Clinical pharmacology in neonates: small size, huge variability

Abstract

Drug therapy is a powerful tool to improve neonatal outcome. Despite this, neonatologists still routinely prescribe off label compounds developed for adults, and extrapolate doses from those used in children or adults. Knowledge integration through pharmacokinetic (PK) modeling is an important method to improve the current situation. Such predictive models may convert neonatal pharmacotherapy from explorative to confirmatory. This will be illustrated by two research projects related to the prediction of neonatal renal clearance and neonatal glucuronidation. Finally, these kind of models will also improve the current knowledge of neonatal (patho)physiology. In the meanwhile, both the field of clinical pharmacology (e.g. PK/PD modeling, pharmacogenetics) and neonatology (e.g. whole body cooling, lower limit of viability) matured, resulting in new research topics. However, both the modeling as well as the newly emerging topics need to be tailored to the characteristics of neonates to turn them into effective tools. Consequently, the field of neonatal pharmacotherapy needs dedicated neonatologists who continue to raise awareness that off label practices, eminence based dosing regimens and the absence of neonatal drug formulations, all reflect suboptimal care.

Introduction

When a compound is prescribed, the aim is to attain targeted effects (e.g. bactericidal, analgesic, blood pressure normalisation), preferably without disproportional side-effects (e.g. drug toxicity, hypotension, tachycardia). Clinical pharmacology aims to estimate these (side)-effects at the level of the population or – preferably - the individual [1,2]. Pharmacokinetics (PK) describes the relationship between a concentration in a specific compartment (e.g. plasma, cerebrospinal fluid, bronchial epithelial lining fluid) and time (concentration/time, ‘what the body does to the drug’). Pharmacodynamics (PD) describe the relationship between a concentration in a specific compartment and (adverse)-effects (concentration/effect, ‘what the drug does to the body’). The concepts of clinical pharmacology obviously also apply to neonates, but their physiological characteristics warrant a tailored approach [3,4].

Covariate analysis is an important tool to translate information from the newborn population into the individual neonate in need of individualized pharmacotherapy [3,4]. Covariates are specific characteristics that explain in part the inter- and intra-individual PK/PD variability. The most obvious covariates in neonates relate to growth and development, reflected and quantified by birth weight, current weight, or age – either postnatal, gestational or postmenstrual age. There is already at least one order of variability in weight (<0.5 up to 5 kg) while both the height velocity rate (10-20 cm/year) and the increase in body weight (50 % increase in the first 6 weeks) reflect the dynamics of a rapidly evolving biological system in perinatal life. The maturation related variability is further aggravated by interfering disease characteristics (e.g. renal failure, sepsis, growth restriction) or treatment modalities (e.g. co-medication, extracorporeal membrane oxygenation, whole body cooling). Moreover, maturation (e.g. receptor expression, receptor activity, cellular metabolism, enzyme activity) interrelates with growth. Some tissues may be more sensitive to specific compounds in early life, irrespective of a given concentration or exposure, whereas others will be less sensitive. This will affect population specific PD. Beyond clinical pharmacology, this is in fact the most crucial characteristic of perinatology: despite the limited size, there is extensive variability, related to rapidly evolving physiology [3,4].

Drug therapy is a powerful tool to improve neonatal outcome. Despite this, neonatologists still routinely prescribe off label compounds, extrapolate doses and indications from those used in children or adults without fully considering perinatal physiology. PK/PD modeling is an emerging tool to improve the current situation since modeling may convert neonatal pharmacotherapy from explorative to confirmatory. This will be illustrated by modeling projects on neonatal renal clearance and neonatal glucuronidation. Such models also improve the knowledge of neonatal (patho)physiology. This will be illustrated by recent data on maturational differences in renal compared to hepatic elimination routes and drug-drug interactions. Finally, both clinical pharmacology (e.g. PK/PD modeling, pharmacogenetics) and neonatology (whole body cooling, lower limit of viability) matured, resulting in new research topics.

Modeling as a structured approach to improve knowledge of pharmacotherapy

A powerful tool to improve neonatal pharmacotherapy and facilitate clinical studies is knowledge integration through pharmacokinetic (PK) modeling. PK modeling is through mechanism based PK or physiology-based (PB) PK [4-7]. Mechanism based models apply a bottom-up ‘from compound to model’ concept: based on drug specific observations, covariates are described, resulting in mechanism-based models. PB-PK applies a top-down ‘from physiology to clinical observations’ concept: based on available data on neonatal physiology (e.g. weight, cardiac output, renal function), a PB-PK model is developed [4-7]. These models hold the promise to predict PK/PD of compounds if the normal clearance routes are known. Such approaches were recently reported for drug related clearance through either glomerular filtration rate (GFR) [8,9] or glucuronidation in early life [10].

For the renal model, a covariate model characterizing developmental changes in clearance of amikacin in neonates was developed based on birth weight, postnatal age and ibuprofen exposure [8]. Assuming that such a model contains physiological information on GFR ontogeny, the amikacin covariate model was subsequently applied to datasets of other aminoglycosides and glycopeptides (netilmicin, gentamicin, tobramycin and vancomycin). It was hereby documented that the descriptive and predictive properties of the models developed using the amikacin covariate model were good, and fairly similar to the independent reference models and similar clearance values were obtained using both approaches [9]. This study hereby confirmed that neonatal covariate models may contain physiological information since information derived from one drug can be used to describe other drugs. This approach may be used to optimize sparse data analysis and to derive individualized dosing algorithms for drugs in newborns. Such dosing algorithms obviously need prospective validation, since Zhao W et al. illustrated the potential hazards related to the transfer of published vancomycin models to different clinical settings. The predictive performance of 6 earlier published models was evaluated and differences in predictive performances of models for vancomycin pharmacokinetics in neonates were found. This means in essence that a model described in a cohort in a given hospital does not automatically predict the data in another cohort in another hospital. In this specific example, it was concluded that dosage individualisation of vancomycin in neonates should consider not only clinical characteristics, but also the methods used to measure serum creatinine and vancomycin [11]. A similar effort has been reported for glucuronidation, based on morphine glucuronidation in neonates, subsequently extrapolated to zidovudine glucuronidation [10].

Developmental pharmacology reflects developmental physiology

The different routes of elimination do not mature simultaneously but all have their process specific maturational pattern (e.g. glomerular filtration versus renal tubular excretion, CYP2D6 versus CYP3A4/5, or renal versus hepatic clearance) [1-4]. This implies that it is important to integrate ontogeny-related knowledge of the different elimination pathways to predict compound specific, phenotypic in vivo observations in neonates: there is no such thing as a neonatal kidney or liver, but only newborns in need of improved pharmacotherapy. This also means that specific settings (renal failure, hepatic failure or drug-drug interactions) described in children or adults need to be interpreted cautiously in the newborn by taking the available knowledge on developmental physiology into account.

Metabolites like O-desmethyl tramadol (analgesia), morphine glucuronides (analgosedation) or 1-hydroxy-midazolam (sedation) may accumulate in neonates not because the metabolic clearance is already very effective, but because the subsequent renal elimination of the metabolites is even less effective [12]. The same holds true for propylene glycol, an alcohol commonly used as an excipient in drug formulations. In adults, primary renal elimination accounts for 45 %, hepatic metabolic clearance through alcohol dehydrogenase for 55 %. In neonates, the contribution of primary renal elimination was much more limited (15-25 %) while the major route for clearance was through hepatic metabolism [13]. The phenotypic result is that propylene glycol clearance in neonates is low, and strongly depends on hepatic metabolic capacity. This likely explains the side effects described with Kaletra syrup [1 ml contains 80 mg of lopinavir co-formulated with 20 mg of ritonavir + excipients, including 356.3 mg of ethanol (42.4% v/v) and 152.7 mg of propylene glycol (15.3% w/v)] administration in neonates. This formulation contains propylene glycol and ethanol and simultaneous exposure results in accumulation through competition for hepatic clearance. In contrast, associated renal failure, commonly described as a risk factor for propylene glycol accumulation in adults is likely of limited relevance in neonates [13].

Similar to excipient-excipient interactions, drug-drug interactions are only anecdotically explored in neonates. Salem et al. recently reported a ‘PBPK-model’ to predict drug-drug interactions throughout pediatric life [14]. For a theoretical compound metabolized 50 % by CYP2D6 and CYP3A4 pathways at birth, co-administration of ketoconazole (3 mg/kg, CYP3A4 inhibitor) resulted in a 1.65-fold difference between inhibited versus uninhibited concentration time profiles compared to 2.4-fold in 1 year olds and 3.2-fold in adults. Obviously, neonates could be more sensitive to such interactions (PD) than adults.

Emerging clinical research topics

Pharmacogenetics

The idea of individualized clinical pharmacology through integration of pharmacogenetics (PG) reflects the fact that specific (side)effects are not merely randomly distributed. Pharmacogenetics explores interindividual differences in drug response related to genetic variations, i.c. polymorphisms. Genetic variations can affect drug disposition through modifying receptor sensitivity or differences in drug metabolism. This obviously holds the promise to tailor perinatal clinical pharmacology beyond the usual covariates like age, weight or disease characteristics [15]. There are illustrations on the integration of PG and age-related maturation (i.e. ontogeny) to improve prediction of phenotypic drug metabolism [cytochrome P450 (CYP) C219, CYP 2D6, or N-acetyl transferase (NAT) 2] [16,17,18]. Based on in vivo observations of pantoprazole (CYP 2C19), tramadol (CYP2D6) and isoniazide metabolic clearance respectively, an age related impact of specific polymorphisms was documented. Obviously, this is limited to iso-enzymes that already are sufficiently active in early life. This likely explains the absence of a link between CYP2C8 and CYP2C9 polymorphisms and the response to ibuprofen to induce closure of the patent ductus arteriosus [19].

More importantly, PG studies in perinatal life should go beyond confirmation of associations described in adults and explore the impact of PG as covariate limited to perinatal life during which the genotype-phenotype concordance still exists [15,16]. This includes issues like fetal malformations, breastfeeding or neonatal clinical syndromes. A recent illustration of such an approach is the impact of polymorphisms on the severity of neonatal abstinence syndrome following maternal opioid intake [20]. While the median length of stay for term neonates was 35 days, specific Cathechol-O-methyltransferase (COMT, 158A>G) and μ–opioid receptor (OPRM1, 118A>G) polymorphisms were associated with a median reduction of 10.8 and 8.5 days respectively. Interestingly, breastfeeding itself (univariate) resulted in a reduction of 18 days [20].

This confirms findings reported by the group of Koren and coworkers on the links between genetic polymorphisms in mothers and their infant and the variation in response to standard doses of maternal opioids (i.c. codeine) [21]. The authors hereby explored the associations between polymorphisms in CYP2D6, UDP-glucuronosyltransferase 2B7 (UGT2B7), P-glycoprotein (ABCB1), OPRM1, COMT genes and central nervous depression in 111 infants during breastfeeding. A model combining specific maternal risk genotypic polymorphisms (CYP2D6 and ABCB1) was associated with central nervous depression in both mothers (OR 2.74) and their offspring (OR 2.68) [21].

Thermopharmacology in neonates

Therapeutic hypothermia is an effective and valid treatment in term newborns following perinatal asphyxia. This relates to the effect of hypothermia on cerebral metabolism, but hypothermia itself also affects physiological functions, such as circulation and metabolic activity [22]. Consequently, this modality may affect physico-chemical properties and the PK/PD of the drugs. These aspects are covered by thermopharmacology, a term recently introduced into neonatal pharmacology by van den Broek et al [23,24]. However, in the current practice, hypothermia is an additional intervention in neonates with peripartal asphyxia. Based on the currently available in vivo observations, it seems that primary renal elimination of e.g. aminoglycosides is not further reduced when hypothermia is applied and that asphyxia itself already reduces clearance [25]. In contrast, metabolic clearance seems to be reduced as illustrated for lidocaine (dependent on hepatic blood flow), morphine (glucuronidation) or phenobarbital (dependent on hepatic drug metabolism) [23,24]. Besides PK effects, there are also some PD related effects since the transition rate from a continuous normal voltage to discontinuous normal voltage aEEG background level seems to be reduced in hypothermic asphyxiated newborns when compared to normothermic setting. Similar, bacterial growth also depends on temperature, and hypothermia may affect the PK/PD relation of antibiotics in neonates.

Clinical pharmacology at the limit of viability

The overall outcome - including both mortality as well as (co)morbidity - in extreme preterm neonates strongly depends on the age at birth, weight at birth and gender. Despite this, the pharmacological tools commonly used at the limit of viability when compared to less immature neonates are not fundamentally different, the dosing is commonly extrapolated form neonatal practices in less immature neonates while none of these practices takes gender specific modalities into account. Perhaps we should consider additional subpopulations within neonatal intensive care. The post hoc analysis on the age-dependent magnitude and direction of neurodevelopmental outcome following prophylactic thyroid hormone supplementation may hereby serve as an illustration [26]. A gestational age-dependent effect of thyroxine on neurodevelopmental outcome was found in post-hoc subgroup analyses up until the age of 10 years. Thyroxine treatment was associated with improved mental, motor, and neurological outcomes in infants <28 weeks gestation, but with worse mental and neurological outcome in infants of 29 weeks gestation or beyond [26].

Drug therapy is a powerful tool to improve neonatal outcome, but may also result in side effects. Adverse drug reactions (ADR) in critically ill newborns may produce harm that further increases morbidity or mortality. However, differentiation of ADRs from reactions related to the immaturity or disease (i.e. renal, hepatic dysfunction) is difficult, while we take it for granted that the overall rate of (co)morbidity in extreme preterm neonates at the limit of viability is high [27]. Similar to pharmacogenetics, pharmacovigilance also needs to be tailored to the specific characteristics of newborns. The decades needed to document appropriate oxygen targets for preterm infants, or more recently, the dexamethasone observations or the trends towards less invasive ventilation strategies should at least raise awareness that any treatment intended to improve outcome in neonates, may in itself have unanticipated (long term outcome) side-effects [28,29,30]. Unfortunately, in a recent systematic review on perinatal interventions, Teune et al. documented that only 40/249 (16 %) randomized controlled trials followed neonates after discharge from the hospital, without any improvement in consecutive time intervals (15 % before 1990, 15 % between 1990-2000 and 19 % since 2001) [31].

Concluding remarks

Effective pharmacotherapy depends on predictable pharmacokinetics and –dynamics. For both aspects, newborns can differ significantly from children or adults. Extrapolation of safety and efficacy can be made more reliable when methods developed in other fields of clinical pharmacology can be integrated in neonatal clinical research and care. These methods include modeling, pharmacogenetics and pharmacovigilance, but will need to be tailored to the needs and characteristics of neonates to turn them into effective tools. Consequently, the field of neonatal pharmacotherapy needs dedicated neonatologists who continue to raise awareness that off label practices, eminence based dosing regimens and the absence of neonatal drug formulations remain reflections of suboptimal care.