Current methods of analyzing patients’ adherence to complex drug resistant-tuberculosis (DR-TB) regimens can be inaccurate and resource-intensive. Our method analyzes hair, an easily collected and stored matrix, for concentrations of 11 DR-TB medications. Using LC-MS/MS, we can determine sub-nanogram drug levels that can be utilized to better understand drug adherence.
Drug resistant-tuberculosis (DR-TB) is a growing public health threat, and assessment of therapeutic drug levels may have important clinical benefits. Plasma drug levels are the current gold standard assessment, but require phlebotomy and a cold chain, and capture only very recent adherence. Our method uses hair, a matrix that is easily collected and reflective of long-term adherence, to test for 11 anti-TB medications. Previous work by our group shows that antiretroviral drug levels in hair are associated with HIV outcomes. Our method for DR-TB drugs uses 2 mg of hair (3 cm proximal to the root), which is pulverized and extracted in methanol. Samples are analyzed with a single LC-MS/MS method, quantifying 11 drugs in a 16 min run. Lower limits of quantification (LLOQs) for the 11 drugs range from 0.01 ng/mg to 1 ng/mg. Drug presence is confirmed by comparing ratios of two mass spectrometry transitions. Samples are quantified using the area ratio of the drug to the deuterated, 15N-, or 13C-labeled drug isotopologue. We used a calibration curve ranging from 0.001-100 ng/mg. Application of the method to a convenience sample of hair samples collected from DR-TB patients on directly observed therapy (DOT) indicated drug levels in hair within the linear dynamic range of nine of the eleven drugs (isoniazid, pyrazinamide, ethambutol, linezolid, levofloxacin, moxifloxacin, clofazimine, bedaquiline, pretomanid). No patient was on prothionamide, and the measured levels for ethionamide were close to its LLOQ (with further work instead examining the suitability of ethionamide’s metabolite for monitoring exposure). In summary, we describe the development of a multi-analyte panel for DR-TB drugs in hair as a technique for therapeutic drug monitoring during drug-resistant TB treatment.
In the twenty-first century, drug-resistant TB (DR-TB) is an evolving catastrophe for already weak national TB control programs, with confirmed cases doubling in the past 5 years alone, accounting for nearly one-third of all deaths related to antimicrobial resistance globally1,2. Successful treatment of DR-TB has conventionally required longer and more toxic second-line regimens than treatment for drug-sensitive TB. Moreover, patients with DR-TB often have significant pre-existing challenges to adherence, which contributed to the emergence of resistance initially3.
Unlike HIV infection where viral loads can be used to monitor treatment, surrogate endpoints of treatment response in TB are delayed and unreliable on an individual level4. Monitoring patient adherence, an important predictor of subtherapeutic anti-TB drug concentration and treatment failure, is also challenging. Self-reported adherence suffers from recall bias and the desire to please providers5,6. Pill counts and medication event monitoring systems (MEMS) can be more objective7 but do not measure actual drug consumption8,9,10. Drug levels in biomatrices can provide both adherence and pharmacokinetic data. Therefore, plasma drug levels are commonly used in therapeutic drug monitoring11,12. In the context of drug adherence monitoring, however, plasma levels represent short-term exposure and are limited by significant intra- and inter-patient variability when determining appropriate adherence reference range. “White coat” effects, where adherence improves prior to clinic or study visits, further complicates the ability of plasma levels to provide accurate drug adherence patterns13.
Hair is an alternative biomatrix that can measure long-term drug exposure14,15. Many drugs and endogenous metabolites incorporate into the hair protein matrix from the systemic circulation as hair grows. As this dynamic process continues during hair growth, the amount of drug deposited in the hair matrix depends on the continuous presence of the drug in circulation, making hair an excellent temporal readout of drug intake. Hair as a biomatrix has the additional advantage of being easily collected without the need for cold chain for storage and shipment compared to blood. Moreover, hair is non-biohazardous, which provides additional feasibility advantages in the field.
Hair drug levels have long been used in forensic applications16. Over the last decade, hair antiretroviral (ARV) levels have demonstrated utility in assessing drug adherence in HIV treatment and prevention, to which our group contributed. ARV levels in hair have been shown to be the strongest independent predictors of treatment outcomes in HIV infection17,18,19,20,21. To determine whether hair levels of DR-TB patients will have the same utility in predicting treatment outcome, we used LC-MS/MS to develop and validate a method for analyzing 11 DR-TB medications in small hair samples. As an initial assessment of the assay’s performance, we measured DR-TB drugs levels in a convenience sample of patients with DR-TB receiving directly observed therapy (DOT) in the Western Cape, South Africa22.
All patients provided written informed consent prior to hair sample collection. We obtained Institutional Review Board approval from the University of Cape Town and the University of California, San Francisco.
1. Hair sampling
2. Drug extraction
3. LC-MS/MS preparation
4. Data analysis
An illustration of a chromatogram with confirmed levels of all 11 DR-TB drugs is shown in Figure 1. The retention time for each analyte can change when using different instruments and columns, so the exact retention time should be determined individually.
The Extracted Ion Chromatograms (EICs) for one particular drug (isoniazid, INH) in one of the calibrators (blank hair sample spiked with DR-TB drug reference standards) are shown in Figure 2. The quantifier and qualifier transitions are used to qualitatively confirm the presence of the drug, as the ratio between area of quantifier and area of qualifier remains constant across samples. The internal standard is also monitored to ensure that each sample injection is normalized.
For purposes of demonstration, we analyzed a convenience sample of 15 hair samples among a total study population of 96 patients taking DR-TB drugs under DOT conditions from Western Cape, South Africa. Table 6 presents representative levels of DR-TB drugs across the lowest and highest levels measured for each analyte. Although data for 15 patient samples are presented, each analyte did not have 15 levels reported because each patient is on a different combination of DR-TB medications. None of the patients were on prothionamide, and only a single patient was taking pretomanid.
Figure 1. An illustration of a representative chromatogram showing peaks of the 11 analytes in the DR-TB method (EMB= ethambutol; INH= isoniazid; PZA= pyrazinamide; ETH= ethionamide; PTH= prothionamide; LFX= levofloxacin; MFX= moxifloxacin; LZD= linezolid; PTM= pretomanid; BDQ= bedaquiline; CLF= clofazimine). Because the sensitivity of the method for each analyte is different, INH, LZD, LFX, MFX and LZD were spiked at 20 ng/mg hair while BDQ, CLF, EMB, ETH, PTH and PTM were spiked at 2 ng/mg hair. Please click here to view a larger version of this figure.
Figure 2. Two Extracted Ion Chromatograms (EICs) from an injection of calibration point 9 (C9), isoniazid (INH) at 20 ng/mg. The top EIC shows both the INH quantifier transition (blue, labeled INH-2) and the INH qualifier transition (red, labeled INH-3). The bottom EIC shows response of INH-d4, the internal standard (IS) used to quantify INH. Please click here to view a larger version of this figure.
Figure 3. Screenshots of the process of quantitation. The top portion is a partial sample list showing injection data for one analyte (INH, isoniazid) across 12 calibration points (labeled C0-C11), three QC levels, and six samples. Bottom left portion is the calibration curve, ranging from 0.5 ng/mg –100 ng/mg. Opaque blue dots are calibration points. Transparent blue squares are quality control points. The R-value is shown in the top left (0.99722) with weighting 1/x. The two chromatograms in the bottom right illustrate a sample with INH (top chromatogram) and a sample without INH (bottom chromatogram). Please click here to view a larger version of this figure.
Drugs in each mix | Concentration of each drug in mix | Volume of mix added to 50mL vol. flask |
Mix 1: CLF-d7, EMB-d4 | 10 μg/mL | 40 μL |
Mix 2: LFX-d8, PTH-d5 | 10 μg/mL | 10 μL |
Mix 3: BDQ-d6, LZD-d3, MFX 13C-d3, OPC (IS for PTM) | 10 μg/mL | 20 μL |
Mix 4: PZA 15N-d3 | 10 μg/mL | 200 μL |
Mix 5: INH-d4 | 10 μg/mL | 100 μL |
Table 1. Concentration and amount of each internal standard to add to a 50 mL volumetric flask.
Drug | Stock concentration | Volume added |
BDQ | 0.5 mg/mL | 8 μL |
CLF | 0.5 mg/mL | 8 μL |
EMB | 1 mg/mL | 4 μL |
PTH | 1 mg/mL | 4 μL |
PTM | 1 mg/mL | 4 μL |
INH | 1 mg/mL | 40 μL |
LFX | 1 mg/mL | 40 μL |
LZD | 1 mg/mL | 40 μL |
MFX | 1 mg/mL | 40 μL |
PZA | 1 mg/mL | 40 μL |
Table 2. Amount of each drug reference standard to add to “Ref Std Mix 1” vial.
Label name | Vial drawn from | Volume added |
C0 | N/A | 0 μL |
C1 | Ref Std Mix df1000 | 5 μL |
C2 | Ref Std Mix df1000 | 10 μL |
C3 | Ref Std Mix df1000 | 20 μL |
C4 | Ref Std Mix df100 | 5 μL |
C5 | Ref Std Mix df100 | 10 μL |
C6 | Ref Std Mix df100 | 20 μL |
C7 | Ref Std Mix df10 | 5 μL |
C8 | Ref Std Mix df10 | 10 μL |
C9 | Ref Std Mix df10 | 20 μL |
C10 | Ref Std Mix df1 | 5 μL |
C11 | Ref Std Mix df1 | 10 μL |
Table 3. Amount of each Ref Std Mix intermediate to add to the 12 calibration points.
Total Time (min) | Flow Rate (μL/min) | A (%) | B (%) |
0 | 450 | 95 | 5 |
0.3 | 450 | 95 | 5 |
2.3 | 450 | 0 | 100 |
5 | 550 | 0 | 100 |
11 | 550 | 0 | 100 |
11.1 | 550 | 95 | 5 |
13 | 450 | 95 | 5 |
16.75 | 450 | 95 | 5 |
Table 4. The flow rate and mobile phase gradient used for each injection.
Calibration point | Actual concentration of BDQ, CLF, ETH, EMB, PTH, PTM (ng/mg) | Actual concentration of INH, LFX, LZD, MFX, PZA (ng/mg) |
C0 | 0 | 0 |
C1 | 0.005 | 0.05 |
C2 | 0.01 | 0.1 |
C3 | 0.02 | 0.2 |
C4 | 0.05 | 0.5 |
C5 | 0.1 | 1 |
C6 | 0.2 | 2 |
C7 | 0.5 | 5 |
C8 | 1 | 10 |
C9 | 2 | 20 |
C10 | 5 | 50 |
C11 | 10 | 100 |
Table 5. Final concentration of analytes in each calibration point.
Drug | LOD (ng/mg hair) |
LLOQ (ng/mg hair) |
ULOQ (ng/mg hair) |
Sample values (ng/mg hair) Samples: UC-04, UC-08, UC-11, UC-16, UC-25, UC-36, UC-69, UC-83, UC-89, UC-90, UC-91, UC-104, UC-105, UC-108, UC-109 |
Bedaquiline | 0.005 | 0.05 | 10 | 0.21, 0.38, 0.56, 0.86, 0.90, 1.04, 1.29, 2.15, 2.29, 5.64 |
Clofazimine | 0.005 | 0.05 | 10 | 0.37, 0.61, 1.84, 2.20, 2.90, 3.41, 3.90, 6.03, 8.25, 10.66, 11.01 |
Ethambutol | 0.005 | 0.05 | 10 | 0.04, 0.05, 0.25, 0.42, 0.43, 0.5, 0.68, 0.92, 0.95, 1.01, 1.53, 1.54, 9.76 |
Ethionamide | 0.01 | 0.01 | 10 | <LOD, <LOD, 0.01, 0.01, 0.01, 0.02, 0.02, 0.17 |
Isoniazid | 0.05 | 0.5 | 100 | <LOD, <LOD, 0.12, 0.26, 0.84, 0.94, 1.36, 2.88, 4.03, 4.04, 9.14 |
Levofloxacin | 0.1 | 0.5 | 100 | 8.01, 8.42, 15.37, 24.41, 39.45, 42.12, 56.15, 75.58, 119.96 |
Linezolid | 0.1 | 0.5 | 100 | 0.87, 1.09, 3.51, 5.51, 7.80, 9.21, 15.68, 18.32, 19.13, 21.22 |
Moxifloxacin | 0.05 | 0.5 | 100 | 0.35, 0.49, 1.58, 1.59, 6.23, 7.06, 13.14, 17.37, 21.72, 55.88, 86.64 |
Pretomanid | 0.005 | 0.05 | 10 | 0.57 |
Prothionamide | 0.002 | 0.01 | 10 | |
Pyrazinamide | 0.05 | 1 | 100 | 1.14, 1.74, 1.86, 3.21, 5.94, 11.39, 12.36, 12.71, 12.85, 14.38, 16.13, 44.17, 69.66 |
Table 6. Representative levels of drugs measured in 15 patients taking DR-TB medications under DOT. The limit of detection (LOD), lower limit of quantification (LLOQ) and upper limit of quantification (ULOQ) of the method for each drug are given for comparison.
We report here the protocol for the method we developed and validated for quantifying 11 anti-TB medications utilized in the treatment of DR-TB in small hair samples using LC-MS/MS. No other method for quantifying these 11 drugs in hair has been previously developed, validated and published. Our method can quantify sub-nanogram levels of drugs in only 20-30 hair strands of approximately 3 centimeters (cm) in length (~2 mg) and has already been validated22. The low weight of hair analyzed means that patients involved in the study can participate discreetly and potentially return for repeat testing without fear of exposing bald scalp. We have previously published data on the association between DR-TB drug levels in hair and DR-treatment outcomes23. Therefore, the development and validation of this multi-analyte panel method represents a significant advance in the field of DR-TB therapeutic drug monitoring.
Hair requires different homogenization techniques than those required with liquid biomatrices. Pulverization of hair strands allowed efficient access of extraction solvent to analytes in the hair matrix. Thus, one important feature of our method is the quick and easy extraction process of drugs from hair using the pulverized samples. Incubation time during the extraction process is only two h, due to the large accessible surface area of pulverized hair, and there is no clean up step, due to the small sample size (2 mg). Care must be taken, though, to limit drug degradation during the extraction process. The protocol uses a two-cycle pulverization, with a 45 s cooling period in between the cycles. This process avoids overheating and potentially degrading the drugs in the hair.
Unlike many hair analyses for drugs of abuse, this method does not use a washing step. DR-TB drugs come in capsule or tablet form, limiting possible sources of external contamination and the subsequent need to wash hair prior to analysis. Future studies could analyze wash solvent from DR-TB patient hair to assess external contamination.
Although hair pulverization promotes efficient drug extraction, it has its own limitations. Our laboratory has found that if hair is pulverized in the bead ruptor and left at room temperature, the concentration of some of the 11 drugs decreases over weeks and months. This may be due to the large surface area of the pulverized hair exposed to the atmosphere that can promote oxidation and other degradation reactions. If a stability study of the drugs in hair is desired, hair can be cut with scissors into small segments of <1 cm, homogenized by hand, and then left at room temperature for weeks or months during the stability study. When this cut hair is pulverized on the day of analysis, we have not observed any significant drug degradation over time. Hence, in performing the described protocol, we recommend that hair be pulverized on the day it is extracted. Likewise, all drug mixes below 10 µg/mL concentrations should be prepared on the day of extraction.
No previously published methods are available to assess the suitability of the linear dynamic ranges (LLOQ-ULOQ) we established for each TB drug in the multi-analyte method. However, the convenience sample of hair samples from Western Cape, South Africa, indicates the suitability of the linear dynamic range of this method. With the exception of ethionamide, pretomanid, and prothionamide, more than 95% of the drug levels we measured in these patients are within the linear dynamic range of each analyte. Only one patient was taking pretomanid (which was detected), and no patients were taking prothionamide. For ethionamide, we hypothesize that the drug may not deposit to the hair matrix well, as our LOD is 0.01 ng/mg hair (or 10 pg/mg hair) and yet only one of the eight patients taking ethionamide has levels greater than 0.02 ng/mg hair. Further examination is warranted to determine the pharmacokinetics of different TB drugs in hair. For example, a potential alternative for monitoring drugs like ethionamide is to develop a method targeting their metabolite(s) instead. We have made a similar observation for delamanid, a novel DR-TB medication, which was initially part of this panel. A method targeting delamanid’s metabolite is currently in the process of being validated in our laboratory, because the metabolite is found in higher concentrations than the parent drug. The same procedure can be performed for ethionamide. The drug concentrations in Table 6 are presented as a group because the individual results and clinical outcomes are not the focus of this method paper. Individual assessment of this group of patients has been published elsewhere23.
The patients contributing small hair samples for the demonstration study were administered a variety of drug regimens via DOT in an inpatient setting, and all regimens were documented according to nursing records during the inpatient period. However, as is common among DR-TB patients, previous, poorly documented drug regimens had also been administered prior to their inpatient stay. This led to detection of drugs in patient hair that were not noted on their inpatient records. Therefore, we could not use these samples to determine specificity of the method, as we could not determine if these samples were truly false positives. Instead, we tested hair from patients who were not taking DR-TB drugs. No DR-TB drugs were detected in these samples, indicating that the method is specific.
Although our method demonstrates the utility of using hair in measuring DR-TB drugs, hair analysis has its own set of limitations. Because hair is a solid matrix, spiking of drug reference standards during method validation does not allow for the standards’ full integration into the matrix as with urine and blood. Thus, recovery assessment is limited to detection of drug after spiking onto the solid matrix, and not actual retrieval from the matrix. Likewise, because hair is an alternative matrix that is still being explored for testing, no readily available reference ranges for medications are available to assess method suitability. More pharmacokinetic studies on the incorporation of drugs into hair will be useful to further understand the utility of hair drug levels in adherence monitoring. Finally, the proper collection of hair samples at field sites has its own unique challenges. While collection and storage of hair samples requires fewer resources than other biomatrices, care must be taken to identify the distal and proximal ends of any hair strands longer than 2 cm. Longer hair strands may have different drug concentrations along the strand, depending on medication use over time. Proper labeling allows for analysis of specific segments of the strands; in the case of our method, the three centimeters of hair closest to the scalp was used to determine the most recent data on medication adherence. Proper labeling requires training and quality assurance procedures at the sites.
In summary, we have developed the first multi-analyte panel for analyzing TB medications used for DR-TB via LC-MS/MS in small hair samples. Given the feasibility of collecting and storing hair in resource-limited settings, our method represents a potentially significant advance in the field of TB therapeutic drug monitoring. Objective measures of drug exposure that take into account both adherence and individual pharmacokinetic variability may provide early indication of ineffective treatment regimens, thereby aiding both individual treatment as well as limiting community transmission of DR-TB24.
The authors have nothing to disclose.
The authors would like to thank Professor Keertan Dheda, Dr. Ali Esmail, and Marietjie Pretorius at the University of Cape Town Lung Institute who facilitated the collection of hair samples for the study. The authors further gratefully acknowledge the contributions of the participants of this study.
2 mL injection vials | Agilent Technologies | 5182-0716 | |
250 uL injection vial inserts | Agilent Technologies | 5181-8872 | |
Bead ruptor 24 | OMNI International | 19001 | |
Bead ruptor tubes (2 mL bead kit, 2.8mm ceramic, 2 mL microtubes) | OMNI International | 19628 | |
Bedaquiline | Toronto Research Chemicals | B119550 | |
Bedaquiline-d6 | Toronto Research Chemicals | B119552 | |
Clofazimine | Toronto Research Chemicals | C324300 | |
Clofazimine-d7 | Toronto Research Chemicals | C324302 | |
Disposable lime glass culture tubes | VWR | 60825-425 | |
Ethambutol | Toronto Research Chemicals | E889800 | |
Ethambutol-d4 | Toronto Research Chemicals | E889802 | |
Ethionamide | Toronto Research Chemicals | E890420 | |
Ethionamide-d5 | ClearSynth | CS-O-06597 | |
Formic acid | Sigma-Aldrich | F0507-100mL | |
Glass bottles | Corning | 1395-1L | |
Hot Shaker | Bellco Glass Inc | 7746-32110 | |
HPLC | Agilent Technologies | Infinity 1260 | |
HPLC grade acetonitrile | Honeywell | 015-4 | |
HPLC grade methanol | Honeywell | 230-1L | |
HPLC grade water | Aqua Solutions Inc | W1089-4L | |
Isoniazid | Toronto Research Chemicals | I821450 | |
Isoniazid-d4 | Toronto Research Chemicals | I821452 | |
LC column, Synergi 2.5 um Polar RP 100 A 100 x 2 mm | Phenomenex | 00D-4371-B0 | |
LC guard cartridge | Phenomenex | AJ0-8788 | |
LC guard cartridge holder | Phenomenex | AJ0-9000 | |
LC-MS/MS quantitation software | Sciex | Multiquant 2.1 | |
Levofloxacin | Sigma-Aldrich | 1362103-200MG | |
Levofloxacin-d8 | Toronto Research Chemicals | L360002 | |
Linezolid | Toronto Research Chemicals | L466500 | |
Linezolid-d3 | Toronto Research Chemicals | L466502 | |
Micro centrifuge tubes | E&K Scientific | 695554 | |
Moxifloxacin | Toronto Research Chemicals | M745000 | |
Moxifloxacin-13C, d3 | Toronto Research Chemicals | M745003 | |
MS/MS | Sciex | Triple Quad 5500 | |
OPC 14714 | Toronto Research Chemicals | O667600 | |
Pretomanid (PA-824) | Toronto Research Chemicals | P122500 | |
Prothionamide | Toronto Research Chemicals | P839100 | |
Prothionamide-d5 | Toronto Research Chemicals | P839102 | |
Pyrazinamide | Toronto Research Chemicals | P840600 | |
Pyrazinamide-15N, d3 | Toronto Research Chemicals | P840602 | |
Septum caps for injection vials | Agilent Technologies | 5185-5862 | |
Turbovap LV evaporator | Biotage | 103198/11 |