Simultaneous Determination of Six Antiretroviral Drugs in Human Hair Using an LC-ESI⁺-MS/MS Method: Application to Adherence Assessment

Abstract

BACKGROUND:

The determination of antiretroviral drugs in hair is receiving considerable research interest to assess long-term adherence to antiretroviral therapy (ART). Currently in China, lamivudine, zidovudine, nevirapine, efavirenz, ritonavir, and lopinavir are combined as first- and second-line free therapy regimens and are recommended for people living with human immunodeficiency virus (HIV) (PLWH). Simultaneous determination of the six antiretroviral drugs in human hair is important for accurately and widely assessing long-term adherence in Chinese PLWH receiving different ART regimens.

METHODS:

Six drugs were extracted from 10-mg hair samples incubated in methanol for 16 h at 37 °C and then analyzed by liquid chromatography with tandem mass spectrometry (LC-MS/MS) using a mobile phase of 95% methanol, with an electrospray ionization (ESI) source in multiple reaction monitoring and positive mode.

RESULTS:

The LC-ESI⁺-MS/MS method exhibited a linear range (R² > 0.99) within 6–5000, 10–5000, 6–50000, 12–50000, 8–5000, and 8–12500 pg/mg for lamivudine, zidovudine, nevirapine, efavirenz, ritonavir, and lopinavir. For all six drugs, the limits of quantification ranged between 6 and 12 pg/mg. The intra-day and inter-day coefficients of variation were within 15%, and the recoveries ranged from 91.1% to 113.7%. Furthermore, the other validation parameters (i.e., selectivity, matrix effect, stability, and carryover) met the acceptance criteria stipulated by guidelines of the United States Food and Drug Administration and European Medicines Agency. Significant intergroup differences were observed between high- and low-adherence groups, with high inter-correlations in the hair content of the six drugs.

CONCLUSIONS:

The developed method demonstrated good reliability, to comprehensively and accurately assess adherence in PLWH receiving different ART regimens.

Introduction

Currently, antiretroviral therapy (ART), consisting of multiple antiretroviral drugs, is the cornerstone for preventing and managing acquired immunodeficiency syndrome (AIDS) infection. It can substantially maintain virological suppression and reduce morbidity and mortality in people living with human immunodeficiency virus (HIV) (PLWH).^{1, 2} Optimal adherence to ART is another key to achieve positive treatment outcomes.^{3, 4} Previous studies have revealed that optimal adherence has a high correlation with treatment outcomes^{5, 6} and is the main predictor for virological suppression, immune restoration, and clinical progression.^7–9 Therefore, an assessment of adherence would help predict ART treatment outcomes.

Previously, several methods have been reported to assess ART adherence.^10–13 Among them, determining the content of antiretroviral drugs in different bio-matrices has received considerable research interest,^{12, 14, 15} especially hair.^16–19 Researchers postulate that hair as a reliable bio-matrix can reflect adherence over a long-term regimen, whereas other bio-matrices (e.g., plasma and urine) measure adherence within several hours or days.²⁰ The drug contents in the 1-cm hair segment closest to the scalp can retrospectively reflect drug usage during the past month as the growth rate of human hair is approximately 1 cm per month.²⁰ Furthermore, previous studies have verified that determining the hair antiretroviral drug content was reliable for long-term ART adherence assessment.^{21, 22} Thus, the hair antiretroviral drug content would be appropriate for assessing long-term ART adherence.

Currently available studies have mostly measuring the hair content of a single antiretroviral drug to assess ART adherence, rather than all kinds of drugs from an identical regimen.^{21, 23–25} This might be mainly due to the common opinion that patients would co-administer all kinds of drugs based on the regimen to achieve better treatment outcomes. Thus, determining a single drug in hair could reflect medication adherence in a patient. However, patients might often miss drug doses and/or other drugs from identical regimens or administer wrong drugs owing to adverse drug reactions and ART regimen changes. Accordingly, the hair content of a single drug may only reflect the real usage of the determined drug, rather than all drugs in an identical regimen. Moreover, there exists an inter-individual variation in the hair content of an identical drug among PLWH administering the same drug dose.¹⁸ This is mainly attributed to the individual differences in drug metabolism induced by differences in physiological characteristics of patients. Additionally, it can result from differences in drug dissolution out of hair, which might be attributed to the inter-individual differences in irradiation by sunlight and life habits (e.g., hair washing frequency).²⁶ Therefore, the hair content of a single drug from a regimen might limit accurate and comprehensive assessment of long-term ART adherence in patients currently administering with multiple drugs. In contrast, simultaneously determining the hair content of all drugs from an identical regimen would help eliminate these limitations.

Moreover, ART regimens composed of different types of antiretroviral drugs appear to vary with patients. In China, three regimens from six mainstream ART drugs, lamivudine (3TC), zidovudine (AZT), nevirapine (NVP), efavirenz (EFV), ritonavir (RTV), and lopinavir (LPV), are recommended by the National Free Antiretroviral Treatment Program for most Chinese PLWH on free ART treatment.⁴ Among these, 3TC+AZT+NVP and 3TC+AZT+EFV as two first-line regimens are applied as initial therapy, and 3TC+AZT+LPV/r (a drug tablet combined with RTV and LPV) is a second-line regimen for PLWH with virological failure. For widespread application, a sensitive method for simultaneously determining the six drugs needs to be developed.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS) with high sensitivity and specificity has been widely employed to determine the hair content of antiretroviral drugs.^27–30 For the aforementioned six drugs, previous studies have reported the LC-MS/MS determination for a single drug or a few drugs in human hair.^{18, 19, 23, 25, 31} However, no study has developed an LC-MS/MS method for simultaneously determining six drugs in hair. Furthermore, the six antiretroviral drugs possess different lipophilicities. 3TC and AZT, with negative logP values, are markedly hydrophilic, while NVP, EFV, RTV, and LPV demonstrate lipophilicity, with logP >2 (data obtained from chemicalize.com³²). This implies that they belong to two drug types, presenting different polarities. Traditionally, electrospray ionization (ESI) in combination with a mobile phase composed of a low ratio of organic solvent has been used for determining polar drugs, including 45% methanol for 3TC²⁴. In contrast, atmospheric pressure chemical ionization (APCI) accompanied by a mobile phase with a high ratio of organic solvent (90% methanol) was employed for low-polarity drugs.¹⁹ Consequently, simultaneously determining the six drugs with different polarities requires an optimal balance between the ionization source and mobile phase. Compared with APCI, ESI would be more suitable for assessing polar drugs (i.e., 3TC¹⁸) and could also be applied to low-polarity drugs (e.g., NVP^{18, 23}, EFV^{25, 31}, RTV, and LPV³¹). Moreover, the mobile phase with a high ratio of organic solvent would enhance the ionization of low-polar compounds.³³ Therefore, the combination of an ESI source and mobile phase with a high ratio of organic phase would be suitable for simultaneously determining the six drugs possessing different polarities.

In this study, we aimed to develop a simple and sensitive LC-ESI⁺-MS/MS method with a mobile phase composed of a high ratio of organic solvent for simultaneously determining the six antiretroviral drugs in human hair and validate it under the United States Food and Drug Administration (US FDA) and European Medicines Agency (EMA) guidelines.^{34, 35} The developed method would also be verified with population analysis for its reliability in accurately and comprehensively assessing ART adherence in PLWH.

Materials and Methods

CHEMICALS

Standards, 3TC, AZT, NVP, EFV, RTV, and LPV were purchased from TargetMol (Shanghai, China); corresponding deuterated internal standards (IS), 3TC-d3, AZT-d4, NVP-d3, EFV-d5, RTV-d6, and LPV-d8 were obtained from Toronto Research Chemicals (Toronto, ON, Canada). Ammonium acetate and high-performance liquid chromatography (HPLC)-grade methanol were supplied by Tedia (Fairfield, OH, USA) and Sigma Aldrich (St. Louis, MO, USA), respectively. Tri-distilled water was from Watsons (Hong Kong, China).

Stock solutions of 3TC, NVP, and their corresponding ISs were individually dissolved in methanol at a concentration of 1 mg/mL. AZT, EFV, RTV, and LPV were used at a concentration of 400 μg/mL, with 50 μg/mL IS concentrations. They were all stored at −20°C. The working solutions with mixed six drug standards were obtained by diluting stock solutions with methanol to final concentrations at 0.1–10000 ng/mL. IS solutions were diluted with methanol at 5 ng/mL for 3TC-d3, NVP-d3, RTV-d6 and LPV-d8, 20 ng/mL for AZT-d4, and 100 ng/mL for EFV-d5.

PARTICIPANTS AND SAMPLE COLLECTION

In total, 93 PLWH were randomly recruited from the Guangxi Zhuang Autonomous Region, China. All participants enrolled in this study provided written informed consent before inclusion. This study was in accordance with the Declaration of Helsinki and was approved by the Health Science Research Ethics Boards of Southeast University and Guangxi Center of Disease Control in China and the Institutional Review Board (IRB) at the University of South Carolina, USA.

Participants self-reported their demographic information, ART regimens, and ART adherence during the previous month and provided their hair strands from the occiput with the assistance of a local research staff member. Among them, 34 participants (male/female = 26/8) were using the 3TC+AZT+NVP regimen, 36 participants (male/female = 25/11) were on 3TC+AZT+EFV, and 23 participants (male/female = 9/14) were on 3TC+AZT+LPV/r. Additionally, participants following identical ART regimens were divided into high- and low-adherence groups according to the days of their oral drug administration, i.e., >27 and ≤ 27 days in the past month.⁴ To match the one-month time span of questionnaire information from the participants, only the 1-cm hair segment closest to the scalp was collected as analysis sample.

HAIR SAMPLE PREPARATION

The analysis samples were dried using pure N₂ at 50°C after double rinsing with methanol. The dried samples were cut into 1–2 mm pieces using scissors and weighed at 10 mg before transfer into a clean centrifuge tube. Then, the samples were incubated in 950 μL methanol with 50 μL IS working solutions at 37°C for 16 h.^{18, 19} The incubation solution was mixed by vortexing for 2 min and separated using centrifugation with 8056.8 g for 5 min. The supernatant (800 μL) was transferred into another clean centrifuge tube and evaporated with pure N₂ at 50°C. Finally, the residue was re-dissolved in 50 μL mobile phase for detection.

LC-MS/MS ASSAYS

Simultaneous determination of antiretroviral drugs was performed using an HPLC-MS/MS system. Briefly, the reconstituted solution (10 μL) was injected into the 1200 HPLC system (Agilent, Waldbronn, Germany) and separated with a Platisil ODS C18 column (5 μm, 150 mm×4.6 mm; Dikma), protected by a C18 guard cartridge (5 μm, 10 mm×4.6 mm; Dikma). A mixture of methanol and 0.004 mol/L ammonium acetate in tri-distilled water (95:5, v/v) was used as the mobile phase, which was filtered through a 0.22 μm microporous membrane and ultrasonic-processed for 10 min after preparation. The column oven temperature was 40 ± 1°C, and the flow rate was 300 μL/min.

MS detection was performed with a 3200 QTRAP tandem mass spectrometer (ABI, Foster City, CA, USA), equipped with an ESI source operating in MRM-positive mode. Liquid nitrogen was gasified as the nebulizing gas. The symmetric heaters were at 400°C and the ion spray voltage was 4500 V. Curtain gas was maintained at 10 psig, collision gas at medium (approximately 7 psig), and ion source gas 1 and gas 2 at 40 psig. The precursor ions and product ions of all analytes and their optimum conditions were re-optimized based on our previous studies^{18, 19} (see Table S1 and Fig S1 in Supplemental Materials) and the dwell time was 100 ms.

METHOD VALIDATION

This method was validated under the US FDA and EMA guidelines.^{34, 35} Hair samples from a healthy volunteer without any antiretroviral drug administration were collected as blank hair matrices. For validation, 10 mg double-rinsed and nitrogen-dried blank matrices were blended with 50 μL standard working solutions, 50 μL IS working solutions, and 900 μL methanol and then treated in the same manner as aforementioned natural hair samples.

Linearity and Limit of Detection (LOD)/Limit of Quantitation (LOQ)

Standard curves were prepared by spiking working solutions at desired ranges to the blank hair matrices and pretreated as aforementioned. The LOD was the lowest standard concentration detected at a signal-to-noise ratio(S/N) of 3 and the LOQ was the lowest standard concentration at S/N = 10.

Precision and Recovery

Intra-day and inter-day precision were validated as coefficients of variation (CV), calculated with five replicates at LOQ, low, medium, and high concentrations in a day and over five consecutive days, respectively. Recovery was determined by comparing the drug concentration calculated with the standard curve and the actual spiked drug concentration. It was calculated in triplicate at low, medium, and high concentrations within one day.

Selectivity and Matrix Effect

Selectivity and matrix effect were validated using six lots of blank hair matrices from six healthy volunteers. Selectivity was the peak area ratio of the interference response near the retention time from the blank matrices with the target drug at LOQ level or its corresponding IS. Matrix effect was evaluated by measuring the matrix factor (MF) and IS normalized MF in the six blank matrices, which were separately added at low and high concentrations, and determined in triplicate. Additionally, the effect of the hair matrix on sensitivity was estimated as the variation in the sensitivity of standard curves with and without blank hair matrices (see §1 and Fig S2 in Supplemental Materials).

Stability and Carryover

Stability was the percentage of the drug concentration calculated using a standard curve versus the actual drug standard concentration under six different conditions (i.e., bench-top stability, long-term stability, stock solution stability, auto-sampler stability, high-temperature stability and freeze-thaw stability, details are provided in §2 in Supplemental Materials), which was validated in triplicate at low and high concentrations. Carryover effect was the peak area ratio of the carryover in the blank samples following the highest concentration standard samples to the drug at LOQ level.

STATISTICAL METHODS

Statistical analyses were performed using the SPSS 22.0 for windows (IBM, Armonk, NY, USA). All data on drug content were log-transformed to match the normal distribution. Chi-square test was used to assess differences in gender distribution between different adherence groups. The t-test for two independent samples was used to examine intergroup differences in drug content among participants following an identical regimen. Pearson correlation analysis was used to assess correlations in hair content between different drugs among participants with identical regimens. Furthermore, covariance and partial correlation analyses were separately performed to examine the intergroup differences in hair drug contents and correlations among them when gender and age were controlled.

Results

CHROMATOGRAPHY

The six drugs and their ISs were well chromatographically separated, without interference peaks (Fig.1). All drugs in the natural hair samples from PLWH showed well-resolved peaks under the same conditions. Furthermore, the ion chromatograms of all analytes in the blank matrix spiked with standard solutions at LOQ levels revealed good sensitivity of this instrument (see Fig.S3 in Supplemental Materials). The analysis time was <13 min, which is suitable for routine fast analysis.

Figure 1.

The chromatograms of the six analytes (a) 3TC, (b) AZT, (c) NVP, (d) EFV, (e) RTV, and (f) LPV in blank hair matrix, blank hair matrix spiked with standard solutions, and natural hair samples from PLWH, and (g-h) the corresponding ISs of the six analytes spiked with the blank hair matrix. The standard concentration was 100 pg/mg for each drug when spiked with the blank hair matrix. The drug contents of natural hair samples were 101, 126, 100, 104, 140, and 150 pg/mg for 3TC, AZT, NVP, EFV, RTV, and LPV, respectively. The IS concentrations were 5 ng/mL for 3TC-d3, NVP-d3, RTV-d6, and LPV-d8, 20 ng/mL for AZT-d4, and 100 ng/mL for EFV-d5. 3TC, lamivudine; AZT, zidovudine; NVP, nevirapine; EFV, efavirenz; RTV, ritonavir; LPV, lopinavir.

METHOD VALIDATION

As listed in Table 1, the coefficients of correlation (r) were all >0.99, indicating that this method showed good linearity in the set concentration range. The LOQs were <12 pg/mg, which were better than or matched previous methods (i.e., 6 vs. 15 ¹⁸ and 10 ²⁴ pg/mg for 3TC; 10 vs. 36 ¹⁹ pg/mg for AZT; 6 vs. 6 ¹⁸ and 250 ²³ pg/mg for NVP; 12 vs. 16 ¹⁹, 625 ²⁵ and 50 ³¹ pg/mg for EFV; 8 vs. 10 ³¹ and 12 ¹⁹ pg/mg for RTV; 8 vs. 50 ³¹ and 10 ¹⁹ pg/mg for LPV).

Table 1.

The linearity, limits of detection and quantitation, and retention time for the six drugs in hair.

Compound	Standard curve	R 2 a	Linear range (pg/mg)	LOD b (pg/mg)	LOQ c (pg/mg)	Retention time (min)
Lamivudine	y=0.01129x+0.00404	0.999	6-5000	2	6	6.4
Zidovudine	y=0.02958x+0.06405	0.994	10-5000	3	10	6.8
Nevirapine	y=0.06132x+0.0549	0.997	6-50000	2	6	7.5
Efavirenz	y=0.05434x-0.02495	0.999	12-50000	4	12	8.9
Ritonavir	y=0.03949x+0.04965	0.999	8-5000	3	8	9.2
Lopinavir	y=0.03670x + 0.00279	0.998	8-12500	3	8	10.5

^aR² is the square of the correlation coefficient.

^bLimits of detection (LOD) were set at S/N = 3.

^CLimits of quantitation (LOQ) were set at S/N = 10.

Intra-day and inter-day precisions both had CVs <15%, and the recovery ranged from 91.1% to 113.7% (Table 2), which were acceptable under the US FDA and EMA guidelines^{34, 35}. Additionally, this method showed good selectivity with the values <15% (see Table S2 in Supplemental Materials). As listed in Table 3, the CVs of MF for the six drugs and their corresponding ISs were <15% and those of IS normalized MF were <15%, indicating that no significant matrix effect was observed for detecting the six drugs using this method.

Table 2.

Intra- and inter-day coefficients of variation and recovery for the six drugs in hair.

Compound	Nominal a (pg/mg)	Intra-day CV (n=5, %) b	Inter-day CV (n=5, %) b	Recovery (M±SD, %) c
Lamivudine	6	3.1	6.3	‒
	10	4.3	7.6	112.5±4.8
	2000	3.0	5.4	94.4±2.8
	4000	3.2	4.3	104.2±3.3
Zidovudine	10	12.4	12.0	‒
	20	6.0	7.4	98.0±5.9
	2000	3.5	6.2	96.7±3.4
	4000	3.9	5.2	108.8±4.2
Nevirapine	6	6.3	5.7	‒
	10	3.8	8.0	107.6±4.1
	20000	4.3	4.8	93.1±4.0
	40000	9.2	7.5	95.6±8.8
Efavirenz	12	6.9	8.0	‒
	20	4.7	7.3	98.8±4.6
	20000	7.8	6.7	91.1±7.1
	40000	4.3	6.3	95.7±4.1
Ritonavir	8	9.1	9.2	‒
	20	3.1	5.1	105.4±3.2
	2000	3.6	4.1	112.0±4.0
	4000	2.6	3.0	112.1±2.9
Lopinavir	8	7.5	7.5	‒
	20	3.8	4.2	95.2±3.6
	4000	2.4	3.3	113.7±2.7
	10000	2.3	4.1	110.7±2.5

^aIntra- and inter-day coefficients of variation (CVs) were calculated at four levels, limits of quantitation (LOQ), low, medium, and high concentrations.

^bIntra-day and inter-day precisions were estimated by CVs.

^cRecovery was evaluated at three levels, low, medium, and high concentrations.

Table 3.

The matrix effect with matrix factor (MF) and its coefficient of variation (CV_MF) for the six drugs in different blank hair matrices.

Low concentration	Lot A	Lot B	Lot C	Lot D	Lot E	Lot F	Mean	SD	CVMF, %
MF for ESs
Lamivudine	1.10	1.15	1.16	1.19	0.89	1.07	1.09	0.11	10.0
Zidovudine	1.15	1.17	1.04	1.13	1.09	1.02	1.10	0.06	5.6
Nevirapine	1.17	1.14	0.89	1.14	0.92	1.06	1.05	0.12	11.6
Efavirenz	1.07	1.04	1.08	1.05	1.16	0.95	1.06	0.07	6.5
Ritonavir	1.06	1.02	1.13	1.07	1.03	1.11	1.07	0.04	3.9
Lopinavir	1.00	1.11	1.07	0.98	1.14	1.09	1.06	0.06	6.0
MF for ISs
Lamivudine-d3	1.08	1.21	1.05	1.21	0.99	1.02	1.09	0.09	8.5
Zidovudine-d4	1.02	1.03	0.91	0.97	0.97	0.94	0.97	0.05	4.7
Nevirapine-d3	1.08	1.12	0.83	1.11	1.08	1.03	1.04	0.11	10.3
Efavirenz-d5	1.12	1.01	1.16	1.04	1.19	0.88	1.07	0.11	10.7
Ritonavir-d6	1.06	1.11	1.12	1.13	0.92	0.95	1.05	0.09	8.8
Lopinavir-d8	0.88	1.22	1.00	0.98	1.18	1.03	1.05	0.13	12.3
IS normalized MF for ESs
Lamivudine	1.02	0.95	1.10	0.98	0.90	1.04	1.00	0.07	7.3
Zidovudine	1.13	1.13	1.14	1.16	1.13	1.08	1.13	0.03	2.3
Nevirapine	1.09	1.02	1.06	1.02	0.85	1.03	1.01	0.08	8.2
Efavirenz	0.96	1.03	0.93	1.00	0.98	1.08	0.99	0.05	5.5
Ritonavir	1.00	0.92	1.01	0.95	1.12	1.17	1.02	0.10	9.5
Lopinavir	1.14	0.91	1.07	1.00	0.96	1.06	1.02	0.08	7.9
High concentration	Lot A	Lot B	Lot C	Lot D	Lot E	Lot F	Mean	SD	CVMF, %
MF for ESs
Lamivudine	1.14	1.17	1.15	1.19	1.02	1.06	1.12	0.07	6.1
Zidovudine	0.95	1.03	1.05	1.01	0.90	0.93	0.98	0.06	6.3
Nevirapine	0.90	0.90	1.01	1.05	0.89	0.84	0.93	0.08	8.4
Efavirenz	1.09	1.11	1.12	1.05	1.07	1.01	1.07	0.04	3.6
Ritonavir	1.16	1.09	1.18	0.95	1.27	1.07	1.12	0.11	9.8
Lopinavir	0.85	0.86	0.93	1.15	1.12	0.95	0.98	0.13	13.1
MF for ISs
Lamivudine-d3	1.02	1.18	1.05	1.08	1.04	1.03	1.07	0.06	5.5
Zidovudine-d4	1.07	0.92	0.94	0.90	1.01	0.98	0.97	0.06	6.4
Nevirapine-d3	1.04	1.04	1.07	1.09	1.04	0.96	1.04	0.05	4.4
Efavirenz-d5	1.06	1.03	1.15	1.13	0.97	1.03	1.06	0.07	6.4
Ritonavir-d6	1.10	1.00	1.14	0.93	1.23	1.12	1.09	0.11	9.8
Lopinavir-d8	1.02	1.00	0.99	1.22	1.07	1.13	1.07	0.09	8.3
IS normalized MF for ESs
Lamivudine	1.13	0.99	1.09	1.10	0.97	1.02	1.05	0.06	6.0
Zidovudine	0.89	1.12	1.13	1.12	0.89	0.95	1.02	0.12	11.6
Nevirapine	0.87	0.87	0.94	0.96	0.85	0.87	0.89	0.04	4.9
Efavirenz	1.03	1.07	0.98	0.93	1.11	0.98	1.02	0.07	6.5
Ritonavir	1.05	1.09	1.03	1.02	1.03	0.95	1.03	0.05	4.5
Lopinavir	0.83	0.87	0.94	0.95	1.05	0.84	0.95	0.10	11.1

Notes: IS, internal standard; ES, external standard.

The deviation against initial concentration and CVs of stability under the six different conditions were <15% (see Table S3 in Supplemental Materials). All values met the acceptance criteria, indicating that drugs were stable under the six conditions. The carryover in the blank sample was <20% of each drug at the LOQ level, except for LPV (29.4%), and <2% for each IS (see Table S4 in Supplemental Materials).

COMPARISON OF HAIR DRUG CONTENTS BETWEEN PATIENTS WITH HIGH- AND LOW-ADHERENCE

The developed method was employed to determine the six drugs contents in natural hair samples for estimating its reliability in comprehensively assessing ART adherence.

As listed in Table 4, for patients administering 3TC+AZT+NVP, those with high-adherence were mostly male relative to those with low-adherence (p<0.01), as examined by the Chi-square test. The high-adherence group revealed significantly more days of oral drug administration within the past month (p<0.001), but no significant difference was observed between them in terms of age (p>0.05). Moreover, they presented significantly higher hair contents of AZT and NVP than those with low-adherence (p<0.001) and higher 3TC contents (p=0.054), whereas the difference was close to statistical significance. Covariance analysis further verified that the intergroup differences in the three drugs contents remained significant (ps<0.05) when gender and age were used as covariates, indicating no significant influence on drug content (ps>0.106).

Table 4.

Comparisons between high- and low-adherence groups in gender, age, ART adherence, and the hair drug content for patients with identical ART regimens.

	High adherence group a	Low adherence group a	Statistical value b
			Chi-square/t-test	Covariance analysis
3TC+AZT+NVP
Number	21	13	–	–
Gender, male/female	17/4	9/4	χ2=9.529**	–
Age, years	42.76±10.51, 23–68	45.54±8.30, 28–53	t32= −0.874	–
Adherence (days/month) c	29.71±0.72, 28–30	20.69±5.14, 10–27	t12.112=4.990***	–
3TC content, pg/mg d, e	374±243, 110–1076	211±165, 0–524	t13.912=2.107+	F1. 30=6.770*, ηp2=0.184
AZT content, pg/mg d, e	1249±1632, 133–7769	205±450, 0–1593	t13.520=5.554***	F1. 30=53.150***, ηp2=0.639
NVP content, pg/mg d, e	13417±1975, 10352–18167	4301±6618, 0–14353	t12.024=4.603***	F1. 30=38.467***, ηp2=0.562
3TC+AZT+EFV
Number	22	14	–	–
Gender, male/female	14/8	11/3	χ2=5.444*	–
Age, years	43.82±8.14, 31–64	40.29±11.10, 19–55	t34=1.315	–
Adherence (days/month) c	30.00±0.00, 30	21.64±3.67, 15–26	t13.000=7.017***	–
3TC content, pg/mg d, f	383±210, 110–910	120±203, 0–790	t13.978=4.038**	F1. 32=22.734***, ηp2=0.415
AZT content, pg/mg d, f	1625±1201, 82–3586	133±387, 0–1458	t16.414=7.397***	F1. 32=64.591***, ηp2=0.669
EFV content, pg/mg d, f	5950±2787, 2179–12762	1391±2074, 0–6039	t13.444=4.381***	F1. 32=25.356***, ηp2=0.442
3TC+AZT+LPV/r g
Number	12	11	–	–
Gender, male/female	3/9	6/5	χ2=1.087	–
Age, years	36.42±10.30, 22–50	37.55±10.26, 25–52	t21= −0.285	–
Adherence (days/month) c	29.92±0.29, 29–30	22.82±3.31, 20–27	t10.086= 6.478***	–
3TC content, pg/mg d, h	520±339, 73–1358	278±162, 0–537	t12.563=1.743	F1.19=5.474*, ηp2=0.224
AZT content, pg/mg d, h	570±386, 192–1554	194±218, 0–670	t11.053=2.873*	F1. 19=17.778***, ηp2=0.483
RTV content, pg/mg d, h	784±260, 426–1231	258±187, 0–468	t10.361=2.881*	F1. 19=16.315***, ηp2=0.462
LPV content, pg/mg d, h	6611±2973, 3248–12195	1912±1725, 0–4333	t10.587=2.905*	F1. 19=14.992**, ηp2=0.441

Notes:

⁺p=0.054<0.1

^*p<0.05

^**p<0.01

^***p<0.001.

ART, antiretroviral therapy; 3TC, lamivudine; AZT, zidovudine; NVP, nevirapine; EFV, efavirenz; RTV, ritonavir; LPV, lopinavir; 3TC+AZT+NVP, 300-mg 3TC+600-mg AZT+400-mg NVP per day; 3TC+AZT+EFV, 300-mg 3TC+600-mg AZT+600-mg EFV per day; 3TC+AZT+LPV/r, 300-mg 3TC+600-mg AZT+1000-mg LPV/r per day.

^aHigh adherence represents days with oral drug administration > 27 days/month and low adherence represents days ≤ 27 days/month.

^bThe difference in gender distribution was examined using the Chi-square test, and the differences in age and adherence were examined by t-test for two independent samples. The inter-group comparisons in the drug contents were examined with the t-test for two independent samples and covariance analysis with gender and age as covariates.

^cAdherence for patients was defined as the duration of antiretroviral drug administration within the past month. Data are presented as mean ± standard deviation (M±SD) and range.

^dThe antiretroviral drug content was presented as M±SD and range where the content below the limits of quantitation (LOQ) but over the limit of detection (LOD) was replaced with LOD and the content below LOD with 0.

^eFor patients using 3TC+AZT+NVP with low adherence, one participant has a 3TC content below LOD. Four participants have an AZT content below LOQ but over LOD, and five participants have an AZT content below LOD. Two participants have a NVP content below LOQ but over LOD, and one participant has an NVP content below LOD.

^fFor patients using 3TC+AZT+EFV with low adherence, one participant has a 3TC content below LOQ but over LOD, and three participants have a 3TC content below LOD. Three participants have an AZT content below LOQ but over LOD, and six participants have an AZT content below LOD. Two participants have an EFV content below LOQ but over LOD, and one participant has an EFV content below LOD.

^gLPV/r is the drug tablet combined with RTV and LPV.

^hFor patients using 3TC+AZT+LPV/r with low adherence, one participant has a 3TC content below LOD. Three participants have an AZT content below LOD. Two participants have an RTV content below LOD, and one participant has an LPV content below LOD.

Similarly, for patients receiving 3TC+AZT+EFV, those with high-adherence were mostly male when compared with those presenting low-adherence (p<0.05), as examined by the Chi-square test, while displaying significantly more days of oral administration within the past month (p<0.001), with no significant difference in terms of age (p>0.05). They showed significantly higher hair contents of 3TC, AZT, and EFV (ps<0.01), retaining significantly higher contents of the three drugs (ps<0.001), as examined by covariance analysis with gender and age as covariates, where gender did not influence the contents of 3TC, AZT, and EFV (ps>0.495), and age did not influence AZT and EFV (ps>0.165), but influenced 3TC (p=0.026).

For patients using 3TC+AZT+LPV/r, those presenting high-adherence exhibited significantly more days of oral administration over the past month (p<0.001), with no significant differences between them in terms of gender and age (ps>0.05). They showed significantly higher hair contents of AZT, RTV, and LPV (ps<0.05) than those with low-adherence, except for 3TC (p>0.05); however, intergroup differences were significant for all four drugs (ps<0.05) as shown by covariance analysis with gender and age as covariates, where gender influenced AZT, RTV and LPV (p<0.041), with a trend to influence 3TC (p=0.099), and age influenced 3TC and RTV (ps<0.044), with a trend to influence AZT and LPV (ps<0.093).

Furthermore, for 3TC and AZT that all patients administered, patients with high-adherence showed significantly higher contents when compared with those with low-adherence, as examined by covariance analysis with gender, age, and the type of ART regimen as covariates (ps<0.001, see §3 and Table S5 in Supplemental Materials).

Additionally, the inter-correlation in hair contents of two drugs from the identical regimen was validated to verify the method’s reliability (see §4 and Table S6 in Supplemental Materials).

Discussion

This study developed and validated a simple and sensitive LC-ESI⁺-MS/MS method for simultaneously determining six antiretroviral drugs, 3TC, AZT, NVP, EFV, RTV, and LPV, in human hair. To the best of our knowledge, this is the first successful attempt to simultaneously determine these six drugs in hair, although they have been previously determined in blood using high-resolution MS.³⁶ This is also the first ESI method for determining hair AZT, exhibiting a better performance than previous APCI method. Additionally, this method showed reliability in comprehensively and accurately assessing adherence in PLWH administering different ART regimens.

This study displayed obvious improvements in LOQs. This could result from the optimization of ion sources and sample preparation as follows. First, ESI was used as the ionization source for MS detection instead of APCI, which enhances the responses of polar drugs and shows good sensitivity for low-polarity drugs. Accordingly, this ESI method showed better LOQ for AZT and matched LOQs for EFV, RTV, and LPV when compared with previous APCI method. Second, the mixture with methanol at a markedly high ratio (i.e., 95%) was used as the mobile phase in this study. Previously, a lower ratio of organic reagents was typically used for chromatographic separation of drugs, including 45% methanol for 3TC,²⁴ 50% acetonitrile for NVP,²³ 55% acetonitrile for EFV,³¹ and 65% acetonitrile for RTV and LPV.³¹ However, methanol at a higher ratio helps elution of low-polarity drugs from the injected sample solution and chromatographic column stationary phase owing to the theory of “like dissolves like.” Moreover, methanol enhances the ionization of compounds.³³ Third, this study utilized 10 mg as the optimal hair weight for determining the six drugs. Previously, a 2-mg hair weight was selected for determination of 3TC,^{18, 24} NVP,^{18, 23} EFV,³¹ RTV, and LPV.³¹ Moreover, a method employing 0.2-mg hair samples for EFV²⁵ has been reported. It postulated that small hair weight facilitates sample collection and reduces the matrix effect; however, the method’s sensitivity would be decreased, as ascertained previously.¹⁸ Additionally, the LOQ of EFV based on 0.2-mg samples²⁵ was reportedly higher than that of others obtained using 2-mg³¹ and 20-mg¹⁹ samples. Finally, to enhance the extraction efficiency of drugs from hair samples, this study followed an optimized incubation process (i.e., incubation with methanol for 16 h at 37°C), as previously confirmed.¹⁸

This study further revealed that patients with high-adherence had significantly higher hair contents than those with low-adherence for all drugs from the identical regimen (Table 4). Notwithstanding the use of different regimens, patients with high-adherence presented significantly higher hair 3TC and AZT contents than those with low-adherence (Table S5). These results indicated that the hair content of all drugs from the identical regimen was consistent with self-reported adherence. It was inferred that determining the hair contents of all drugs from the identical regimen would be more comprehensive for assessing long-term adherence than that of a single drug. Additionally, our previous study has determined that the hair content of AZT or EFV was positively related to adherence in PLWH.¹⁹ This indicated that this method was reliable in comprehensively and accurately assessing ART adherence. However, this study noted the influence of age and gender on the hair drug content for assessing adherence in PLWH, especially for 3TC, as discussed above. This might be mainly attributed to the differences in drug metabolism related to age and life habits between genders (e.g., the frequency of hair washing and care). Some previous studies have corroborated that older adults show higher hair antiretroviral drugs contents than young adults owing to differences in drug metabolism abilities.^{37, 38} Furthermore, gender was reportedly associated with hair antiretroviral drugs contents.^{37, 39} In contrast, some inconsistent findings that showed no association between hair drug content and age or gender have also been reported (see the review⁴⁰). Thus, it is necessary to reclarify the associations of hair drug content with age and gender in a future large-scale cohort study.

This study further investigated whether strong and even markedly strong positive inter-correlations exist in the hair content between any two drugs from an identical regimen (Table S6). Furthermore, 3TC and AZT showed strong correlations among patients using different regimens contained 3TC and AZT (Table S5). This indicated that the hair contents of different drugs might show high consistency in assessing ART adherence, although large inter-individual and inter-drug differences were observed in drug contents.^{18, 19} Notably, RTV and LPV showed the highest coefficient of correlation among the six drugs (Table S6). This was mainly because RTV and LPV are protease inhibitors possessing similar chemical structures and physicochemical properties, including lipophilicity.^{29, 41} Accordingly, they would have the same mechanism of in vivo metabolism, incorporation into hair, and dissolution from hair samples with methanol owing to their similarity. Moreover, they were combined as a single tablet (i.e., LPV/r) for oral administration; their in vivo metabolic processes were synchronous. Thus, RTV and LPV demonstrated markedly strong inter-correlations in the hair content for patients using LPV/r. Additionally, the strong inter-correlation between AZT and EFV or RTV and LPV in this study was consistent with those observed in our previous study (i.e., rs=0.644 and 0.616 vs. r=0.52, ps<0.05, and rs=0.964 and 0.958 vs. r=0.80, ps<0.01,¹⁹). This could provide evidence for this method’s reliability.

Although we successfully developed an LC-ESI⁺-MS/MS method for simultaneous determination of the six drugs in hair and applied it for natural hair sample analysis in PLWH, this study has several limitations. First, only six mainstream ARV drugs in China were identified in our study. Second, the study only recruited a small number of PLWH from a particular region. Lastly, this study briefly reported the effects of sociodemographic factors related to hair drug content (e.g., age and gender) on adherence assessment. Although consistent with some previous studies, the controversies regarding these data persisted owing to the small sample size. Therefore, future investigations should develop a reliable method to simultaneously determine as many drugs as feasible, as well as utilize it for adherence assessment of a large-scale cohort of PLWH from multiple regions worldwide. Furthermore, the method should explore the impact of sociodemographic factors on adherence assessment using a large cohort of PLWH, to provide additional guidelines for assessing adherence based on hair contents of antiretroviral drugs.

Conclusions

In this study, a simple and sensitive LC-ESI⁺-MS/MS method was developed to simultaneously determine six antiretroviral drugs (i.e., 3TC, AZT, NVP, EFV, RTV, and LPV) in human hair. The developed method exhibited good performance during validation, meeting the acceptance criteria of applicable FDA and EMA guidelines. The population analysis revealed that this method was reliable in assessing adherence in PLWH administering different ART regimens. These present results might provide some guidance for comprehensively assessing long-term ART adherence using multiple hair drug contents, instead of a single hair drug content in clinical settings.