FormalPara Key Points

Dried blood spots with a reduced sample volume requirement, high sample stability, and low shipping costs facilitate therapeutic drug monitoring in remote settings using a centralized laboratory service.

Simple semi-quantitative methods using urine or saliva can serve as point-of-care testing in high-burden settings.

Hair samples can provide information on drug exposure over a longer period of time.

1 Introduction

Anti-tuberculosis (TB) drugs act in a concentration-dependent manner and suboptimal circulating drug concentrations have been associated with poor outcomes, including acquired drug resistance [1,2,3,4]. Individual pharmacokinetic variability is difficult to predict without direct measurement, and early detection of suboptimal drug concentrations enables clinicians to optimize the dose to prevent treatment failures and avoid adverse effects due to toxic drug concentrations [5].

Measuring drug concentrations via serum or plasma has been considered the gold standard for therapeutic drug monitoring (TDM). However, TDM poses many challenges such as uncomfortable sampling methods, requirement for highly trained personnel from the sample collection to analysis, and dry-ice shipping, which all lead to high costs or a lack of availability in TB-endemic settings where poor treatment outcomes are more common and TDM may be of the most benefit [6]. Performing TDM with dried blood spots (DBS), urine, saliva, and hair, in lieu of regular serum or plasma sampling is gaining popularity owing to the relatively simple sample collection and specimens that do not require cold-chain transport [7, 8].

For DBS, a single drop of blood obtained via an automatic lancet can be collected by healthcare workers or patients themselves with minimal discomfort and without the need for trained phlebotomists [9]. The small sample volume makes this method more suitable in pediatric patients as well as patients unable to undergo large-volume venous sampling during intensive pharmacokinetic studies [10]. Dried blood spot cards can be shipped at an ambient temperature reducing the need for shipments on dry ice, thereby reducing shipping costs [9, 10].

Urine collection offers an inexpensive point-of-care testing option with minimal processing for quantifying the excretion of drugs with known and relatively fixed proportions of renal elimination [11,12,13]. For example, colorimetric methods to qualitatively detect isoniazid in urine by the Arkansas method have been commercialized (IsoScreen; GFC Diagnostics Ltd, Oxforshire, UK) and used extensively to estimate adherence in patients with active TB or latent TB infection, or in patients receiving isoniazid preventative therapy [14]. Colorimetric analytical procedures have also been developed to quantitatively measure rifampin, pyrazinamide, and levofloxacin in the urine of patients with TB [11,12,13, 15].

Similarly, saliva offers another biomatrix with simple sampling methods that may be more cost effective, with the ability to be implemented across a wide variety of patient populations [16, 17]. Saliva is a low-protein matrix and the drug concentrations quantified in this matrix may more accurately reflect the proportion of medication that is non-protein bound [16]. The ability of many anti-TB drugs to be distributed into oral fluid makes saliva a promising alternative matrix for performing drug monitoring in the field with simple equipment and very little extra processing [17,18,19,20].

DBS, urine, and saliva metrics provide snapshots of drug concentrations either at one timepoint or over one dosing interval that can be used to estimate important pharmacokinetic parameters such as peak concentration (Cmax) and the total area under the concentration–time curve (AUC) for a dosing interval. However, cumulative exposure throughout the treatment period is not captured by these metrics. Measuring drug concentrations in hair, especially of drugs with short half-lives such as isoniazid [21] and linezolid [22], may be more representative of long-term pharmacokinetic exposure that is dependent upon the four parameters of absorption, distribution, metabolism, and elimination, but also patterns of adherence to prescribed medications, a potentially important feature for anti-TB care where treatment courses are long [23].

The aim of this systematic review was to assess the current state of knowledge of studies comparing TB medication concentrations in DBS, urine, saliva, and hair with plasma or serum concentrations, define the product development stage of these methods based on the published literature, and explore if TDM using these alternative matrices would be feasible for anti-TB care in programmatic settings.

2 Methods

First-line and second-line anti-TB drugs were included in this systematic search [24]. PubMed and Web of Science were searched in May 2022 for the keywords (isoniazid OR rifampin OR pyrazinamide OR ethambutol OR rifapentine OR levofloxacin OR moxifloxacin OR gatifloxacin OR amikacin OR capreomycin OR kanamycin OR streptomycin OR ethionamide OR prothionamide OR cycloserine OR terizidone OR linezolid OR clofazimine OR bedaquiline OR delamanid OR pretomanid OR paraaminosalicylic acid OR imipenem/cilastatin OR imipenem OR cilastatin OR meropenem OR amoxicillin/clavulanate OR amoxicillin OR clavulanate OR thiacetazone) AND (saliva OR urine OR hair OR dried * spot OR volumetric absorptive microsample*) AND (tuberculosis OR TB). There was no limit on publication dates. Reproducibility of results was checked by a second reviewer by conducting a search using the same keywords. Two independent reviewers screened titles and abstracts for eligibility after duplicates were removed. A full-text review was performed on the remaining reports and articles. Non-human studies, commentaries, and studies that did not collect DBS, urine, saliva, or hair samples were excluded. References were screened to include relevant articles. The Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA) was used for this review [25].

Data extraction was performed to determine study population, sample size, sampling, analytical methods used to determine concentrations, comparative serum and/or plasma concentrations, and if the obtained results were used to perform drug monitoring. Ratios of the concentrations of individual drugs within the alternate biomatrix to serum and/or plasma concentrations were calculated if Cmax or AUC values were available.

Risk of bias was assessed for all included studies using the Risk Of Bias In Nonrandomized Studies-of Interventions (ROBINS-I) tool, which evaluates the risk of bias in estimates of effectiveness or safety of an intervention from studies that did not use randomization to allocate interventions [26]. As no validated tool for risk bias assessment was available for pharmacokinetic studies, ROBINS-I was adopted by making changes to the classification of interventions and deviations from intended interventions sections, as they were not applicable to pharmacokinetic studies. For each drug under every biomatrix, the technology readiness level (TRL) was assessed from a scale of 1 (basic research) to 9 (launch operations) [27], and details on the level assessment are described in Table 1.

Table 1 TRL to test readiness of implementing alternative biomatrix in programmatic settings
Full size table

3 Results

A total of 671 articles were found in PubMed and 335 in Web of Science for the search terms resulting in 777 articles after 229 duplicate reports were removed. Of the remaining articles, 648 records were excluded as they were not relevant based on title and abstract screening. A full-text assessment was performed for 129 articles and 58 articles were excluded for reasons stated in Fig. 1. Four articles were included from searching references, leading to a final total of 75 articles included in the systematic review.

Fig. 1
figure 1

Flowchart of the search of reports included in this systematic review. Chart from Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. doi: 10. 1136 / bmj. n71 (http://www.prisma-statement.org)

Full size image

3.1 DBS

Table 2 summarizes the information of studies focusing on the development and validation of a bioanalytical method to quantify anti-TB drugs in the DBS matrix (n = 8). The majority of studies (87.5%) were performed on patients with TB while one study was conducted in healthy volunteers [28], and one in pediatric patients [29]. Most of the studies were small in size and ranged from 6 to 26 subjects. Plasma and DBS samples were collected 1 week to 10 days after treatment initiation. Dried blood spots were generated through a finger prick [29, 30] or pipetting venous dried blood spot (VDBS) onto paper, [28, 31] or both [32, 33]. For comparative plasma samples, both intensive and sparse sampling methods were applied, while finger-prick DBS specimens were mostly collected by a sparse sampling strategy. Considering the quantification method, liquid chromatography-tandem mass spectrometry was the most common apparatus applied for both DBS and plasma matrices (seven of eight studies). The methods were also validated with criteria according to the guidelines for bioanalytical method validation with accuracy and precision ≤ 20% relative error and coefficient of variation respectively for quality-control samples at the lower limit of quantification and ≤ 15% for other quality-control samples. As clinical validation is highly recommended [9], the agreement between DBS and plasma analysis data was assessed in all studies including two or more methods such as simple linear regression, Passing–Bablok regression, Deming regression, Bland–Altman plots, and predictive performance of plasma concentrations from DBS. Pharmacokinetic parameters, including Cmax and/or AUC, were calculated for plasma and/or DBS in two studies [29, 33]. Information on sample size [34], the duration between treatment initiation and sample collection [30, 32, 34, 35], DBS sampling times [34, 35], and drug concentrations [28,29,30,31,32, 34, 35] was not provided in some studies. All included studies were estimated to have a low overall risk of bias under various categories (Table 2). As studies comparing DBS and plasma presented DBS-plasma ratios, the TRL score was 8 for rifampin, ethambutol, and linezolid. Albeit the small sample size, measuring DBS was performed in mostly patients with TB, and the TRL scores for isoniazid, pyrazinamide, moxifloxacin, and clarithromycin were 7, indicating the technology of using DBS was demonstrated in an operational environment (Table 6). The TRL score for rifapentine was 6 as the study was performed in healthy volunteers rather than patients with TB [28].

Table 2 List of studies reporting DBS sampling, serum/plasma-DBS comparative methods, and risk of bias
Full size table

3.2 Urine

A total of 43 articles were found to determine rifampin, rifapentine, isoniazid, pyrazinamide, ethionamide, levofloxacin, and cycloserine in urine. Study populations comprised healthy volunteers and adult and pediatric patients with drug-susceptible drug-resistant TB or patients with latent TB infection. The sample size ranged between one and 650 participants. Dosage, sample collection, and drug analytical methods are listed in Table 3. Visual detection using the Arkansas method was the most common method of testing adherence among patients taking isoniazid. Seventeen studies quantitatively measured drugs in urine, and seven of the 17 studies compared urine concentration with serum. Only one study [36] described a procedure for reporting the absence of isoniazid in urine to the treating physician to monitor adherence.

Table 3 List of studies reporting urine sampling, serum/plasma-urine comparative methods, and risk of bias
Full size table

Studies were assessed for the risk of bias. All participants included in one study [37] were male, causing a moderate risk of bias in the selection of participants into the study. Of the four participants enrolled into one study [38], results were reported for three participants, leading to a moderate risk of bias due to missing data. Another study [39] reported only cumulative apparent excretion for a metabolite of isoniazid instead of the parent compound, causing a moderate risk of bias in selection of the reported results. High-performance liquid chromatography was used to measure serum concentrations for rifampin [11, 15], pyrazinamide [12], and levofloxacin [13] whereas, colorimetry with a spectrophotometer was used to measure urine concentrations, leading to a moderate risk of bias in the measurement of outcomes due to the different analytical instruments used. All other studies [14, 36, 40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73] had a low overall risk of bias (Table 3). Urine had a TRL score of 7 (Table 6) for all drugs except ethionamide found in this systematic search, as most studies were performed in patients with TB in different settings, but the absence of urine-serum/plasma ratios prevents urine from being used prospectively to perform TDM. Both studies testing the presence of ethionamide were performed in healthy volunteers, resulting in a TRL score of 6.

3.3 Saliva

Studies comparing saliva and serum were found for two first-line drugs, rifampin and isoniazid, and five second-line anti-TB medications, levofloxacin, moxifloxacin, linezolid, amikacin, and clarithromycin. Patients with TB and healthy volunteers comprised the study population and sample sizes ranged from 6 to 45 participants. Liquid chromatography-tandem mass spectrometry was the most common instrument for drug quantification, followed by spectrophotometry (Table 4). A novel mobile ultraviolet-visible spectrophotometry was repurposed to detect levofloxacin [18, 20] and linezolid [19] in saliva. The duration between treatment initiation and sample collection [74] and saliva sampling times [75] were not provided for two studies. The risk of bias was assessed, and one study [76] had a moderate risk of bias because of the selection of participants in the study as all participants were female. Remaining studies [18,19,20, 50, 74, 77,78,79,80,81] had a low overall risk of bias (Table 4). The TRL score for all saliva studies was 8 (ultraviolet-visible as they were performed mostly in patients with TB on drug regimens similar to those found in programmatic settings and most studies performed saliva-plasma/serum comparisons.

Table 4 List of studies reporting salivary sampling, serum/plasma-saliva comparative methods, and risk of bias
Full size table

3.4 Hair

A total of 13 articles reported on measured hair concentrations of three first-line TB drugs (isoniazid, pyrazinamide, and ethambutol) and eight second-line drugs (levofloxacin, moxifloxacin, linezolid, clofazimine, bedaquiline, pretomanid, ethionamide, and delamanid). Apart from parent compounds, three articles also measured TB drug metabolites in hair (acetyl-INH [21, 82] and DM-6705 [83], a metabolite of delamanid). Study populations comprised both adults and pediatric patients, and sample sizes ranged from two to 264 participants. Liquid chromatography-tandem mass spectrometry was used in all studies to quantify the various anti-TB drugs from hair (Table 5). Only two [22, 84] of the 13 studies performed comparative pharmacokinetic studies in plasma as well as hair samples, and simple scatter plots were used to demonstrate correlations. All studies were assessed for the risk of bias and two studies had a moderate risk of bias because of the selection of participants as one study [85] had 98% female participants and the other [22] enrolled all male participants. Other studies [21, 82,83,84, 86,87,88,89,90,91,92] had a low overall risk of bias. Similar to urine and saliva, the TRL score for hair was 7 (Table 5) as all studies were performed in patients with TB in operational settings.

Table 5 List of studies reporting hair sampling, serum/plasma-hair comparative methods, and risk of bias
Full size table

4 Discussion

This systematic review sought to explore opportunities for performing TDM for anti-TB drugs in alternative biological matrices to serum or plasma, specifically DBS, urine, saliva, and hair. We found that numerous classes of anti-TB drugs have been studied in quantitative or semi-quantitative assays in the alternative matrices, but few have been carried forward beyond diagnostic accuracy work to translate into dose adjustment. Studies within certain matrices such as DBS and saliva have been more comprehensive in reporting diagnostic accuracy, comparing levels to relevant pharmacokinetic parameters in serum or plasma, while studies in urine and hair have focused primarily on predicting medication adherence (Table 6).

Table 6 Data indicating TRL of each alternative matrix for drugs found in this systematic review
Full size table

Performance characteristics for each alternative biomatrix described in this systematic review are important to consider. For instance, from our search results for DBS, comparisons between plasma and DBS were performed for rifampin, pyrazinamide, ethambutol, moxifloxacin, and linezolid. The study by Martial et al. [29], conducted in children, had DBS to plasma ratios of 0.75 for rifampin, 0.81 for pyrazinamide, and 0.51 for ethambutol. While the ratios were acceptable for rifampin and pyrazinamide, ethambutol concentrations in DBS may be unsuitable to predict plasma concentrations because of low precision. The authors attribute the lower ratio of rifampin and pyrazinamide to peripheral distribution variability in children [29]. Linezolid showed good agreement between DBS and plasma with a ratio of 1.2 and a narrow range [33]. Linezolid concentrates more in erythrocytes than plasma and the differences in binding capacity cause linezolid concentrations to be higher in blood, hence, the authors proposed conversion factors to determine corresponding plasma values [33, 93]. High sample stability was also observed, making monitoring with DBS feasible for linezolid, which can reduce under-exposures or over-exposures in as many as 40% of patients [93]. These features may have broad applicability given the widespread roll out of linezolid in rifampin-resistant TB regimens for both improved efficacy and mitigating common exposure-related toxicities of linezolid [94]. For drugs such as rifapentine, isoniazid, moxifloxacin, and clarithromycin, studies with fewer than 30 participants were found, and the absence of reported DBS-plasma/serum ratios precluded prediction of clinical applicability. Although DBS can be a more convenient alternative to collecting whole blood for drug quantifications, especially in very young children and other participants unable to undergo multiple large-volume blood draws, there is a need for validated sample collection and measurement techniques [95] before blood spots can be used in lieu of plasma/serum for drug monitoring.

In contrast to the other biomatrices, urine has been utilized to monitor adherence to anti-TB treatment for over five decades. This earlier usage was borne from the misguided assumption that treatment failure arose from a patient’s inability or unwillingness to take medications as prescribed. Currently, variable adherence is understood as an expected response to TB treatment, but prescribed dose and individual pharmacokinetic variability also largely influence drug exposure and treatment outcome [96, 97]. Thus, there have been advances to use urine colorimetric methods for quantification within a medication dosing interval in an attempt to make a more precise dose adjustment in response to an individual’s pharmacokinetic variability. For example, the earlier visual detection of color change upon adding chemicals to the patented IsoScreen kit to detect isoniazid semi-qualitatively has been adapted to measure concentrations of various drugs [12, 13]. To reduce the use of laboratory demands further, a mobile phone color reader with a standardized light box has been used to quantify rifampin concentrations in urine [15]. However, only a few of the identified studies in this review quantitatively measured concentrations of drugs in urine, distinct from the semi-quantitative methods used for the measurement of adherence [11,12,13, 15, 50]. Although testing for adherence has been well validated for rifampin and isoniazid, a lack of reported urine-plasma/serum ratios in quantitative studies makes it difficult to identify urine threshold concentrations that may be predictive of optimum plasma exposure. Furthermore, while urine assays may be relatively simple to implement owing to an easier sample collection for all ages, including the presence of special urine collection bags for pediatric patients, and simple quantification methods, the identified studies did not consistently report on factors such as patient hydration, urine pH [98], and the presence of other co-morbid conditions affecting renal clearance.

Most studies of the saliva matrix reported concentrations in ratio to serum or plasma values allowing interpretation as to whether some drugs were more or less fitting for this platform. For example, rifampin, arguably the most important anti-TB drug, had the lowest ratio of 0.07 of saliva:plasma concentrations observed in one study [74], making the use of saliva to predict plasma concentrations challenging. Rifampin saliva concentrations were low despite assured adequate dosing [74, 77], likely due to strong binding of rifampin to plasma proteins and poor diffusion into the salivary glands [99]. A wide range of saliva-plasma ratios was reported for isoniazid, levofloxacin, and linezolid that could be due to varying dosing and sampling methods across studies. The highest ratio was observed for clarithromycin of 3.07 in Bolhuis et al. [80]. Higher ratios may allow for easy detection in saliva. This may be promising for other infectious diseases, as clarithromycin or other macro/azalides are more indicated for treating non-tuberculous mycobacteria. More important than the actual ratio is the inter-patient and intra-patient variability in the ratio as it would allow the incorporation of an appropriate correction factor where the ratio is reproducible. To illustrate, isoniazid is not bound to plasma proteins and can easily diffuse into saliva [100], yet the inter-study variability of saliva-plasma ratios among Anusiem et al. [50], Gurumurthy et al. [74], and van den Elsen et al. [77] suggests that salivary flow and pH might influence concentrations and well-designed pharmacokinetic studies would be needed before a reliable correction factor can be applied. However, saliva TDM appears possible in the treatment of rifampin-resistant/multi-drug-resistant TB for the key drugs of the fluoroquinolone class (levofloxacin and moxifloxacin) and linezolid. These drugs have also been measured using a novel, mobile, micro-volume, ultraviolet-visible spectrophotometer [18, 19], which can quantify salivary drug concentrations as demonstrated at the bedside in at least one study among patients with drug-resistant TB in Tanzania [20].

The systematic review did identify a relatively recent increase in the number of studies attempting to quantify drug exposure from hair samples in a range of cohorts with both drug-susceptible and drug-resistant TB. As a representative example, in a study by Mave et al. [21], hair samples were collected at 2, 4, and 6 months after isoniazid therapy initiation where isoniazid and acetyl-isoniazid concentrations were decreasing over time, which the authors suggested might indicate important changes in adherence patterns. Additionally, for a drug such as isoniazid that is unstable in plasma, DBS, and urine over long periods and requires cold-chain transport from serum or plasma, hair may offer an advantage for the measurement of cumulative drug exposure over time due to the relative stability of isoniazid in this biomatrix [21]. Overall, however, comparative studies of hair concentrations with gold standard plasma or serum concentrations were few as plasma and serum measurements cover different durations of exposure compared with hair. Concentrations in hair are an indicator of the average level of drug over a period of weeks or months, and contemporaneous plasma or serum measurement would only reflect a more recent drug intake, usually during a single dosing interval. In future studies, a different type of comparison between plasma or serum and hair could involve comparing a steady-state drug concentration in serum over a clinically relevant period (utilizing peak and trough concentrations) with hair concentrations in the same span of time.

This systematic review was not without limitations. A validated tool for assessment of the risk of bias of bioanalytical-pharmacokinetic types of studies was not available, but we instead modified the ROBINS-I for this purpose. Hence, a validated tool would be needed to properly assess the risk of bias in pharmacokinetic studies to avoid inappropriate risk classification. Some studies were performed in healthy volunteers or spiked samples, which could limit the extrapolation of findings to patients with TB, particularly those treated with multi-drug regimens.

Despite these limitations of early-stage studies, TDM using DBS, urine, saliva, or hair would be of immense benefit in TB-endemic regions and therefore randomized controlled trials enrolling diverse populations including adults, adolescents, and children with drug-susceptible drug-resistant TP from various ethnicities are needed. Dosage regimens in these studies must be most indicative of dosages administered in clinical and programmatic settings, and paired plasma/serum-alternative matrix sampling should be obtained for full pharmacokinetic curves and for additional population-pharmacokinetic studies that inform dose adjustment strategies. Population pharmacokinetic modeling and pharmacokinetic-pharmacodynamic studies could help predict the most appropriate individual dose, and model-informed precision dosing could also be utilized in predicting sampling schedules and exposures in alternative matrices [101]. Variable factors need to be taken into consideration to provide high-level evidence for TDM and these include volume and hematocrit effects for DBS [9]; pH, fraction of drug eliminated renally, hydration, renal function for urine [98]; salivary flow and pH [17]; and understanding relevant serum exposures from hair concentrations [23]. Having validated analytical methods for plasma and or serum and the alternative matrix, and the ability to calculate plasma-matrix ratios from AUC values form important components of a rigorous pharmacokinetic study design [17]. Last, with TDM more commonly performed among both inpatients and outpatients [7], there is also a need to determine the cost effectiveness and financial implications that TDM might pose to individuals and service providers in TB-endemic settings [102, 103].

5 Conclusions

Despite the readiness of alternative matrix assays to be performed in operational settings and considerable promise for the use of alternative matrices for personalized dose adjustment, assays from DBS, urine, saliva, and hair must be subjected to well-designed studies with diverse study populations on TB treatment, using consistent sample collection methods and validated analytical techniques for both serum or plasma and the alternative biomatrix to increase the uptake in guidelines and accelerate implementation in programmatic TB treatment.