Introduction

Theoretically, therapeutic drug monitoring (TDM) concerns only a small number of classes of drugs that represent a few dozen compounds in total. Improvement to patient care based on blood concentration measurements of these drugs has been clearly established and consequently the analytical methods used for quantification have to provide accuracy, specificity, selectivity and productivity. Nowadays, immunoassays are used worldwide, mainly for the TDM of certain immunosuppressants, antibiotics or anticonvulsants. However, these technologies are unspecific for certain drugs and are subject to interferences. Moreover, the costs of reagents and consumables are substantial. Apart from immunoassays, TDM can be performed using liquid chromatographic techniques coupled to specific detectors such as UV-diode array detectors (LC-DAD) which are able to cover a very large panel of relevant compounds. Nevertheless, the limited specificity of UV spectra, their variability as a function of pH and the fact that a lot of compounds present poor UV absorbance, render LC-DAD that are neither very specific, reliable nor universal. Initially, the LC-MS techniques were mainly used in research, due to the high financial investments involved and the technical difficulties encountered in setting up procedures for routine activity. Since then, these technologies have been made available for an increased number of users as the cost has been significantly reduced and the application has been simplified. Obviously, the question for the head of a laboratory is no longer whether he can afford to invest in an LC-MS system or not, but to decide which methods should be transferred from classical immunoassays or other chromatographic techniques to LC-MS, and which new quantification procedures should be developed using this technology. As an illustration of this tendency, numerous LC-MS procedures dedicated to TDM have been published in the last few years.

In this paper, the authors focus on the most recent studies reported for the improvement of TDM of immunosuppressants, antifungal drugs, antiretroviral drugs, antidepressants and antipsychotics, and discuss the current role of LC-MS in the management of these drugs.

Immunosuppressants

The TDM of immunosuppressants currently concerns cyclosporine A (CsA), everolimus, sirolimus, tacrolimus and mycophenolic acid (MPA). The therapeutic ranges of CsA and above all MPA are much higher than those of everolimus, sirolimus and tacrolimus. Three principal analytical technologies are widely used in their TDM: immunoassays, liquid chromatography coupled with diode-array detection (LC-DAD), and liquid chromatography coupled with single or tandem mass spectrometry (LC-MS/(MS)) and recent reviews underline the increasing importance of LC-MS [1–3]. Indeed, numerous studies comparing performances between immunoassays and other chromatographic techniques showed that results obtained using immunoassays are overestimated due to cross-reactivity of antibodies with metabolites.

We will only consider here papers published since these lastest reviews, including methods describing the quantification of metabolites (see Table 1). Brandhorst et al. recently reported a simple LC-MS/MS method for the determination of MPA and its glucuronide metabolites after protein precipitation of 100 μL of serum [4]. Quantification of MPA in saliva [5] or plasma [6, 7] after simple protein precipitation with a similar sample volume has also been described. These techniques involve chromatographic runs which do not exceed 8 min. Previous techniques [9–13] investigated in a former review [2] generally required longer chromatographic runs and sample preparation procedures. However, as signalled in a short paper [14], though analytical run time can be shortened, appropriate chromatographic separation between MPA and its glucuronide derivative (MPAG) is essential, due to in-source fragmentation of MPAG into the MPA aglycone. LC-DAD is widely used in the clinical laboratory for TDM of MPA [15–20]. It involves longer extraction procedures (generally liquid–liquid extraction (LLE)) [17, 18] and chromatographic runs. Solid-phase extraction (SPE) tends to be used more often now [15, 19]. Interestingly, MPA was investigated in a recent paper presenting an on-line sample preparation procedure and UV detection at 306 nm [15]. Furthermore, among the most recent papers, an easier sample preparation procedure, i.e. protein precipitation, seems to be applicable upfront to this mode of detection [16].

Table 1 Overview of techniques using LC-MS(/MS) for therapeutic drug monitoring of immunosuppressants in humans
Full size table

Few methods [21–26] concerning the determination of CsA (sometimes with its metabolites) have been published since the last reviews [1–3]. Koseki et al. [21] developed a method using 100 μL of blood sample that allows the simultaneous LC-MS/MS determination of CsA and its main metabolites in 10 min. These authors used a triple quadrupole with an APCI source and selected reaction monitoring (SRM) in the positive mode and reported a 1–2,500 μg/L analytical range for each analyte. Vollenbroeker et al. [24] presented an LC-MS method for the determination of CsA and some of its metabolites in blood sample using single ion monitoring (SIM) but with a very long chromatographic run (65 min). Salm et al. [22] published a method for CsA with a chromatographic run of 2 min, linear from 10 to 2,000 μg/L using a simple protein precipitation of 50 μL of blood sample followed by SPE on a C18 cartridge, SRM detection being performed on a triple quadrupole instrument equipped with an electrospray interface working in the positive mode. Keevil et al. [26] had previously published a method with a 1.5-min chromatographic run time obtained by using a triple quadrupole mass spectrometer fitted with a Z Spray ion source.

Although CsA has no chromophores it can also be measured by HPLC with ultraviolet detection at relatively low wavelengths (i.e. <230 nm) [1]. Of note, this technology was considered as the gold standard throughout the world for CsA measurement a decade ago [27]. However, longer extraction procedures (e.g. liquid–liquid extraction) are necessary to remove interferences from matrix, and the lack of specificity of the technique when using low wavelength is obvious.

Due to its low therapeutic concentrations and weak UV absorbency, UV detection seems to be inadequate for tacrolimus. To the best of our knowledge, no new method allowing the specific quantification of tacrolimus has been published since the last reviews [1–3], except a recent paper presenting a quantitative method for three isomeric metabolites of tacrolimus with a run time of 10 min [31]. Older techniques were also characterized by short chromatographic runs, limits of quantification which did not exceed 0.2 μg/L [28, 29] and sometimes included metabolites [30].

The recent LC-MS(/MS) techniques for the quantitation of sirolimus or everolimus in blood were characterized by the small sample volumes needed, fast sample preparation and short chromatographic runs [32, 33]. Using only 100 μL of blood, Taylor et al. [32] described an SRM method with a 2-min run time in positive mode using a triple quadrupole equipped with an electrospray ionization source that allows the quantitation of everolimus from 0.5 to 40 μg/L. Moreover, Korecka et al. [33] recently published an LC-MS/MS method using a triple quadrupole mass spectrometer working in the positive mode for measurement of everolimus in 2.8 min with a limit of quantification of 1 μg/L. This procedure was based on on-line extraction and cleanup of the injected sample with subsequent introduction into the mass spectrometer by using a built-in Valco switching valve.

Methods using UV detection have recently been published for sirolimus and everolimus [34, 35]. Despite good analytical results, they often rely on time-consuming LLE procedures and much longer analytical runs than above-mentioned LC-MS/MS procedures.

As patients’ treatment often consists in a combination of different immunosuppressants, there is great interest in developing methods which include several drugs in a single procedure. Some studies reporting the simultaneous determination of combined immunosuppressants have been published [36–44]. Briefly, the following strategies were employed: (i) direct injection of the supernatant obtained after centrifugation, (ii) injection of a reconstituted sample obtained after drying the supernatant, (iii) on-line sample preparation such as SPE or turbulent flow chromatography. Contrary to the other modes of detection, in these studies the chromatographic separation of all compounds of interest was not as necessary because the authors employed the highly specific tandem mass spectrometry. However, some pitfalls inherent to the use of mass spectrometry, such as matrix effects and/or ion suppression [45], had to be thoroughly investigated when such high-throughput sample preparation was involved.

In conclusion, immunoassays are still widely used for the analysis of immunosuppressants. However, recent studies showed their important drawbacks due to their cross-reactivity with metabolites. Although LC-DAD technologies are less expensive, extraction procedures used are often time-consuming and analytical runs are relatively long. As illustrated by the recent tendency to transfer immunosuppressant measurements from immunoassays to LC and owing to the possibility of simultaneous analysis of several immunosuppressants, LC-MS/MS may become the mainstay technology for this TDM in the near future. The main advantages of the methods are minimum sample preparation, short analytical runs, lower limits of detection and high selectivity (theoretically, absence of interfering compounds). Moreover, although the investment cost is important, profitability can be rapidly obtained because of the minor cost of reagents.

Antifungal drugs

Numerous antifungal drugs have been commercialized in the last few years, with new azole derivatives (posaconazole, voriconazole), new classes of compounds (echinocandins) or new galenic formulations (lipidic forms of amphotericin B). Among antifungal drugs, only those used for the treatment of invasive and disseminated fungal infections are eligible for TDM, namely flucytosine, amphotericin B, caspofungin and azole derivatives (itraconazole, fluconazole, voriconazole and posaconazole). The pharmacokinetics and pharmacodynamics of these compounds have been recently reviewed [46]. Though pharmacokinetic–pharmacodynamic relationships were clearly identified in experimental models, few target plasma concentrations could be determined from studies performed in patients [47]. Thus, the relation between plasma concentrations and efficacy is still ambiguous for most antifungals. Nevertheless, multiple analytical techniques have been published in the last few years for the determination of antifungals in human fluids (see Table 2), and preliminary pharmacokinetic studies suggest that TDM can be useful in most populations, especially patients suffering from major burns or cystic fibrosis [46].

Table 2 Overview of techniques using LC-MS(/MS) for therapeutic drug monitoring of antifungal drugs in humans
Full size table

For azole derivatives, only a few LC-MS(/MS) methods have been published up to now, and most of them concerned itraconazole. Numerous methods have been reported to quantify itraconazole and its coactive metabolite hydroxyitraconazole (OH-ITZ) in biological fluids using HPLC coupled to different detection modes. Among them, a method based on SPE and fluorescence detection with a sensitivity of 10 μg/L [48], one using LLE and spectrofluorimetric detection with a sensitivity of 20 μg/L [49], and one based on a sample preparation using protein precipitation and UV detection with a sensitivity of 25 μg/L [50] are available. However, these methods largely used for TDM often lack the selectivity needed for these analyses. Only a few methods for the determination of itraconazole with higher sensitivity using LC-MS(/MS) have been published. Yao M et al. proposed a method with a single mass spectrometer where the detection of ITZ was performed in SIM mode after protein precipitation of 100 μL of plasma [51]. The authors reported that standard curves were linear over the concentration range 4–1,000 μg/L with a recovery of 96%. However, the validated assay was only applied to a pharmacokinetic study of ITZ in rats following administration of a single dose of itraconazole (15 mg/kg). Recently, Young et al. reported a method using a triple quadrupole mass spectrometer operated in the positive mode and using MRM to quantify ITZ [52]. The authors reported good precision and accuracy but the upper limit of the concentration range was 100 μg/L. Vogeser et al. [53] reported an LC-MS/MS method that involves automated SPE for the quantification of ITZ and its active metabolite in human samples, whereas Kousoulos et al. reported a semi-automated procedure based on a LLE [54]. In this last application the extraction process was performed using two liquid-handling robotic workstations, the detection with a triple quadrupole operated in MRM mode, and ranges of concentrations 2–500 μg/L for ITZ and 4–1,000 μg/L for OH-ITZ were obtained. Apart from the gain in sensitivity and selectivity, the main interest of these LC-MS/MS procedures was to offer run times of less than 5 min when coupled to robotic workstations for the extraction procedure.

Numerous clinical reports have been published about voriconazole pharmacokinetics. In these studies various analytical HPLC methods were employed for the determination of voriconazole in serum [55–58]. The first LC-MS method was described by Zhou et al. [59] to study the penetration of voriconazole in the eye of rabbits, and so no sample preparation was required because of the clean aqueous humor samples that were employed. Keevil et al. [60] developed an LC-MS/MS method involving precipitation of proteins with zinc sulfate and acetonitrile before on-line chromatographic cleanup and analysis with a triple quadrupole. However, as reported by the authors themselves, this technique was “found to lack interassay reproducibility with the results of a QC sample (HPLC value of 3.026 mg/L) varying from 2.42 to >6.2 mg/L”. Recently, Egle et al. [61] reported a “fully automated analysis of voriconazole from serum by LC-MS/MS with parallel column-switching technique”. Working with a triple quadrupole coupled with an orthogonal electrospray ionisation source and an ion trap, the authors described the quantification of voriconazole in crude serum without any other form of pre-treatment. The procedure was based on automated on-line extraction with a total sample preparation and analysis time of 13 min, showing sufficient performance and precision. Voriconazole was detected by isolation and fragmentation of the positive pseudomolecular ion (m/z = 350.3) and its specific fragment (m/z = 281.3) which results from loss of the triazole moiety. A major point in this method was the very small volume of serum (5 μL) required for the analysis, whereas classical HPLC methods for the detection of voriconazole require at least 500 μL. This could be of interest in some special cases.

Although the routine monitoring of posaconazole plasma levels has not been recommended, an accurate assay could be necessary, at least in the context of clinical trials. Based on earlier toxicokinetic data, an analytical method able to measure concentrations as low as 5 μg/L is needed to accurately characterize the human plasma concentration versus time profile of posaconazole. In a recent study, Shen et al. [62] presented a 96-well sample preparation and LC-MS/MS method that was suitable for quantifying posaconazole over a dynamic range of 5–5,000 μg/L. The positive-ion APCI mass spectrum of posaconazole was dominated by the protonated molecular ion at m/z 701, and only showed a m/z 683 fragment (dehydration). The loss of water is not a recommended choice for a selective MRM transition, but according to the authors the difference in signal to noise ratio between the two transitions was sufficient to allow its selection for quantitation of posaconazole.

Very few experimental or clinical data describing the pharmacokinetics of caspofungin are available [63, 64]. However, a specific and reliable method for the measurement of this drug could be necessary. Such a method should be suitable for the determination of clinically relevant caspofungin levels (i.e. 1.2–10.4 mg/L in serum). Egle et al. [65] published a method using an advanced column-switching technique for fully automated analysis of caspofungin in serum without any pre-treatment. The authors reported an extraction performed automatically on-line, using a diol column, and followed by chromatography on a CN column. MS detection was performed with a triple quadrupole. To the best of our knowledge, no other LC-MS technique has been proposed up to now for the determination of caspofungin. However, the authors did not use any internal standard in their method. In our opinion, this strategy is not pertinent. An internal standard is supposed to compensate for the variability in sample preparation, recovery, injection and possible compound degradation. Although it is more the case for GC-MS than for LC-MS, an internal standard is also indispensable to limit detection variability leading to quantification errors. For example, ionization efficiency (i.e. the portion of the total number of molecules ionized when the compound is introduced into the source) varies during day-to-day operation due to variations in several parameters that are impossible to control (the temperature and pressure in the ion source, for instance). Additionally, the signal is further modified by the effect of matrix components present in serum, plasma or urine. Indeed, when the analyte is introduced into the ion source, it competes for ionization with other compounds simultaneously introduced into the source, which decreases the analyte signal (ion suppression) especially with electrospray ionization. The degree of ion suppression caused by matrix components mainly depends on the chemical structure of the analyte. Therefore, analyte and internal standard have to present similar structures to preserve the ratio of analyte and internal standard detector responses. For this reason, internal standards used in quantitative LC-MS(/MS) assays should either be structural analogues or labelled (e.g. deuterated) analogues of the analyte.

Early determination of amphotericin B was performed using microbiological inhibition assays, which lacked selectivity and sensitivity. HPLC methods were then developed for the quantification of amphotericin B in plasma or serum [66–73]. The most sensitive HPLC method reported for amphotericin B in plasma used solid-phase extraction for sample cleanup with a limit of quantitation of 5 μg/L [73]. All these methods concerned the total (free and bound) concentrations of amphotericin B in blood. Lee et al. reported an LC-MS/MS procedure with solid-phase extraction for the determination of total amphotericin B in human plasma and other biologic matrices as well as the free amphotericin B concentration in plasma [74]. Using a triple quadrupole, electrospray interfaced to a C18 analytic column, the authors reported a LOQ of 1 μg/L for ultrafiltrate and LOQs of 2 mg/L in plasma for total amphotericin B, 50 μg/L in urine, and 400 μg/L in fecal homogenate.

The last few years have seen the availability of several new antifungal drugs, providing increased efficacy and safety, particularly in immunosuppressed patients. The choice of the drugs, their combinations and their dose adjustments are still a matter of debate. In this context, the availability of specific and sensitive analytical techniques is probably necessary.

Antiretroviral drugs

The role of TDM for the optimization of antiretroviral therapy has been the focus of previous reviews [75–82]. There is consensus that most antiretroviral agents can be considered as candidates for TDM, as they are characterized by a large inter-individual variability in drug concentrations. Clinical data also suggest a correlation between drug concentrations and toxicity or efficacy. Indeed, plasma determination of antiretroviral drugs [82] is important in relation to: (i) checking compliance, (ii) drug–drug interactions, (iii) physiopathologic status changes, (iv) pregnancy, (v) virological failure, (vi) pediatric patients. Optimal efficacy and prevention of viral resistances require that antiretroviral drugs are administered in combination regimens, which are generally referred to as highly active antiretroviral therapy (HAART). According to current guidelines, HAART generally consists of two nucleoside reverse transcriptase inhibitors (NRTIs) combined with either one or two protease inhibitors (PIs), or with an non-nucleoside reverse transcriptase inhibitors (NNRTI). Additionally, ritonavir is commonly added as a pharmacokinetic enhancer.

Most of the published procedures for the TDM of antiretroviral drugs were developed using LC-DAD [83]. Briefly, large differences amongst assays in lower limits of quantitation for the measurement of PIs have been reported. The review article by Aarnoutse et al. [84] shows that a number of assays are not sensitive enough to measure concentrations below mean trough levels in the population or below the recommended therapeutic ranges. The selectivity of LC-DAD for antiretroviral drugs is another major point in method validation. Apart from endogenous substances, the potential for drug interferences is high, because of the large number of concomitant drugs administered to HIV-infected patients, as well as the presence of numerous metabolites, particularly for PIs. Most of all, in order to increase sensitivity and reduce plasma volume, LC-DAD methods were usually developed using low wavelength ranges, which are not specific enough.

Numerous methods dedicated to the determination of a single compound have been reported [85–88], but, except in particular cases such as pharmacokinetic studies, it seems obvious to develop methods allowing the detection of several antiretroviral drugs in a single procedure. This approach is clearly recommended for a routine activity. In the review of the HPLC methods articles published since 2000 [84, 89, 90], we identified about 25 studies reporting the use of an HPLC technique for the quantification of several NRTIs, NNRTIs or PIs in a single run. Among them, nine were LC-MS or LC-MS/MS techniques (see Table 3) [91–99]. For example, working on a triple quadrupole mass spectrometer, Crommentuyn et al. developed a procedure for the quantification of six PIs (amprenavir, indinavir, lopinavir, nelfinavir, ritonavir and saquinavir) and the pharmacologically active nelfinavir metabolite M8 in plasma [96]. Simple protein precipitation of 100 μL of plasma afforded a linearity over a concentration range 10–10,000 μg/L for indinavir and saquinavir, 100–10,000 μg/L for amprenavir, 50–10,000 μg/L for nelfinavir and ritonavir, 100–20,000 μg/L for lopinavir and 10–5,000 μg/L for M8. The analytical run time was about 5 min. Egge-Jacobsen et al. [91] reported a method for the quantification of six PIs (amprenavir, indinavir, lopinavir, nelfinavir, ritonavir and saquinavir) and three NNRTIs (nevirapine, efavirenz and delavirdine) using a G1946A single quadrupole mass detector. This procedure was based on an on-line extraction with a previous simple precipitation of 100 μL of plasma and presented LOQs ranging from 20 to 100 μg/L for the investigated compounds. The total procedure lasted about 8 min. Volosov et al. [99] developed a method for quantification of most of the marketed anti-HIV drugs in human plasma, using a triple quadrupole. This procedure included an on-line extraction–cleaning step using 80 μL of plasma. Stavudine, zidovudine and efavirenz were analysed in the negative MS/MS mode, while amprenavir, indinavir, lopinavir, nelfinavir, ritonavir, saquinavir, abacavir, didanosine, lamivudine, zalcitabine, delavirdine and nevirapine were analyzed in the positive mode. The procedure allowed the determination of this panel of drugs with a 4.5-min run time. Between-day imprecision was less than 10% for all analytes, accuracy ranged between 95 and 105% (n = 20), and the method was linear over the measuring range of each drug (2–2,000 μg/L for stavudine, didanosine, zalcitabine and zidovudine, and 10–10,000 μg/L for all other drugs). Therefore, these examples of LC-MS/MS methods illustrate the feasibility of this technique for the TDM of anti-HIV drugs. LC-MS/MS obviously provides satisfactory results especially when compared with those obtained in recently published LC-DAD methods. For example, a procedure for the determination of 16 anti-HIV drugs using LC-DAD required 600 μL of plasma, for off-line SPE and took 35 min for a single analysis [100], with calibration ranges of, among others, 25–10,000 μg/L for didanosine and lamivudine, and 100–10,000 μg/L for most PIs. In our opinion, such a comparison illustrates the opportunity of a substantial gain in sample volume for the analysis, and in analysis time through the automation of the extraction procedure and simplification of chromatographic separation when using LC-MS/MS.

Table 3 Overview of techniques using LC-MS(/MS) for therapeutic drug monitoring of antiretroviral drugs in humans
Full size table

Total plasma concentrations are currently measured for therapeutic drug monitoring of PIs and NNRTIs. However, the pharmacological target of antiretroviral drugs is inside cells. This suggests that a potential improvement of TDM would be to measure cellular concentrations [101–106]. In this context, the difficulties encountered with sample cleaning and the low concentration ranges targeted required the use of a very sensitive and specific technique. Recently, Colombo et al. [101] studied the variability of cellular accumulation of several PIs and NNRTIs, and evaluated to what extent total plasma concentrations correlate with cellular concentrations in 133 HIV-infected patients. In this study, total plasma concentrations were analysed by LC-DAD but cellular concentrations (peripheral blood mononuclear cells) could only be determined by LC-MS/MS using a triple quadrupole to achieve the analytical sensitivity required for the intracellular measurements (i.e. LOQ <0.4 μg/L). Of note, the best correlations between total plasma concentrations and cellular concentrations were observed for nelfinavir, saquinavir and lopinavir, while no correlation was found for nevirapine, which suggests that monitoring of cellular drug concentrations can be of interest for this compound.

Antidepressants and antipsychotics

Antidepressants are drugs widely prescribed for the treatment of a variety of depressive states and psychiatric disorders. They can be divided in five main classes: tricyclic antidepressants, also called imipraminic antidepressants; monoamine oxidase inhibitors (IMAO); selective serotonin reuptake inhibitors (SSRIs); serotonin-noradrenergic reuptake inhibitors (SNaRIs); and other antidepressants. The wide variety of these compounds increases the difficulty of a reliable TDM.

The analysis of tricyclic antidepressants (especially amitriptyline, nortriptyline, imipramine and desipramine) can be performed using immunoassays [107, 108]. However, the antibodies employed cross-react differently with the different drugs and also with their metabolites, rendering precise quantitation almost impossible. Moreover, there are risks of interferences due to drugs with structural similarity to antidepressants [107, 109].

The TDM of antidepressants is generally performed using chromatographic technologies. Most of the published techniques concerned a single drug (bupropion, citalopram [110–112], doxepin [113], fluoxetine [114–119], fluvoxamine, mianserin [120], milnacipran, mirtazapine [121], paroxetine [122, 123], sertraline [124, 125], venlafaxine [126]), sometimes with its metabolites, and employed GC, LC-DAD or LC-MS(/MS). Sample preparations consisted in most cases in LLEs or SPEs. A few multi-compound techniques [126–128] have also been developed using these same detection modes. Owing to the development of high-throughput technologies such as on-line SPE with column switching or turbulent flow chromatography, new opportunities have appeared. These recent papers present methods allowing the determination of several compounds in a single run and require a small sample volume, contrary to SPE or LLE for which not less than 500 μL is classically necessary. The main published techniques based on LC-MS(/MS) are summarized in Table 4.

Table 4 Overview of techniques using LC-MS(/MS) for therapeutic drug monitoring of antidepressants in humans
Full size table

According to this table, acquisitions are usually performed in the positive mode, mainly using electrospray ionization sources, either in SIM or MRM modes with run times varying from 3 to 10 min, the shortest run times being characteristic of MRM acquisitions. When LLE and SPE procedures are employed, sample volumes higher than 500 μL are often necessary to reach the therapeutic ranges of the investigated drugs. Newer methods allow the quantitation of several molecules at the same time and, owing to the emergence of automated systems allowing on-line sample extraction, a lower sample volume is needed.

As an example, Kollroser et al. [127] and Souverain et al. [117] developed methods for the determination of antidepressants using on-line sample extraction with column switching coupled with mass spectrometry, requiring only 50 μL of plasma. More recently, Sauvage et al. [128] developed a quantitative method for 14 antidepressants and their respective metabolites using turbulent flow chromatography coupled with a triple quadrupole (Fig. 1). This procedure used only 100 μL serum and presented calibration curves ranging from 10 to 500 μg/L. Working with a triple quadrupole, Kirchherr et al. also published a method for the quantitation of 48 antidepressants and antipsychotics based on a sample of 50 μL of serum, using a monolithic column with injection intervals of 8 min [129]. These last two applications yielded good performances in terms of limit of quantification, accuracy and inter- or intra-assay precision, at least comparable to those of methods using higher sample volume, longer extraction steps and chromatographic runs.

Fig. 1
figure 1

Example of the application of LC-MS/MS procedure proposed by Sauvage et al. [128] for the quantification of antidepressants in human serum. The chromatogram was obtained from a patient receiving clomipramine. The calculated concentrations were 202.5 and 85.2 μg/L for clomipramine and its demethylated metabolite, respectively

Full size image

The antipsychotics for which TDM is performed belong to a large variety of chemical families (phenothiazines, butyrophenones, benzamides, thioxanthenes, diphenylbutylpiperazines etc.) and present various modes of action and therapeutic ranges [131–133]. Most of the published methods concerning this therapeutic class concerned a single compound (amisulpride, aripiprazole, clozapine, droperidol, haloperidol, olanzapine, perospirone, perphenazine, pimozide, quetiapine, risperidone and its metabolite, sertindole, sulpiride, sultopride, tiapride, ziprasidone, zuclopenthixol). A lesser number dealt with several compounds. In most cases, LC was used, with different detection modes (UV, single or tandem mass spectrometry, electrochemical or fluorimetric detection).

Less methods using LC-MS(/MS) for antipsychotics have been published compared to antidepressants and these are summarized in Table 5. They were meant for the TDM of a single compound, sometimes with its active metabolite (amisulpride [134], aripiprazole [135, 136], clozapine [137, 138], haloperidol [139, 140], olanzapine [141–146], risperidone [147–150], sulpiride [151], sultopride [152], tiapride [153], ziprasidone [154]), or for several compounds [129, 155–157]. Generally, sample extraction procedures were not considerably simplified when compared to those applied with other techniques. However, as described above, Kirchherr et al. [129] proposed a method allowing the determination of several antipsychotics after protein precipitation of 50 μL of serum. The authors reported good accuracy, linearity and inter-and intra-assay precision but each compound was monitored using a single MRM transition. Kollroser et al. [156] also reported a method for three antipsychotics using on-line sample preparation based on Oasis HLB extraction cartridges, and a six-port switching valve coupled with an ion-trap mass spectrometer. The authors reported that the whole procedure lasted 6 min and was accurate enough for a TDM application. Kratzsch et al. [155] described a method for 15 neuroleptic drugs and three of their relevant metabolites with a limit of detection lower than 1 μg/L, allowing the determination of those active at low concentrations. After SPE of 500 μL of plasma, the presence of neuroleptics was screened for in the full-scan mode using a single quadrupole with an APCI source. Positive peaks in the traces were identified by comparing the spectra obtained with those registered in an in-house library. Finally, quantification was performed in the SIM mode.

Table 5 Overview of techniques using LC-MS(/MS) for therapeutic drug monitoring of antipsychotics in humans
Full size table

When analysing results of the Heathcontrol-Therapeutic Drugs Scheme (Cardiff Bioanalytical Services Ltd), which circulates quality control samples containing several antidepressants and antipsychotics, LC-MS and LC-MS/MS do not seem to be widely used yet. Indeed, for the majority of compounds investigated, less than 20% of all laboratories used these technologies. However, future developments of quantitative methods for antidepressants and antipsychotics TDM should obviously be focussed on technologies characterized by low quantification limits, necessitating less sample volume, easier and less time-consuming sample preparation. Therefore, the number of methods using automated extraction techniques, either on-line solid-phase extraction or turbulent flow chromatography, constantly increases. Nevertheless, developers should be informed of the risks of misusing such on-line sample extraction systems. Indeed, due to carry-over phenomena, the signal observed for a sample could be affected by carry-over of the drugs, fixed either on the extraction column or on the different switching valves from a former injected sample. Furthermore, the shorter analysis run time the higher the risks of ion suppression or cross-talking, i.e. when a signal caused by another compound with another MRM transition interferes with the investigated compound. For example, a recent paper described an LC-MS method allowing the simultaneous determination of fluoxetine, citalopram, paroxetine and venlafaxine in plasma [130]. Indeed, amitripyline, maprotiline and venlafaxine are characterized by the same m/z 278 pseudo-molecular ion, but apparently no investigation considering this obvious interference was carried out. Consequently, appropriate chromatographic separation of these potentially interfering compounds should be effective and afford accurate quantitation, though this point was not discussed in the article.

Moreover, as reported in a paper describing the interference of tramadol with a metabolite of venlafaxine [158], tandem mass spectrometry does not avoid such problems, and even with this very selective technique, the choice of appropriate transitions is of the utmost importance. Monitoring only one transition is now advised against, as it is often not sufficient if metabolites had not been considered beforehand. Indeed, for most antidepressants being characterized by a similar tricyclic structure and by an amine-containing side chain, some fragments may not be specific to a given drug. For example, the imipramine transition 281.3/86.1 can interfere with that of a demethylated metabolite of trimipramine, which can have similar retention properties [129]. The same risks apply to nortriptyline and a demethylated metabolite of maprotiline, as well as to phenothiazines that present similar chemical structures and are extensively metabolized, and so induce interferences in the case of co-administrated drugs (personal communication, unpublished data). Therefore, appropriate chromatographic conditions must be chosen to confirm that the signals studied are specific to the investigated compounds. From our own experience, although the selection of a single MRM transition improves the LOQ—as it increases the dwell time, or the number of points acquired over a chromatographic peak—it can be the source of errors even for a TDM method where the drugs to be analysed are known beforehand. In two recent papers [159, 160], although an extensive validation procedure was conducted, problems of isobaric compounds had not been investigated or documented, and potential risks of mass interferences with amitriptyline or maprotiline were not discussed. As a consequence it seems obligatory to use at least two transitions per compound and to check the respective abundances of the confirmation and quantitation transitions, especially when analytical run times become very short. Even if putatively interfering compounds had been extensively studied during method development, metabolites from concomitant administered drugs might interfere.

Conclusion

This paper illustrates the increasing role of LC-MS in the clinical laboratory routine activity. Nowadays, it seems obvious that almost all the drugs subject to TDM can be analysed using LC-MS(/MS). Therefore, it is no longer a question of evaluating the performance or the feasibility of this technology in routine activity, but to determine which classes of compounds have to be analysed with priority using it. The best example of current applications is the tendency to transfer the routine activity of immunosuppressants from immunoassays to LC-MS which allows the automation of sample preparation, shortens analytical run times and is probably cost-effective despite heavy investment costs. This single example illustrates most of the advantages of LC-MS techniques and suggests their suitability for routine TDM.

However, this expansion, not to speak of vulgarisation, implicates a necessary reassessment of the function of the TDM specialist (PharmD, clinical chemists, etc.) and the development of new positions such as analytical engineers in the laboratory. Indeed, whereas analysis automation using immunoassays permitted one to obtain results only by “pushing a button”, these new methods necessitate a higher degree of analytical competence. Knowledge and skills should be upgraded and criteria for assessing an applicable and accurate method have to be clearly defined and respected. When constraints and limitations (e.g. the use of at least two specific transitions per compound, the research of potential interferences and the ion suppression phenomenon) are ignored or incompletely understood, quantification or even identification errors can occur, and the advantages of LC-MS are lost.