The therapeutic monitoring of antimicrobial agents

Abstract

Aims

To review the basis and optimal use of therapeutic drug monitoring of antimicrobial agents.

Methods

Antimicrobial agents for which a reasonable case exists for therapeutic drug monitoring were reviewed under the following headings: pharmacokinetics, why monitor, therapeutic range, individualisation of therapy, sampling times, methods of analysis, interpretative problems and cost-effectiveness of monitoring.

Results

There is a strong historical case for monitoring aminoglycosides. The recent move to once-daily dosing means that criteria for therapeutic drug monitoring need to be redefined. Vancomycin has been monitored routinely but many questions remain about the most appropriate approach to this. A case can be made for monitoring teicoplanin, flucytosine and itraconazole in certain circumstances.

Conclusions

The approach to monitoring aminoglycosides needs to be redefined in the light of once-daily dosing. It is premature to suggest that less stringent monitoring is necessary as toxicity remains a problem with these drugs. The ideal method of monitoring vancomycin remains to be defined although a reasonable case exists for measuring trough concentrations, mainly to ensure efficacy. Teicoplanin is monitored occasionally to ensure efficacy while flucytosine is monitored occasionally to avoid high concentrations associated with toxicity. Itraconazole has various pharmacokinetic problems and monitoring has been suggested to ensure that adequate concentrations are achieved.

Introduction

Most antibiotics, such as the β-lactams, macrolides and quinolones have a wide therapeutic index and therefore do not require therapeutic drug monitoring. Some, such as the aminoglycosides and vancomycin, have a narrow therapeutic index, and toxicity may be severe and irreversible. Therapeutic drug monitoring may be appropriate for these drugs. Other drugs such as teicoplanin, flucloxacillin and the antifungal agents itraconazole, flucytosine and fluconazole are monitored in certain circumstances.

The aminoglycosides

There is a strong case for monitoring aminoglycoside concentrations [1]. The recent move to once-daily dosing has resulted in the need to re-evaluate monitoring strategies.

Pharmacokinetics

The pharmacokinetics of the aminoglycosides are relatively simple. These drugs are hydrophilic with low protein binding, and are eliminated renally. They have no stereoisomerism and are not subject to genetic polymorphism. The volume of distribution (V) approximates that of the extracellular fluid volume, and the clearance (CL) that of the glomerular filtration rate [2]. The λ_z-phase t_1/2 is approximately 2.5 h in patients with normal renal function, although there is a slow terminal elimination phase of 100–150 h, related to distribution into a ‘deep compartment’.

Aminoglycoside pharmacokinetics vary markedly according to the state of the disease that is being treated. Infections are associated with altered hydration and permeability of biological barriers. The V and CL may therefore change dramatically during therapy. Severe infection, or burns, can be associated with an increased V, which may require higher doses to achieve desired peak concentrations [3, 4]. Pre-renal failure may follow dehydration resulting in impaired aminoglycoside clearance, which may subsequently improve during fluid replacement and resolution of the infection.

Why monitor aminoglycosides?

The aminoglycosides have a low therapeutic index. Bactericidal efficacy is directly related to peak concentrations (concentration-dependent killing) while toxicity is related to total drug exposure [1]. Nephrotoxicity (usually reversible) and ototoxicity (often irreversible) are the major forms of toxicity [5].

The desired plasma concentration-time profile for aminoglycosides differs from that of most other drugs involved in therapeutic drug monitoring. For most drugs the aim is to have minimal fluctuations (i.e. a flat profile) within the dosing interval to keep the concentrations within a ‘therapeutic range’. A single plasma concentration measurement, representative of accumulation, is usually all that is required for monitoring.

For the aminoglycosides, the desired concentration-time profile includes a high peak concentration (for efficacy) followed by a low trough concentration (to prevent accumulation). The dose interval is substantially longer than the half-life. For therapeutic drug monitoring of the aminoglycosides, it has been traditional to monitor both a peak and a trough concentration. With the recent move to once-daily dosing, this approach needs to be reconsidered.

Therapeutic ranges

For many years, peak concentrations of 6–10 mg l⁻¹ and trough concentrations of ≤2 mg l⁻¹ were advocated for gentamicin, tobramycin and netilmicin, with double these values for amikacin. SI units are not used because gentamicin is not a single entity but a mixture of substances with differing side groups. It, therefore, does not have a molecular weight and so measurement is in mass rather than molar units. Although this does not pertain to tobramycin or netilmicin, the same units are used for convenience. The therapeutic ranges have been based largely on studies that have demonstrated poor outcome associated with suboptimal peak concentrations [6–8].

There are no clearly established therapeutic ranges for once-daily dosing. Some groups have arbitrarily suggested that trough concentrations should be less than 2, 1 or 0.5 mg l⁻¹ [9–11]. However, there is no rationality for this approach because concentrations at 24 h will be unrecordable, using conventional assay technology, in patients with normal renal function [12]. Concentrations of 0.5–2 mg l⁻¹ at 24 h would indicate an overdose in these patients. Clearly the concept of a trough concentration is not relevant to once-daily dosing.

There is little need for monitoring peak concentrations during once-daily dosing. Peak concentrations will always be well in excess of those seen during multiple-daily dosing and therefore within the useful range of bactericidal effect of most bacteria (usually 5–10 times the MIC). Peak concentration measurement is therefore unnecessary except for pharmacokinetic dose individualisation [12].

Individualisation of therapy

Various methods of dose prediction have been used successfully. A summary of these can be found in Begg et al. [1] and Morike et al. [13]. Nomograms relating dosing to estimates of renal function were the first attempts to individualise dosing, utilising the pharmacokinetic relationship between aminoglycoside clearance and renal function.

The next dosing method to gain favour was the dose-individualisation of Sawchuk & Zaske, [14] based on a one-compartment pharmacokinetic model. The advantage of this method was that after the first dose, future doses could be based upon an assessment of the patient’s own pharmacokinetics (i.e. V and CL) for the relevant aminoglycoside. This method was shown to have advantages over nomograms and methods based on physician intuition, in terms of achieving desired peak and trough concentrations [15, 16]. Bayesian methods have also been used in the dose prediction of aminoglycosides [17]. These have the advantage of obtaining useful information from a smaller number of samples than with the Sawchuk & Zaske approach [14], but in practice performance is similar for the two approaches.

With the advent of once-daily dosing, the scene has virtually returned to the formative years of aminoglycoside dosing. Most proponents of once-daily dosing have simply employed an arbitrary initial dose, usually [3–7 14;mg 14;kg⁻¹ day⁻¹, based on the total amount that would have been prescribed over 24 14;h during multiple-daily dosing. This approach to choice of dose seems reasonable, at least for the starting dose. It remains unclear how drug concentrations measured for monitoring should be interpreted, and indeed at what time points drug concentrations should be measured.

Some groups [12, 18, 19,] have advocated sampling at a time or times earlier than 24 h after the dose when the concentration is still measurable. The exact timing of these samples can vary in relation to renal function. MacGowan & Reeves [18] proposed the measurement of a single sample at 8 h. However, it was unclear how dosing should proceed as a result of this estimate, although this method would lend itself to Bayesian forecasting. Nicolau et al. [19] proposed a dose of 7 mg kg⁻¹ followed by the measurement of a single serum concentration between 6 and 14 h after the first dose. This concentration is compared with a nomogram which indicates whether the dose interval should be 24, 36 or 48 h. Although dose intervals longer than 24 h make good theoretical sense, there is little clinical evidence to support this. A recent report noted an 18% incidence of ototoxicity (5/28) in patients with haematological malignancies treated using this approach [20]. This is of some concern and suggests that some patients may be receiving excessive doses.

Begg et al. [12] suggest a target area under the curve (AUC) approach, based on the notion that the same total dose should be given over 24 h as would be given using dose-individualisation with multiple-daily dosing. In order to assess the AUC in a given patient, measuring a concentration at or around 1 h post-infusion and a second concentration somewhere between 6 and 22 h after the first dose is advocated, the time of the second sample adjusted in relation to renal function [12]. From these two concentrations and assuming a one-compartment model, the AUC is calculated. Subsequent doses can be adjusted to achieve any desired AUC. The desired target AUC related to doses of 5, 6 and 7 mg kg⁻¹ day⁻¹ in a patient with a population mean value for V of 0.25 l kg⁻¹ and CL of 4 l h⁻¹ are 72, 86 and 101 mg l⁻¹ h. Bayesian approaches can be used with the AUC method [21] and have the advantage that useful information can be obtained from a single measured concentration although greater accuracy can be achieved if two concentrations are measured.

Dosing in renal impairment

The pharmacokinetics of the aminoglycosides are altered predictably when glomerular filtration is impaired. Patients with renal dysfunction, elderly patients and neonates will often have impaired clearance and dosing should allow for this. It should be noted that although there is a close relationship between aminoglycoside clearance and clinical estimates of renal function such as calculated creatinine clearance, this relationship is by no means exact and extrarenal losses, such as through sequestration into third spaces and removal of fluid from third spaces, can also occur [22].

Once-daily dosing during increasing renal impairment is complicated since the concentration-time profile approaches that of a continuous infusion. Therefore, the advantage of once-daily dosing (high peak concentrations followed by a decline to unrecordable concentrations) is progressively lost with greater renal dysfunction.

Sampling times

With multiple-daily dosing, peak concentrations should be measured at least 0.5 h after the end of the infusion or after a bolus dose to allow for distribution to be complete. The exact time of sampling is not important but it should be recorded accurately.

The ideal time to measure the trough concentration is the end of the dosage interval. Any time within 0.5 h before the next dose is close enough in practice. The exact time should be recorded to allow pharmacokinetic interpretation. For once-daily dosing, the timing of sampling is less clear and is dependent on the method of dose individualisation to be followed.

How often to sample is also an important question. Aminoglycoside therapy is not usually for longer than 7 days except for bacterial endocarditis, or by inhalation for cystic fibrosis. With multiple-daily dosing frequency of monitoring should reflect this clinical situation; i.e. individualised—more often if unstable, less often if stable. With once-daily dosing it is logical to follow the same rules until new information allows these rules to be re-formulated.

For longer term therapy, such as for bacterial endocarditis, repeat monitoring may be performed at increasing intervals if the measured concentrations are relatively stable. Since the target concentrations for the treatment of endocarditis are often lower than conventional target concentrations, the need for stringent monitoring may be less. The aminoglycosides are used in this context more for a greater than additive effect with a β-lactam antibiotic given concurrently than for primary bactericidal efficacy.

Methods of analysis

The aminoglycosides are generally analysed using immunoassays such as EMIT (Enzyme Multiplied Immunoassay Technique Syva and Behring diagnostic products, CA, USA) or FPIA (Fluorescence Polarization Immunoassay; Abbott, Abbott Park, IL, USA). Some aminoglycosides, such as tobramycin, netilmicin and amikacin, can be analysed using h.p.l.c., but this is not feasible for gentamicin because it is not a single substance. The EMIT and FPIA assays are relatively cheap, are not labour intensive, and have good reproducibility. They are not particularly sensitive, however. The limits of quantification are around 0.29 mg l⁻¹ for the FPIA assay and 0.25 mg l⁻¹ for the EMIT assay, levels which may be low enough for monitoring of trough concentrations using older dosing regimens but not low enough for assessing trough concentrations during once-daily aminoglycoside dosing.

There are various quality assurance programmes (QAP) available worldwide. In Australasia the main QAP is that of the Royal College of Pathologists of Australia. In the United Kingdom, the equivalent is UKNEQAS Scheme organised by the Antimicrobial Reference Laboratory, Southmead Hospital, Bristol.

Interpretative problems

A minor laboratory problem relates to inactivation of gentamicin in vitro in the presence of ticarcillin and some other β-lactam antibiotics [23]. The effect is to lower the measured concentration of the aminoglycoside. The longer the time interval between collection and measurement, the greater the effect. Interaction between aminoglycosides and ticarcillin is due to drug complexing and can occur if the two drugs are mixed in a syringe or giving set. Such mixing should be avoided.

Cost effectiveness/benefits of monitoring

Monitoring of aminoglycoside concentrations, particularly with effective dose prediction, has been shown to be cost-effective [24, 25]. Once-daily aminoglycoside dosing is cheaper than multiple-daily dosing in terms of nursing time and infusion equipment and may require less monitoring, although the exact strategy for monitoring has yet to be established [26]. It is possible that total daily doses of aminoglycoside lower than those used traditionally may prove to be effective. If so, toxicity will inevitably be less, and the need for monitoring will also be less. It may be that, except in certain specific circumstances such as rapid fluctuations in renal function or impaired renal function, no monitoring is necessary. However, the conclusion that monitoring is unnecessary is premature.

Vancomycin

Therapeutic drug monitoring is frequently employed during vancomycin therapy. Many have questioned the need for this, especially in uncomplicated patients with normal renal function [27–29]. However, most authors agree that monitoring is often useful.

Pharmacokinetics

The pharmacokinetics of vancomycin are relatively simple, with low protein binding, renal elimination with no metabolism, and no pharmacogenetic problems. V is around 0.4 l kg⁻¹ and the CL approximates that of glomerular filtration rate. t_1/2 is approximately 6 h in patients with normal renal function [2]. A complicating feature is that the pharmacokinetic profile is best described by a 2- or 3-compartment model which makes calculations difficult with a handheld calculator.

Why monitor vancomycin?

Vancomycin has a low therapeutic index, with nephrotoxicity and ototoxicity complicating therapy [30]. It has been traditional to monitor peak and trough concentrations, much as for the aminoglycosides. However, because the bactericidal action of vancomycin is quite different from that of the aminoglycosides, many have questioned the need to measure peak concentrations [31]. Vancomycin, like β-lactam antibiotics, works best if the concentration at the site of activity is maintained above the minimum inhibitory concentration (MIC) throughout the dose interval (so called time-dependent killing). This profile argues against the need for peak concentration measurement, and suggests that a continuous infusion may be the ideal. In practice, a continuous infusion is rarely administered, and spaced dosing at 6 or 12 h intervals is more common [31]. Efficacy can usually be assumed if the trough concentration is sufficiently above the MIC of the infecting organism.

It is likely, although unproven, that toxicity is related to total vancomycin exposure. It would seem logical to monitor some index of vancomycin accumulation, such as a trough concentration.

Therapeutic ranges

Many putative ‘therapeutic ranges’ are illusionary. This is the case with vancomycin, at least for peak concentrations. Ranges for peak concentrations of 20–40 mg l⁻¹ have been widely quoted, but with little supportive evidence. The original report of Geraci et al. [32] suggested that peaks greater than 50 mg l⁻¹ should be avoided, based on two cases of ototoxicity at concentrations greater than 80 mg l⁻¹. Repetitive citation of this paper in the literature has resulted in the ‘establishment’ of the peak concentration strategy. One of the problems with peak concentration measurement is that the range is meaningless unless the timing of sampling is also stated. A peak concentration of 40 mg l⁻¹ has an entirely different meaning if it is from a sample taken just after the end of the infusion or from a sample taken 1 or 2 h later. A survey of Australasian hospitals indicated that peaks were sampled from immediately post-infusion to 3 h later and yet were all considered with reference to the same therapeutic range of 20–40 mg l⁻¹ [31]. The folly of this is obvious. In summary, there is little justification for peak concentration measurement except for use in pharmacokinetic modelling.

A stronger but incomplete case can be made for trough concentration monitoring. The given range of 5–10 mg l⁻¹ has reasonable literature support and reflects the need for the concentration of antibiotics to be above the MIC of the organism for the duration of the dose interval. Concentrations below the MIC have been associated with therapeutic failure. The MIC of vancomycin is approximately 1.5 mg l⁻¹ for many susceptible organisms. The protein binding of vancomycin is approximately 50% which would argue that the minimum total concentration should be at least 3 mg l⁻¹. Trough concentrations above 10 mg l⁻¹ have been associated with an increased risk of nephrotoxicity [30, 33, 34]. It should be noted that nephrotoxicity with vancomycin alone is not common, usually around 5%, and is usually reversible [33]. A range of 5–10 mg l⁻¹ for trough concentrations appears to have some validity.

Individualisation of therapy

Approaches to dosing of vancomycin have included empirical, nomogram, individualised, and Bayesian methods [35]. It is generally accepted that vancomycin should be administered by slow infusion (for at least 1 h) to avoid the ‘red man syndrome’ [36].

Empirical methods usually involve administration of a total daily dose in adults of 2 g daily in 2–4 divided doses. The relationship between vancomycin clearance and renal function has enabled dosing guidelines to take renal function into account [37]. This process has been refined into nomograms such as those of Matzke et al. [38] and Moellering et al. [39].

Individualised dosing methods based on the approach of Sawchuk & Zaske [14] for aminoglycosides have been used with varying success. The approach is based on a one-compartment model, and may lead to underestimation of the AUC [40]. At steady state this is unlikely to create significant error. Bayesian predictive models have also been used [41, 42]. They require a minimal number of samples, and can accommodate 1- or 2-compartment models. Pryka et al. [43] found that the 2-compartment model had less bias and more precision in non steady-state situations, but once steady-state had been achieved the performance for the two models was similar. From a practical point of view, the individualised and Bayesian methods are probably equally useful at achieving target concentrations and subtle differences are unimportant [35]. There is far greater uncertainty about what the target concentrations of vancomycin should be.

Sampling times

The lack of support for peak concentration measurement suggests that only trough concentrations should be monitored. Trough concentrations should be measured just prior to the next dose, within half an hour of the end of the dosage interval. If pharmacokinetic modelling is contemplated, the measurement of another concentration is required early in the dose interval. A trough concentration is all that is necessary for Bayesian forecasting. The exact timing of the sampling is unimportant but needs to be known for interpretation.

The question of how often to sample is also important. As with aminoglycosides, the clinical state may be changing during the early part of therapy. The timing of repeat sampling depends on the clinical circumstances. If very high doses are being used, or if the clinical condition is changing rapidly, sampling should be more frequent.

Methods of analysis

As with the aminoglycosides, vancomycin is generally analysed using immunoassays such as EMIT or FPIA. For research purposes vancomycin is also analysed by h.p.l.c. [44]. The limit of quantification for the EMIT assay is 5 mg l⁻¹ and 2.0 mg l⁻¹ for FPIA. Quality control programs are the same as for the aminoglycosides.

Interpretative problems

Vancomycin degrades spontaneously at 37° C [45]. An inactive metabolic product (CDP-1) accumulates in some circumstances (eg. marked renal failure and patients on dialysis) and is detected by non-specific assays such as FPIA [46].

Cost-effectiveness/benefits of monitoring

Information about the true cost-effectiveness of monitoring of vancomycin is lacking. One study demonstrated a decrease in nephrotoxicity associated with a therapeutic drug monitoring service [47] and another has demonstrated cost savings related to a monitoring service [48].

Teicoplanin

Teicoplanin, like vancomycin, is a glycopeptide antibiotic which acts in a time-dependent manner. Its pharmacokinetics include a V of around 1 l kg⁻¹, protein binding of 90% and a prolonged terminal t_1/2 of 150 to 180 h, which is important during long term therapy [49]. Teicoplanin is cleared unchanged renally, and doses should be reduced appropriately in patients with renal dysfunction. Dose-related nephrotoxicity and ototoxicity appears to be much less of a problem than with vancomycin [50].

Many methods have been used to quantify teicoplanin concentrations but none is entirely satisfactory for routine clinical monitoring. [49]. FPIA is the most convenient method, but it may not be accurate at low concentrations [51]. Therapeutic drug monitoring of teicoplanin is performed occasionally to ensure efficacy. Trough concentrations of >10 mg l⁻¹ have been recommended for most infections and >20 mg l⁻¹ for endocarditis [52]. Monitoring has been advocated particularly in patients who have high clearance, such as children [53]. ‘Red man syndrome’ appears to be much less of a problem with teicoplanin than with vancomycin [54].

Antifungal agents

Individualisation of antifungal therapy has been proposed and reviewed [52, 55]. No specific recommendations have been developed, although under certain circumstances, drug monitoring may assist therapy. The case for monitoring is perhaps strongest for flucytosine and itraconazole. There may be a case for occasional monitoring of other azoles, such as ketaconazole and fluconazole, to ensure that adequate concentrations are achieved. As is the case with all therapeutic drug monitoring, if it provides an answer to a clearcut question, then it is worthwhile.

Flucytosine

Flucytosine is a synthetic antifungal agent that is selectively converted by fungal cells to fluorouracil which replaces uracil thereby disturbing protein synthesis. Flucytosine is well absorbed, with a V similar to that of body water and it is eliminated unchanged through the kidneys. Doses should be appropriately adjusted for renal dysfunction [56]. Monitoring with flucytosine is not so much to ensure adequate concentrations for efficacy but to avoid toxicity. At high concentrations, bone marrow toxicity (usually reversible) has been reported [57].

Stamm et al. [58] recommended avoiding peak concentrations of >100 mg l⁻¹ (2 h after an oral dose, or 30 min after an i.v. dose) to prevent bone marrow toxicity. They advocated more frequent monitoring in patients with renal dysfunction or existing bone marrow depression. Toxicity appears to be a particular problem when concentrations rise as a result of renal insufficiency induced by amphotericin B which is often used concurrently [56]. While a lower end of the ‘therapeutic range’ has not been established for flucytosine, concentrations <25 mg l⁻¹ should be avoided to prevent emergence of resistance [59].

Itraconazole

Itraconazole is a relatively new synthetic triazole antifungal agent. It is used increasingly in leukaemia and AIDS patients and during transplantation, especially for the treatment and prophylaxis of aspergillus infections.

Itraconazole is extensively metabolised, has non-linear pharmacokinetics and is subject to enzyme inhibition. [56, 60]. Absorption is variable, being poor in the unfed state (high gastric pH) and in patients with AIDS or bone marrow transplants [61]. These are all good reasons for therapeutic drug monitoring, especially to ensure that concentrations are adequate. Trough concentrations of itraconazole above 250 μg l⁻¹ have been suggested as satisfactory for most fungal infections [62]. A combined total concentration of 1000 μg l⁻¹ for the parent plus active metabolite hydroxyitraconazole, has also been recommended [63]. Adverse effects related to itraconazole are largely gastrointestinal and appear to be dose-related.

A problem during therapeutic drug monitoring is that concentrations are assay-dependent, with the metabolite being recorded as parent in non specific microbiological assays. The preferred method of analysis is h.p.l.c., preferably with both parent and metabolite assessed [64].

Further work is necessary before the place of therapeutic drug monitoring for the therapy of itraconazole is firmly established.

Conclusion

Aminoglycosides should continue to be monitored until it is clear that there are circumstances where monitoring is unnecessary. Justification for vancomycin monitoring is less clear, and there may be circumstances, such as short courses and short dose intervals in relatively uncomplicated patients, in which monitoring may not be necessary. For the other drugs it is likely that monitoring is only feasible in larger teaching hospitals where expertise is available for useful interpretation.