Article Text

Download PDFPDF
Therapeutic drug monitoring in neonates
  1. Steven Pauwels1,2,
  2. Karel Allegaert3,4
  1. 1Department of Cardiovascular Sciences, KU Leuven, Leuven, Belgium
  2. 2Department of Laboratory Medicine, University Hospitals Leuven, Leuven, Belgium
  3. 3Department of Development and Regeneration, KU Leuven, Leuven, Belgium
  4. 4Neonatal Intensive Care Unit, University Hospitals Leuven, Leuven, Belgium
  1. Correspondence to Dr Karel Allegaert, Neonatal Intensive Care Unit, University Hospital, Herestraat 49, Leuven 3000, Belgium; karel.allegaert{at}


Therapeutic drug monitoring (TDM) aims to integrate drug measurement results into clinical decision making. The basic rules apply when using TDM in neonates (aminoglycosides, vancomycin, phenobarbital, digoxin), but additional factors should also be taken into account. First, due to both pharmacokinetic variability and non-pharmacokinetic factors, the correlation between dosage and concentration is poor in neonates, but can be overcome with the use of more complex, validated dosing regimens. Second, the time to reach steady state is prolonged, especially when no loading dose is used. Consequently, the timing of TDM sampling is important in this population. Third, the target concentration may be uncertain (vancomycin) or depend on specific factors (phenobarbital during whole body cooling). Finally, because of differences in matrix composition (eg, protein, bilirubin), assay-related inaccuracies may be different in neonates. We anticipate that complex validated dosing regimens, with subsequent TDM sampling and Bayesian forecasting, are the next step in tailoring pharmacotherapy to individual neonates.

  • Neonatology
  • Pharmacology
  • Toxicology

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic

  • The general rules for therapeutic drug monitoring (eg, narrow therapeutic index, good correlation between concentration and effect, limited predictability) also apply in neonates.

  • However, neonates do have some specific characteristics that warrant particular attention.

What this study adds

  • Because of large variability in pharmacokinetics in neonates, therapeutic drug monitoring (TDM) is more common in these patients, especially as validated dosing regimens are lacking.

  • Following initiation of treatment, the timing of TDM sampling is crucial since sampling before steady state has been reached in more likely in neonates.

  • Because of differences in composition (eg, protein, bilirubin), there may be specific TDM assay inaccuracies in neonatal samples.


Drug therapy is a powerful tool for improving outcome. The purpose of administering a given dose of a given drug is to achieve a particular effect, commonly through attaining a therapeutic concentration range, while avoiding disproportionate side-effects in individual newborns. Moreover, if the therapeutic target concentration is known, it should be achieved as quickly and consistently as possible, irrespective of the indication (eg, infectious disease, seizures, cardiac arrhythmia).1 ,2 However, due to maturational changes, there is large inter- and intra-individual variability in pharmacokinetics (PK; concentration-time profile) in neonates.3–5 Renal elimination clearance in early life is low and mostly depends on glomerular filtration, which is affected by age, co-administration of non-selective cyclo-oxygenase inhibitors, growth restriction and peripartum asphyxia.6 ,7 Phenotypic drug metabolism is influence d by constitutional (eg, age), environmental (eg, drug–drug interaction, feeding) and genetic (eg, polymorphisms) characteristics.4 ,5

This limited predictability and wide variability can be partly managed through the use of more complex validated dosing regimens as recently demonstrated for amikacin and morphine.8 Therapeutic drug monitoring (TDM) can contribute to tailored drug prescribing, with individualised dosing in neonates used to maximise therapeutic benefits while minimising toxicity or side effects.1 ,9 In addition to therapeutic drug measurement, TDM supports clinical decision-making and is used to improve therapeutic responses or reduce adverse events.10 We examine the principles of the appropriate use and interpretation of TDM in neonates, and highlight some aspects important in the clinical decision-making process. This discussion will be based on general rules for assessing TDM usefulness or futility (box 1) with specific emphasis on neonates.

Box 1

Basic rules to assess the usefulness or futility of therapeutic drug monitoring (TDM)

Reasons for using TDM

  • There is only a weak correlation between the dose administered and the concentration reached (ie, poor predictability), and concentration is related more closely to the therapeutic effect or toxicity than the dosage.

  • There is a quantitative relationship between concentrations (eg, plasma) and (side) effects, the therapeutic concentration range is narrow, and underexposure and overexposure (resulting in poorer outcome and more toxicity, respectively) should be avoided.

  • As neonatal pharmacokinetics and associated factors are have been sufficiently well characterised, the results of TDM can be used to adjust treatment.

  • The analytical technique is sufficiently specific, precise and accurate in the population, is cost effective, and takes account of the characteristics of neonatal samples (eg, low volume, hyperbilirubinemia, plasma protein).

Reasons for not using TDM

  • The value of TDM is limited and there are more convenient methods for assessing the effects of dosage if clinicians can titrate based on easily available outcome variables like blood pressure, analgesia or level of sedation.

  • There is a broad concentration range before toxicity or subtherapeutic levels are evident (eg, penicillin in neonates).

  • There are uncertainties about dosage (administration, product quality), sample collection (timing) and assay validity.

CRITERION 1: interindividual variability in PK

As mentioned earlier, drug disposition in neonates varies greatly1–3 and depends on maturation as well as non-PK factors, such as accuracy of dose administration and preparation.11 Specific formulation-related issues in neonates (eg, highly concentrated formulations, low doses, low infusion rates, dead spaces) mean non-PK factors are important in this population.12 Maturational PK variability and non-PK factors can result in a poor correlation between dose administered and concentration achieved in neonates.

This problem can be partly overcome by the use of more complex dosing regimens once they have been validated.5 ,7 ,8 Prospective evaluation of a model-based neonatal amikacin dosing regimen with 10 different dosing suggestions in neonates (based on weight and postnatal age) resulted in target peak and trough concentrations being attained in almost all (>90%) patients.13 ,14 In contrast, simple standardisation as provided in the National Institute of Health and Care Excellence (NICE) guidelines (5 mg/kg gentamicin every 36 h in neonates) results in only 65% of peak (8–12 mg/L) and trough levels (1 mg/L) being achieved.15 These contrasting results should be considered when making clinical decisions: one may either use a more complex dosing regimen with TDM monitoring restricted to specific settings (eg, peripartum asphyxia, cooling, ibuprofen, extracorporeal membrane oxygenation) or use a very simple standardisation dose with systematic TDM monitoring and subsequent dose adjustment.2 ,13 ,15 ,16 The disadvantage of the second approach is that the effective concentrations are only reached later.

A recently published audit on vancomycin trough levels using the British National Formulary for Children documented that only 13% of the first vancomycin trough levels were within target (10–15 mg/L), reflecting the need to adjust dosing guidance.17 These findings strongly support the use of TDM for vancomycin in neonates until more sophisticated dosing regimens have been validated.1 ,6 ,18–20

CRITERION 2: the target range is known and narrow

Because of immaturity, co-morbidity from inappropriate drug selection (eg, resistant pathogen) or insufficient dosing it is not always easy to identify therapeutic failure in neonates. Although clinicians may feel confident using reference handbooks for drug dosing, there is still important and unexplained variability between dosing regimens suggested in handbooks and the results seen in clinical practice.8 ,18 ,21 Moreover, drugs also have side-effects, which may be reduced with the use of TDM.

Extended interval dosing for aminoglycosides has resulted in higher peak and lower trough levels in neonates. This approach is supported by observations of reduced nephrotoxicity in children, but there is only indirect (ie, more TDM observations on target) evidence in neonates.22 Extended interval dosing for aminoglycosides results in the administration of higher doses in order to achieve higher peak concentrations (Cmax, improved bactericidal effect) and a Cmax/pathogen specific minimal inhibitory concentration (MIC) ratio >8, with longer time intervals between administration resulting in lower trough concentrations (less adaptive resistance, reduced toxicity).2 ,23 Microbial less adaptive resistance is likely more relevant in neonates because neonates are presumed to be less susceptible to aminoglycoside-related oto- and nephrotoxicity, due to maturation-driven transporter activity and subsequent more limited capacity for intracellular accumulation and toxicity.9 ,23

In contrast, advice on vancomycin target concentrations is more uncertain in neonates. Unlike aminoglycosides, vancomycin also has a time-dependent effect. Consequently, a 24 h area under the curve (AUC24h) divided by MIC ((AUC24h/MIC) >400) has been suggested, although this target has been validated in adults with invasive Staphylococcus aureus infections.1 ,9 The subsequently suggested target trough level (10–15 mg/L) is a simple mathematical translation from this AUC target (400), assuming that the specific MIC value of the pathogen is <1. Consequently, when continuous vancomycin is used, the same AUC target results in a higher target (median, not trough) vancomycin concentration of 16–18 mg/L (400/24 h).20 Moreover, this drug is commonly administered for bloodstream infections (other site) with Staphylococcus epidermidis (other pathogen) in neonates. Finally, vancomycin is in part protein bound in plasma, and free fractions are likely higher in neonates. It is the free drug that interacts with the pathogen, diffuses to the tissues and is available for renal elimination.6 ,9 ,24 Based on these arguments, specific PK/pharmacodynamic targets for vancomycin in neonates should be validated. Currently, there is an ongoing effort to do this through the Neovanc research initiative.19

Finally, target concentrations for efficacy or toxicity may also depend on other clinical characteristics. This has been nicely illustrated for phenobarbital used to control seizures in neonates. Van den Broek documented that the same phenobarbital concentration was more effective in the setting of hypothermia compared to normothermia (‘thermopharmacology’).25 Similar, digoxin toxicity depends on the associated potassium levels.1

CRITERION 3: factors affecting PK are sufficiently well known

In considering a TDM result, we should be aware of various other important factors such as growth/weight and development/age, which display co-linearity.8 ,26 Uncertainties about drug dosage or time of sample collection affect the usefulness of TDM in neonates. Sampling interval and the timing of sample collection after treatment initiation are both very important because of the steady state condition where overall drug intake is in balance with its subsequent elimination.2 ,15 For intermittent and continuous administration using regular dosing without a loading dose, steady state is reached after about four times the elimination half-life.

The prolonged elimination half-life (lower clearance and higher distribution volume) is very relevant in neonates as illustrated in figure 1 for the vancomycin TDM sampling strategy. The concentration-time profile in an individual newborn exposed to 10 mg/kg vancomycin every 8 h has been modelled.7 Figure 1 shows that accumulation takes time, resulting in different trough concentrations when TDM sampling is performed at 24, 36 or 48 h (arrows). In contrast, when a loading dose (20 mg/kg) followed by continuous administration is applied (figure 1), steady state is reached sooner. Similar concerns apply to aminoglycosides regarding decisions whether to perform TDM early, but not yet in steady state, or late, during steady state but after initiation of treatment with the risk of either prolonged over- or underexposure.2 ,15 ,16

Figure 1

The estimated concentration-time trends in a neonate treated with 10 mg/kg vancomycin intravenously every 8 h.7 Because attainment of steady state takes time, different trough concentrations are predicted when samples are collected at 24, 36 or 48 h (A). When a loading dose (20 mg/kg) is used, the steady state is reached much more quickly (B).

Rule 4: the analytical techniques are specific, precise and accurate in neonatal samples

Clinicians generally feel comfortable with a reported TDM value, but likely fail to appreciate inaccuracies related to the technique. Bio-analysis has general as well as specific neonatal issues. Similar to the clinical factors we commonly collect (eg, gestational age, weight, length, or blood pressure) differences in immunoassays and analytical platforms may affect comparability and precision, with for example, more extensive between-assay differences for digoxin, phenobarbital, tobramycin and vancomycin.27 Between-assay differences in vancomycin plasma concentrations of up to 20% were documented.24 This is further illustrated in figure 2, which shows the range of results found with different assays as well as within the same assay applied in different Belgian laboratories to measure one specific serum vancomycin sample (target value: 21.9 mg/L).27 This range may affect interpretation as the therapeutic range and dosing regimen are commonly derived using one specific assay, while a specific assay is not always correlated with another assay.24 ,28

Figure 2

Range of results for a specific serum vancomycin sample measured using different assays (the 12 different assay numbers are shown on the x-axis) and with the same assay but in different laboratories (box-whisker plot for every assay where n is the number of participating laboratories using the same assay) (the target value of 21.9 mg/L is indicated by broken line).28

Some assays display interference due to cross-reactivity with target drug metabolites or interference from other compounds present in plasma, as observed for vancomycin.29 This has been demonstrated by Zhao et al who evaluated the predictive performance of different published PK models for vancomycin in neonates. The limitations of the transferability of a vancomycin PK model to another dataset were explained by the analytical techniques used for creatinine and vancomycin.6 In addition, neonatal plasma has a different biological (eg, proteins, bilirubin) and xenobiotic (eg, other drugs, different concentrations or metabolites) composition compared to adult plasma. This may also result in different values caused by these matrix effects. Finally, commercial assays are not (yet) available for some drugs.30–32

Liquid chromatography-tandem mass spectrometry (LC-MSMS) assays are gaining popularity for addressing most of these issues.30–32 These assays require lower sample volumes and are less prone to interference or cross-reactivity, while method development for most analytes is relatively straight forward. As these methods are commonly developed in house, extended validation is needed to cover issues like imprecision, accuracy, limit of quantification and matrix effects.30–32 The use of other matrices, like dried blood spots, may have benefits as discussed by Patel et al.33


The large variability in PK makes TDM attractive, but specific aspects of its use in neonates should be considered. In tables 1 and 2, we summarise current and future TDM use in neonates, and include our opinions on routine or targeted TDM. TDM is commonly used in neonates for aminoglycosides (targeted) or vancomycin, phenobarbital and digoxin (routinely), but future TDM use in neonates may also depend on new drugs or practices,2 ,13 ,15 ,16 ,34–36 as there is accumulating evidence concerning concentration–efficacy and concentration–toxicity relationships.35 ,36

Table 1

Examples of current practice and our opinion on routine or targeted therapeutic drug monitoring (TDM) in neonates

Table 2

Examples of potential future indications for considering therapeutic drug monitoring (TDM) in neonates

Once complex dosing regimens have been validated (eg, tobramycin, amikacin), more targeted, restrictive TDM seems reasonable, allowing this method to be considered in more specific settings, like prolonged administration in specific subgroups (ie, renal impairment due to extremely preterm birth, peripartum asphyxia or extracorporeal membrane oxygenation).1 ,2 ,8 This is not yet the case for vancomycin, as reflected in the very common practice of performing routine TDM for vancomycin in neonates.9 ,19 Because of the PK concept of steady state, the timing of sample collection after initiation of treatment is also important. Finally, bio-assays also display inaccuracies, and these inaccuracies may even be in neonatal plasma. These aspects were highlighted to illustrate the potential value of multidisciplinary collaboration between neonatology, clinical pharmacology and laboratory medicine to further tailor TDM to the needs of neonates.1 ,2 ,9 ,10

An emerging concept to further individualise drug therapy is the Bayesian forecasting approach, which combines TDM measurements or surrogates (eg, INR) with relevant clinical factors (neonatal age, weight, creatinine and genetic polymorphisms). Bayesian forecasting uses population-specific PK (ie, neonates) as a starting point. Using specific patient factors (age, weight, disease characteristics), individual PK estimates can be obtained and subsequently improved by integrating the individual TDM observations in the estimates.37 We refer to the previously mentioned model-based amikacin dosing regimen as an example.13 ,14 The initial prescription can be based on the information available (weight, postnatal age, ibuprofen co-administration and peripartum asphyxia). However, after TDM sampling, the individual dose can be tailored to the individual observations, also taking into account the difference between predicted and observed TDM.13 ,14 ,37

Bayesian forecasting is the next step after the ‘one dose fits all’ approach,13 introducing complex dosing regimens15 ,16 ,37 for individualised dosing.10 ,37 ,38 Various software programmes which can be accessed online, are emerging to facilitate its clinical application.10 ,37 ,38 In adult intensive care patients, Pea et al39 documented that a TDM-guided approach coupled with dose adjustment based on Bayesian forecasting was an effective tool to improve vancomycin dose accuracy. A similar forecasting model has recently been suggested to result in more efficient dose individualisation of warfarin in adults and children.40

In conclusion, TDM should be tailored to the needs of neonates if the clinician wishes to integrate drug measurements into the clinical decision-making process. Pharmacokinetic and non-PK variability can only be partly managed by more complex, validated dosing regimens, with the timing of TDM sampling following the initiation of treatment being more important in neonates. The target concentration may be uncertain (vancomycin target) and inaccuracies in TDM assays may be specific for neonatal samples. The use of more complex dosing regimens, with TDM sampling combined with Bayesian forecasting PK programmes, are the next steps in tailoring drug therapy to the needs of neonates.


We thank Dr Christel Van Campenhout and Dr Marianne Demarteau for permission to use a figure from the report from the Belgian External Quality Evaluation Scheme of the Scientific Institute of Public Health.28



  • Funding SP and KA are supported by the Fund for Scientific Research, Flanders (Clinical Fellowship 1700314N and Fundamental Clinical Investigatorship 1800214N, respectively). SP is supported by the Fund for Clinical Research from the University Hospitals Leuven. The research on neonatal clinical pharmacology is further supported by the Agency for Innovation, Science and Technology in Flanders (IWT) through the SAFEDRUG project (IWT/SBO 120033).

  • Competing interests None declared.

  • Provenance and peer review Commissioned; externally peer reviewed.