion source and H‐electrospray ionization (ESI) insert (Thermo Scientific, Fairlawn, New Jersey). Data collection and processing were performed using the Chromeloen V7 software (Thermo Scientific, Fairlawn, NJ).

2.5. UHPLC–MS/MS Settings

Both methods utilized the same mobile phase A and mobile phase B (4 mM ammonium formate in MeCN) as well as maintained the autosampler and column temperatures at 15°C and 40°C, respectively.

The MS system was operated in the positive selected‐reaction monitoring (SRM) scan mode with the following settings: positive ion spray voltage: 500 V, sheath gas: 50 arbitrary units, auxiliary gas: 10 arbitrary units, sweep gas: 0 arbitrary units, ion transfer tube temperature: 250°C, vaporizer temperature: 400°C, chrom. Peak width: 6.0 s, dwell time: 50 ms, Q1 resolution (FWHM): 0.7, Q3 resolution (FWHM): 0.7, CID gas: 1.5 mTorr, and source fragmentation: 0. Analyte precursor and product m/z values, collision energies, and RF lens settings are presented in Table 2.

TABLE 2.

Analyte‐specific MS system settings in positive mode. AcHZ, HZ, and INH were all derivatized with p‐tolualdehyde; masses shown are derivatized masses.

Compound	Precursor (m/z)	Product (m/z)	ce (V)	RF lens (V)
AcINH	180.033	121	24	41
AcINH‐IS	184.133	125	24	42
INA	124.067	80	23	50
INA‐IS	128.067	84	23	41
AcHZ	177.050	118	15	36
AcHZ‐IS	180.133	118	15	36
HZ	237.167	120	19	57
HZ‐IS	239.100	121	19	49
INH	240.167	80	23	47
INH‐IS	244.133	84	23	50

Abbreviations: AcHZ, acetylhydrazine; AcINH, acetylisoniazid; CE, collision energy; HZ, hydrazine; INA, isonicotinic acid; INH, isoniazid; IS, internal standard.

2.5.1. Polar UHPLC–MS/MS Settings

To accomplish chromatographic separation of AcINH and INA, a Thermo Scientific Accucore Biphenyl column (2.1 × 100 mm, 2.6 μm) with a Phenomenex SecurityGuard ULTRA Cartridge (UHPLC Polar C18 for 2.1 mm ID columns) and a Phenomenex KrudKatcher UHPLC In‐Line Filter was used with a 10‐min‐long stepwise gradient at 0.4 mL/min (Table 3).

TABLE 3.

UHPLC gradient programs for polar and nonpolar methods.

Polar method
Time min	Flow mL/min	%MPA	%MPB	Curve
Initial	0.400	100	0	5
1.75	0.400	100	0	5
1.76	0.400	20	80	5
2.75	0.400	20	80	5
2.76	0.400	100	0	5
10.00	0.400	100	0	5

Nonpolar method
Time min	Flow mL/min	%A	%B	Curve
Initial	0.600	35	65	5
1.00	0.600	35	65	5
1.01	0.600	10	90	5
2.50	0.600	10	90	5
2.51	0.600	35	65	5
4.00	0.600	35	65	5

2.5.2. Nonpolar UHPLC–MS/MS Settings

To accomplish chromatographic separation of INH, AcHZ, and HZ, an Acquity UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm) with an Acquity UPLC Column In‐Line Filter Unit was used with a 4‐min‐long stepwise gradient at 0.6 mL/min (Table 3).

2.6. Method Validation

This assay was developed and validated according to established International Council for Harmonization (ICH) bioanalytical assay validation guidelines, except for matrix effect, validated according to the U.S. FDA guidelines (International Council for Harmonisation (ICH) 2022; Matuszewski et al. 2003; U.S. Food and Drug Administration 2018). All presented tests were performed for both the polar and nonpolar methods. Test solution results were compared with control solution results and nominal concentrations, where appropriate, to show the effect of the test condition and to determine if the test solution would fall within acceptance criteria.

Matrix selectivity was investigated by assessing analyte and internal standard peak areas in blank K₂ ETDA plasma obtained from six donors. Selectivity was also assessed by investigating the effects of 1.25%, 2.5%, and 5% hemolyzed plasma and lipemic plasma (triglycerides: 557 mg/dL) on QL and QH prepared in the respective matrix and compared with nominal concentrations and with control QL and QH samples.

Specificity was investigated in QL and QH samples spiked with clinically relevant concentrations of potential concomitant medications (lamivudine, emtricitabine, tenofovir, zidovudine, nevirapine, darunavir, atazanavir, ritonavir, lopinavir, and efavirenz). Test QC concentrations were compared with nominal concentrations and with control QL and QH samples.

Carry‐over was determined by injecting an extracted blank sample after the analysis of an ULOQ sample, and crosstalk was determined by analysis of a blank‐IS sample and an ULOQ without internal standard. Signals were compared with respective analyte and IS signals from the LLOQ and blank‐IS, respectively.

Intra‐ and inter‐assay accuracy and precision were established in at least three analytical runs. Each run included a calibration curve (calibrator H—calibrator A) and six replicates of each QC level (LLOQ, QL, QM, and QH).

Dilution integrity was assessed by analyzing five replicates of diluted polar or nonpolar dilQC and partial volume accuracy and precision, which was evaluated in case there were samples with insufficient volume to extract 0.1 mL, was determined by analyzing five replicates of diluted QH. Dilution factors of 5× (all analytes, dilution and partial volume) and 10× (AcHZ and HZ, dilution) were tested.

Matrix effects (ME), recovery (RE), and process efficiency (PE) were investigated using a previously described method (Matuszewski et al. 2003). The lack of a significant ME was assessed by performing a slopes analysis resulting from a three‐point curve (calibrators H, E, and B) prepared in plasma from six separate donors. ME, RE, and PE were additionally assessed for each analyte and IS by comparing responses from neat samples, post‐extract spiked samples, and extracted samples at low, mid, and high concentrations.

Stability in matrix and in organic solvents was characterized extensively throughout validation. Bench‐top stability was assessed for QL and QH at room temperature up to 30 min, and freeze–thaw stability was assessed over three rapid freeze–thaw cycles with a heat block heated to 50°C. AcINH stability in plasma at room temperature was assessed up to 30 min. Reinjection reproducibility was assessed by re‐injecting a calibration curve and triplicate QL and QH samples after four (polar method) or three (nonpolar method) days in the autosampler after initial injection. Reinjected QC data was processed on the reinjected curve and on the originally injected calibration curve. Test samples or solutions were compared with nominal concentration and/or to control samples. Long‐term stability of the full range of concentrations of each analyte in all matrices was assessed by comparing results obtained from stored solutions to those obtained from freshly prepared solutions and to nominal concentrations for solutions in matrix from the following solutions: PS1, spiking QL and QH, spiking calibrators H and A, plasma calibrators H and A, and QL and QH in matrix.

Tests were also performed to assess the effects of the rapid freeze/thaw process and to assess whole blood stability by keeping blood tubes on wet ice versus at room temperature for various durations prior to plasma processing. Additionally, the reproducibility of sample results was assessed by incurred sample re‐analysis (ISR) of clinical study samples.

3. Results

3.1. Assay Development

3.1.1. Selection of Instrumentation

The increased sensitivity of the Thermo Scientific TSQ Altis Plus mass spectrometer compared with the Thermo Scientific TSQ Vantage and Quantiva MS systems and the Waters TQ‐S Micro MS system, which were also tested, resulted in improved signal‐to‐noise for each analyte. This improved signal‐to‐noise allowed for improved reproducibility of sample results near the assay LLOQ.

3.1.2. Single vs. Split Polar and Nonpolar Methods

Initial work was performed to try to quantify all analytes with a single method; however, reverse‐phase chromatographic separation of all five analytes was not feasible with a single column in a reasonable timeframe. Early development work for both methods was performed using a 50 × 2.1 mm Accucore Biphenyl column; however, this column was unable to achieve baseline separation of AcHZ and INH, resulted in ion suppression of AcHZ at the retention time of the peak INH signal, and assay performance was generally poor. Achieving baseline separation of AcHZ and INH with the 100 × 2.1 mm HSS T3 column led to improved performance. Additionally, crosstalk was observed from INH to INA and from Ac‐INH into INA and INH, potentially due to source fragmentation. Consistent chromatographic separation of these analytes was achieved by switching the mobile phase from containing formic acid to 4 mm ammonium formate and by switching from a 50 × 2.1 mm Accucore Biphenyl column to a 100 × 2.1 mm Accucore Biphenyl column.

3.1.3. Calibrator Stock Preparation in Plasma

Calibrators for INH, AcINH, and INA were prepared in methanol as spiking solutions. When AcHZ and HZ were prepared in methanol, results were variable, so AcHZ and HZ calibration solutions were prepared in plasma, which improved performance.

3.1.4. Optimization of Extraction Techniques

A variety of solid phase extraction cartridges were tested during development, with extensive testing of the Phenomenex Strata X‐A and X‐AW cartridges; however, a protein precipitation extraction ultimately resulted in the best overall signal, reproducibility, and assay performance. Protein precipitations are frequently performed using MeCN; however, MeCN resulted in suppression of AcHZ, INA, and INH. These effects were improved using MeOH.

3.1.5. Rapid Freeze–Thaw Method for AcINH Degradation

Minutes after spiking AcINH into plasma at the ULOQ, significant degradation into INA and AcHZ was observed compared with respective LLOQ signals. To reduce the effects of this degradation, a rapid freeze–thaw method was developed where frozen plasma samples were kept on dry ice prior to extraction and were then rapidly thawed on a dry heat block preheated to 50°C prior to extraction. The number of samples applied to the heat block was managed such that they were utilized upon thaw, minimizing time in the heat block. Once thawed, 100 μL of the plasma sample was pipetted into MeOH to facilitate protein precipitation. Once precipitated, the degradation of AcINH into other analytes was stopped. After pipetting a study sample, it was immediately placed into a dry freezing block pre‐chilled on dry ice to allow the sample to rapidly freeze, which preserved AcHZ and INA concentrations in the plasma sample.

3.1.6. Derivatization

A previously described method to derivatize AcHZ and HZ with p‐tolualdehyde (Song et al. 2017) was used to add phenyl groups to amino groups on the analytes to improve chromatographic retention and detection accuracy on the mass spectrometer. During the development of the nonpolar method, it was observed that the samples did not appear to achieve complete derivatization after the 30‐min sonication period, as results were variable and analyte response increased over time, particularly for HZ. Sonication times up to 1 h were tested; however, the samples were still not fully derivatized. Extracted QL and QH samples were sonicated for 30 min and injected into the system every 2 h from 0 to 14 h and showed peak area ratios (PARs) stabilized after 6 h in the autosampler. As a result, the derivatized nonpolar samples were stored in the autosampler for at least 6 h prior to analysis.

3.1.7. INH Impurity in AcINH Reference Standard

It was observed that the AcINH reference standard contained a very small percentage of INH impurity. This was not an issue as INH was largely undetectable until concentrations near the assay upper limit of quantitation were prepared. This did not influence the combined preparation of plasma calibrators and quality controls for the assay, as the intended INH concentration in these samples overwhelmed those contributed by the observed AcINH impurity. However, the AcINH‐only sample utilized for stability studies did produce an INH signal that was greater than the INH LLOQ response. This signal remained unchanged throughout the stability studies, indicating that AcINH did not contribute to INH signal during stability tests.

3.2. Assay Validation

3.2.1. Selectivity, Specificity, Carry‐Over, and Cross‐Talk

Selectivity was assessed in plasma from six donors. The resulting analyte SRM channels were compared with LLOQ analyte signals and blank‐IS internal standard signals (Figure 2). No significant analyte (> 20% LLOQ response) or internal standard (> 5% IS response) signals were observed, providing evidence of assay selectivity, and no significant signals were observed in a blank sample injected after a ULOQ sample, indicating insignificant carry‐over. Selectivity was also assessed by comparing test QL and QH samples prepared in hemolyzed and lipemic plasma to control QL and QH solutions and to nominal concentrations. There were no effects seen with lipemic plasma compared with control QCs; however, there were significant matrix effects observed for the QL samples for HZ in 5% hemolyzed plasma (36.1% different from control QC, data not presented). The hemolyzed plasma test was repeated for QL in 5%, 2.5%, and 1.25% hemolyzed plasma (Table 4). The repeated QL in 5% hemolyzed plasma test results were similar to the original test (36.3% different from control QC) and showed matrix effects just within the acceptable threshold in 2.5% hemolyzed plasma (14.9% different from control QC). At 1.25% hemolyzed plasma, there was no significant matrix effect (11.1% different from control QC) observed for HZ. There were no significant matrix effects observed in hemolyzed plasma for any other analyte or for HZ at the QH level.

FIGURE 2.

Representative chromatograms of blank samples (black lines) scaled to LLOQ analyte and internal standard response (pink peaks). Analytes measured by polar method in the top row (AcINH and INA) and by nonpolar method in the bottom row (DAcHZ, DHZ, and DINH; “D” prefix indicates masses of derivatized compounds). For each chromatograph, the analyte is on top, and the internal standard is on the bottom.

TABLE 4.

Effects of hemolyzed plasma at 5%, 2.5%, and 1.25% hemolysis. Significant effects seen at 5% and borderline significant effects seen at 2.5% for HZ. n = 3 for each mean value. % Deviation (%Dev) = (test – control)/(control) × 100.

		QL (30.0 ng/mL)
Compound	Condition	Mean (%CV) ng/mL	%Dev from control
AcINH	Control	28.1 (5.8)	NA
	5% Hemolysis	28.3 (1.3)	0.9
	2.5% Hemolysis	29.6 (1.2)	5.4
	1.25% Hemolysis	29.2 (1.6)	4.0
INA	Control	25.9 (5.5)	NA
	5% Hemolysis	23.9 (1.1)	−7.7
	2.5% Hemolysis	26.0 (1.3)	0.2
	1.25% Hemolysis	26.1 (2.1)	0.7
AcHZ	Control	29.2 (6.1)	NA
	5% Hemolysis	29.2 (1.3)	−0.1
	2.5% Hemolysis	30.3 (2.0)	3.8
	1.25% Hemolysis	30.6 (2.7)	4.5
HZ	Control	30.6 (5.7)	NA
	5% Hemolysis	41.7 (2.1)	36.3
	2.5% Hemolysis	35.1 (1.8)	14.9
	1.25% Hemolysis	34.0 (1.5)	11.1
INH	Control	28.0 (7.3)	NA
	5% Hemolysis	27.6 (0.8)	−1.4
	2.5% Hemolysis	28.8 (2.6)	2.8
	1.25% Hemolysis	29.1 (2.9)	4.0

Specificity was assessed by comparing QL and QH spiked with relevant concentrations of potential concomitant medications (lamivudine, emtricitabine, tenofovir, zidovudine, nevirapine, darunavir, atazanavir, ritonavir, lopinavir, and efavirenz) to control QL and QH solutions and nominal concentrations. There were no effects seen with plasma spiked with concomitant medications compared with control QCs or nominal concentrations.

SRM channel cross‐talk was investigated between the analytes and respective internal standards. No significant crosstalk of internal standard into analyte channels from a blank‐IS sample was observed; however, a significant HZ‐IS response (8.7% of the typical blank‐IS HZ‐IS signal) from an injection of the HZ ULOQ sample without internal standard was observed. Since HZ could only be labeled 2 amu greater than the HZ analyte, this was unsurprising. The typical remedy is to increase the HZ‐IS amount utilized in the extraction; however, this caused unacceptable variability in INA assay performance, so HZ‐IS was returned to its original concentration. No other analytes showed significant crosstalk into any of the internal standard channels.

3.2.2. Calibration Curve and Range

Calibration curves were validated from 10.0 ng/mL for all analytes to either 500 ng/mL for AcHZ and HZ or 5000 ng/mL for INH, AcINH, and INA (Figure 3). Polar analytes were fitted with quadratic curves (1/X ² and 1/X weighting for AcINH and INA, respectively), and the nonpolar analytes were fitted with linear curves (1/X weighting for INH and AcHZ, and 1/X ² weighting for HZ). Calibration level inter‐assay accuracy was shown to be within ±5.9% for all analytes, and the inter‐assay precision was ≤ 13.5% for all analytes.

FIGURE 3.

Representative calibration curves for each analyte. Reported R ² values are from a single representative run.

3.2.3. Accuracy and Precision

Assay accuracy and precision were established using six replicates of each QC level (LLOQ, QL, QM, and QH) in each analytical run. The inter‐assay accuracy for all analytes was within ±5.4%, and the intra‐assay accuracy for all analytes was within ±7.4%. The inter‐assay precision was ≤ 5.6%, and the intra‐assay precision was ≤ 6.1% (Table 5).

TABLE 5.

Inter‐ and intra‐assay accuracy and precision. All QC levels were analyzed with six replicates on each run. Intra‐assay mean concentrations, accuracies, and precisions are presented as ranges (min., max.) over all runs performed. Accuracy and precision of the polar method were assessed over four runs; the HZ‐IS concentration was increased on one run to try and reduce the significance of crosstalk seen between HZ and HZ‐IS, which significantly affected INA performance. Data not included for INA from that accuracy and precision run. % Deviation (%Dev) = (test – control)/(control) × 100.

			Intra‐assay				Inter‐assay
Compound	QC level	Nominal concentration (ng/mL)	Total# of runs	Mean concentration range (ng/mL)	Accuracy range (%Dev from nominal)	Precision range (%CV)	Total # of replicates	Mean concentration (ng/mL)	Accuracy (%Dev from nominal)	Precision (%CV)
AcINH	LLOQ	10.0	4	9.53, 9.86	−4.7, −1.4	1.1, 2.6	24	9.72	−2.8	2.1
	QL	30.0	4	28.6, 29.3	−4.8, −2.3	0.5, 3.3	24	28.9	−3.6	2.2
	QM	800	4	748, 775	−6.5, −3.2	0.7, 2.1	24	765	−4.4	2.0
	QH	4000	4	3765, 3943	−5.9, −1.4	0.7, 2.9	24	3889	−2.8	2.7
INA	LLOQ	10.0	3	9.68, 10.3	−3.2, 3.1	2.0, 6.1	18	10.0	0.4	4.6
	QL	30.0	3	27.8, 28.9	−7.4, −3.6	1.0, 1.8	18	28.4	−5.4	2.1
	QM	800	3	776, 804	−3.0, 0.5	0.9, 2.0	18	795	−0.7	2.2
	QH	4000	3	3901, 4017	−2.5, 0.4	0.6, 2.8	18	3959	−1.0	2.1
AcHZ	LLOQ	10.0	3	10.0, 10.6	0.3, 6.1	1.3, 2.2	18	10.4	4.1	3.1
	QL	30.0	3	29.6, 31.0	−1.5, 3.3	1.1, 4.1	18	30.3	0.8	3.3
	QM	80.0	3	75.8, 84.8	−5.2, 6.0	1.4, 2.0	18	79.0	−1.3	5.6
	QH	400	3	382, 424	−4.4, 6.0	1.3, 2.6	18	399	−0.1	5.1
HZ	LLOQ	10.0	3	10.2, 10.7	1.9, 7.0	1.2, 3.0	18	10.5	5.1	2.9
	QL	30.0	3	29.1, 31.2	−2.8, 4.0	1.2, 5.2	18	30.1	0.4	4.2
	QM	80.0	3	80.1, 84.4	0.2, 5.5	1.7, 2.1	18	82.3	2.8	2.8
	QH	400	3	395, 409	−1.2, 2.2	0.8, 2.9	18	403	0.8	2.6
INH	LLOQ	10.0	3	9.67, 9.84	−3.3, −1.6	1.0, 1.5	18	9.78	−2.2	1.4
	QL	30.0	3	28.4, 28.8	−5.4, −4.0	0.9, 1.5	18	28.6	−4.8	1.4
	QM	800	3	803, 818	0.4, 2.3	0.5, 1.9	18	809	1.1	1.5
	QH	4000	3	4029, 4117	0.7, 2.9	1.4, 2.4	18	4071	1.8	2.1

Abbreviations: AcHZ, acetylhydrazine; AcINH, acetylisoniazid; CV, coefficient of variation; HZ, hydrazine; INA, isonicotinic acid; INH, isoniazid; LLOQ, lower limit of quantitation; QC, quality control; QH, high level QC; QL, low level; QM, mid‐level QC.

Dilution integrity and partial volume accuracy and precision were also determined using five replicates of either polar or nonpolar dilQC or QH, respectively. For all analytes, the dilution accuracy was within ±10.7%, and the precision was established ≤ 2.2%, indicating that 5× dilutions are appropriate for all analytes and 10× dilutions are appropriate for AcHZ and HZ. The partial volume accuracy and precision were established for all analytes within ±9.7% and ≤ 4.1%, respectively.

3.2.4. Matrix Effect, Recovery, and Process Efficiency

Matrix effect was assessed through calculating precision of slopes of three‐point curves prepared in plasma lots from six donors and by comparing peak areas to monitor how the matrix affected each analyte and IS differently, and PAR from post‐extract spiked samples and neat samples prepared at three concentrations.

The matrix effect slope analysis resulted in a %CV ≤ 3.1% for all analytes, indicating that the assay method does not have any significant matrix effects. The comparison of post‐extract spiked and neat samples showed a suppression of signal for some analytes (INA 64.5%, HZ 88.0%, and a slight suppression of AcHZ 96.2%) and an enhancement for others (AcINH 137% and INH F 145%). However, the internal standards demonstrated similar matrix effects, resulting in PAR matrix effects ranging from 99.2% to 113%, indicating the chosen internal standards are appropriate for the assay.

Recovery was assessed through the comparison of PAR from extracted and post‐extract spiked samples and showed recoveries ranging from 96.7% to 106%. Process efficiency was assessed through comparison of PAR from extracted and neat samples and showed process efficiencies ranging from 95.8% to 120%.

3.2.5. Bench‐Top Stability

Bench‐top stability of QL and QH samples was assessed by comparing freshly prepared and extracted QL and QH samples to those left at room temperature for up to 30 min prior to extraction and to nominal concentrations. This test established room temperature plasma stability for QC samples for up to 30 min, with the 30‐min test QCs being within ±9.8% of the control QC and within ±11.6% of nominal concentration for all analytes.

The maximum likely potential contribution of AcINH degradation to the other analytes was also assessed. A plasma sample containing AcINH at 5000 ng/mL (ULOQ) was prepared and extracted either immediately (T0) or sat at room temperature for 5, 10, 15, and 30 min prior to extraction. All analytes were monitored to qualitatively assess how PAR changed over time relative to T0 and the respective LLOQ PAR (AcINH was only compared with T0). The PAR for AcINH (at ULOQ), HZ, and INH did not change over time while INA and AcHZ PAR increased markedly from 5 to 30 min (Table 6), indicating that AcINH degradation could severely impact quantitation of low INA and AcHZ concentration samples without taking special precautions during sample handling processes.

TABLE 6.

AcINH stability in plasma. Peak area ratio (PAR) values are consistent for AcINH, HZ, and INH; however, they increase linearly over time for INA and AcHZ. AcINH reference standard is slightly contaminated with INH, resulting in INH signals approximately 140% of INH LLOQ signal. n = 3 for each mean value. % Deviation (%Dev) = (test – control)/(control) × 100.

Compound	Time (min)	Mean PAR (%CV)	%Dev from T0	Mean LLOQ PAR (%CV)	% of LLOQ PAR
AcINH	0	15.8 (0.7)	NA	NA	NA
	5	16.0 (0.7)	1.0
	10	15.9 (0.8)	0.1
	15	16.0 (1.3)	0.9
	30	15.6 (1.2)	−1.5
INA	0	0.019 (12.9)	NA	0.065 (2.6)	29.4
	5	0.086 (4.0)	351		132
	10	0.161 (2.7)	751		250
	15	0.246 (1.0)	1197		381
	30	0.480 (1.6)	2431		743
AcHZ	0	0.007 (15.5)	NA	0.088 (0.9)	8.2
	5	0.070 (1.9)	870		79.4
	10	0.134 (1.3)	1757		152
	15	0.202 (1.8)	2697		229
	30	0.398 (1.2)	5410		451
HZ	0	0.003 (0.7)	NA	0.144 (1.3)	1.8
	5	0.003 (11.6)	6.4		2.0
	10	0.003 (1.7)	0.4		1.8
	15	0.003 (1.3)	0.6		1.8
	30	0.003 (3.2)	1.9		1.9
INH	0	0.053 (1.7)	NA	0.037 (7.9)	141
	5	0.053 (1.9)	0.2		141
	10	0.053 (0.8)	−0.4		140
	15	0.053 (1.5)	0.1		141
	30	0.052 (2.7)	−0.6		140

Abbreviations: INH/metabolites; PAR, peak area ratio; LLOQ, lower limit of quantitation.

Bench‐top stability of PS1, spiking QL and QH, and spiking calibrator H and A solutions were also assessed by comparing solution aliquots left at room temperature to those retained in storage at −20°C. Test aliquots were retained at room temperature between 2.8 and 5.2 h and all compared within ±6.7% to control aliquots.

3.2.6. Freeze/Thaw Stability

To mimic clinical study plasma samples, QL and QH samples were each prepared and rapidly frozen in 1 mL and 2 mL aliquots in 1.7 mL and 2.0 mL plastic cryovials, respectively. These samples were rapidly thawed at 50°C and then rapidly frozen in a freezing block on dry ice for at least 15 min prior to thawing again. After the third thaw, the samples were extracted in triplicate. Results from the freeze/thaw test QC samples were compared with control QC samples, which had only been thawed once, and to nominal concentrations. For all analytes in all aliquot volumes and in all cryovials, the freeze/thaw test QC samples were within ±7.2% of control QC results and within ±12.1% of nominal concentrations. These results indicate that QC samples are stable through three freeze/thaw cycles using the rapid freeze/thaw technique (Table 7).

TABLE 7.

Freeze/thaw stability of analytes. Rapid freeze/thaw with heat block set to 50°C. 1 mL aliquot samples were in 1.7 mL cryovials and were kept on the heat block for 4.5–5.0 min for each thaw cycle. 2 mL aliquot samples were in 2.0 mL cryovials and were kept on the heat block for 7.0 min for each thaw cycle. The 1 mL and 2 mL aliquot samples each went through three freeze/thaw cycles, while the control samples just went through 1. n = 3 for each mean value. % Deviation (%Dev) = (test – control)/(control) × 100.

		QL (30.0 ng/mL)			QH (4000 or 400 ng/mL)
Compound	Condition	Mean (%CV) ng/mL	%Dev from control	%Dev from nominal	Mean (%CV) ng/mL	%Dev from control	%Dev from nominal
AcINH	Control	29.3 (2.9)	NA	−2.4	4089 (1.5)	NA	2.2
	1 mL Aliquot	28.8 (3.4)	−1.6	−4.0	4093 (1.1)	0.1	2.3
	2 mL Aliquot	28.9 (0.7)	−1.2	−3.6	3882 (2.7)	−5.1	−2.9
INA	Control	30.5 (1.3)	NA	1.7	4257 (2.4)	NA	6.4
	1 mL Aliquot	29.9 (3.8)	−2.1	−0.4	4266 (2.6)	0.2	6.6
	2 mL Aliquot	30.0 (1.1)	−1.6	0.1	4085 (2.3)	−4.1	2.1
AcHZ	Control	29.7 (1.2)	NA	−0.8	404 (1.8)	NA	0.9
	1 mL Aliquot	28.6 (3.0)	−3.9	−4.7	397 (1.2)	−1.6	−0.7
	2 mL Aliquot	29.9 (1.1)	0.4	−0.5	389 (0.8)	−3.7	−2.8
HZ	Control	30.8 (0.5)	NA	2.6	409 (1.6)	NA	2.1
	1 mL Aliquot	29.4 (3.5)	−4.3	−1.9	401 (1.1)	−1.7	0.4
	2 mL Aliquot	31.1 (2.1)	1.1	3.7	398 (1.9)	−2.6	−0.6
INH	Control	28.4 (1.4)	NA	−5.2	4004 (1.4)	NA	0.1
	1 mL Aliquot	26.4 (2.8)	−7.2	−12.1	3895 (2.1)	−2.7	−2.6
	2 mL Aliquot	28.3 (4.3)	−0.6	−5.8	3826 (2.3)	−4.5	−4.3

Abbreviations: INH/metabolites, %CV, QL and QH.

3.2.7. Reinjection Reproducibility

Reinjection reproducibility was established by re‐injecting a calibration curve with six replicates of QL and QH 3 days (polar method) or 4 days (nonpolar method) after initial analysis and quantitating the reinjected QL and QH samples off the reinjected and originally injected curves. Mean results were compared with the original mean QC results and to nominal concentrations. For all analytes, reinjected QC results compared with original results and to nominal concentrations within ±4.9% when measured off the reinjected or originally injected curves. These results show that reinjected samples give reproducible results with full or partial run reanalysis after autosampler storage at 15°C for 3 days (polar method) or 4 days (nonpolar method).

3.2.8. Long‐Term Stability

To assess the long‐term stability of the full range of concentrations for all analytes in all solvents and plasma, PS1, spiking QL and QH, spiking calibrators H and A, plasma calibrators H and A, and QL and QH solutions were assessed by comparing freshly prepared solutions in solvent or plasma to those stored at −20°C or at −80°C, respectively. Long‐term stability was established between 371 and 557 days for PS1 solutions and up to 70 days for spiking QC solutions, 267 days for spiking calibration solutions, 119 days for plasma calibration solutions, and 168 days for QC solutions.

3.3. Application to Clinical Samples

This method was applied to two clinical studies where participants provided written informed consent. In the first, an internal IRB‐approved study for assay development (COMIRB #09‐0761) was used to collect blood from people being treated with INH to assess whole blood stability by investigating the effects of keeping whole blood on wet ice or at ambient temperature prior to plasma processing. The second clinical study were secondary research investigations of a large randomized controlled study of isoniazid preventive therapy in pregnant and postpartum women with HIV, which observed high rates of hepatotoxicity (COMIRB# 20–0149, NCT01494038) (Gupta et al. 2019; Montepiedra et al. 2026). A subset of participants in the study underwent intensive PK sampling, and samples from two of these participants were used to assess ISR prior to application to the rest of the study samples.

3.3.1. Whole Blood Stability

Whole blood stability was assessed with clinical study blood samples that were drawn into multiple pre‐cooled (2°C–6°C) K₂EDTA vacutainer tubes per patient and then placed on ice until receipt in the lab. One blood tube was immediately processed upon receipt in the lab (T0), and additional tubes from the same patient were kept on ice or left at ambient for 1 (T1), 2 (T2), or 4 (T4, ambient only) hours prior to plasma processing. Plasma was processed by centrifuging the blood tubes at 1000 × g for 10 min at 4°C, and then plasma was aliquoted and stored at −70°C or colder until analysis.

T1 and T2 samples stored on ice compared within 7.0% to T0 for all analytes. However, when comparing samples stored at ambient to their respective T0 samples, concentrations were seen to increase for INA and AcHZ over time (48.0% and 25.1% greater than T0, respectively, at T4). Comparing samples at ambient to those stored on ice, increases of up to 26.8% and 33.7% for INA and AcHZ, respectively, were observed at T2. These data indicate that blood samples should be maintained on ice during the collection and plasma harvest process to preserve INA and AcHZ concentrations.

3.3.2. Sample Reanalysis Reproducibility

The rapid freeze/thaw technique was initially tested on 14 samples from two study participants. Four samples were chosen from each of the participants to reanalyze, where the number of reportable results for each analyte was as follows: HZ n = 4, AcHZ n = 6, INH n = 7, and AcINH and INA n = 8. Results from the reanalysis were compared with those from the original analysis with this assay using the ISR calculation (equation 1): ISR% = (repeated result − original result)/(average result) × 100. All results within the assay ranges were compared within ±9.4% except for one sample, which had a reportable INH concentration on the original run (12.0 ng/mL) and was below the limit of quantitation upon reanalysis. These results indicate that sample results can be accurately reproduced using the rapid freeze/thaw process.

4. Conclusions

Accurate and precise quantitation of INH, AcINH, INA, AcHZ, and HZ has been validated in human plasma over the established concentration ranges while limiting breakdown of AcINH into INA and AcHZ using a novel rapid freeze/thaw method. While other groups (Chen et al. 2005; Hee et al. 2015; Huang et al. 2009; Ky Anh et al. 2022; Ng et al. 2007; Song et al. 2017) have developed assays which could quantitate some or all these analytes, to our knowledge, none assessed the effects of AcINH degradation in plasma on AcHZ or INA concentrations, nor developed a method to limit this process. If specialized sample processing, handling, and storage techniques were not followed as specified, there may have been significant changes to AcHZ and INA levels during the sample analysis utilized by other reported methods.

In addition to implementing a novel rapid freeze/thaw procedure to limit AcINH degradation, the maximum likely degradation of AcINH due to rapid freeze/thaw was monitored by extracting an AcINH‐only ULOQ sample on each analytical run. The AcINH only ULOQ was kept in the heat block for a time equal to the longest time any other sample was kept on the heat block and should reflect the approximate maximum contribution of AcINH degradation on the INA and AcHZ concentrations. Degradation in clinical samples with INH and other metabolites present may be more nuanced, however, and should be evaluated in future analyses of clinical samples. Caution should be taken when interpreting results from samples with INA and AcHZ results near the LLOQ validated for this assay, particularly if AcINH concentrations are high or for rapid acetylators that produce relatively high plasma concentrations of AcINH.

Additionally, caution should be taken when interpreting HZ results from hemolyzed samples, as 5% hemolysis was shown to significantly affect HZ results at the QL level, while 2.5% and 1.25% met acceptance (<15% different from control) but showed bias. The mechanism of the matrix effect of hemolyzed plasma on HZ concentration results is unknown but may be due to interactions between thiol‐containing compounds like glutathione released from erythrocytes during hemolysis and HZ interfering with the derivatization process (Giustarini et al. 2004) or other unknown mechanisms. The effect observed at 5% hemolysis represents an extreme circumstance that is not typically encountered, but the mechanism behind the increased HZ concentrations observed in highly hemolyzed plasma should be further investigated.

While this assay performed well in application to clinical study samples, there were some improvements that could be made. Many samples had to be reanalyzed at a dilution for AcHZ, and there were a number of samples that were detectable but BLQ for HZ. Increasing the analytical range for AcHZ above the validated upper limit of 500 ng/mL and decreasing the lower limit of quantitation below 10 ng/mL for HZ could be investigated. Additional longer‐term matrix stability data are also needed, though internal assessments are ongoing.

This method was utilized to accurately determine concentrations of INH and four of its metabolites to assess their PK among non‐pregnant people and pregnant and postpartum women with HIV receiving INH preventative therapy, with future applications to other PK studies (Klein et al. 2024). Characterization of the PK and exposure–response relationships of INH and its metabolites, particularly AcHZ and HZ, with improved approaches to limit degradation of AcINH into INA and AcHZ, will improve our understanding of the underlying mechanisms behind hepatotoxicity and lead to improved patient care.

Funding

We would also like to thank the Colorado Antiviral Pharmacology Laboratory for their contributions in developing the analytical methods. Funding for this work was supported by K08 AI152942 (KMB) and by NICHD contract number HHSN275201800001I (KMB). Overall support for the International Maternal Pediatric Adolescent AIDS Clinical Trials Network (IMPAACT) was provided by the National Institute of Allergy and Infectious Diseases (NIAID) with co‐funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and the National Institute of Mental Health (NIMH), all components of the National Institutes of Health (NIH), under Award Numbers UM1AI068632 (IMPAACT LOC), UM1AI068616 (IMPAACT SDMC), and UM1AI106716 (IMPAACT LC), and by NICHD contract number HHSN275201800001I. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.