Drug metabolism

Two phases of drug metabolism are traditionally distinguished—phase I and phase II—which generally occur sequentially. Many drugs undergo both phase I and phase II, while others only undergo either phase I or phase II. Phase I is characterised by modification reactions, whereby the parent drug is modified by adding reactive and polar groups, resulting in a more hydrophilic compound; in some cases, a polar group is unmasked or revealed by drug metabolism. Phase II metabolism is characterised by conjugation. Metabolites resulting from phase I metabolism may be conjugated to even more polar, more readily excreted large compounds. This oversimplification of drug metabolism does not imply that only the parent compound itself is active and that metabolites are not. Some drugs are inactive and need to be metabolised to exert their effect, for example, carbamazepine, enalapril and oseltamivir. Other drugs may have metabolites with the same therapeutic effect as the parent compound, such as morphine and its metabolite morphine-6-glucuronide or venlafaxine and desvenlafaxine. Finally, some metabolites may be responsible for adverse drug reactions. For example, paracetamol is metabolised to the reactive metabolite N-acetyl-p-benzoquinone imine (NAPQI), which is associated with acute liver injury after overdoses.4

Age in relation to drug metabolism

As many enzymes known to metabolise drugs are also involved in the metabolism of endogenous compounds, such as cortisol, dehydroepiandrosterone, testosterone, vitamins A and D and cholesterol, it is not surprising that their activities may change during childhood consistent with their roles in the regulation of growth and development. Ontogeny of drug metabolism has been studied in vitro in human liver and intestinal microsomes, as well as in in vivo, using the disposition of substrate drugs to elucidate developmental patterns of the individual enzymes.

Although most DMEs show a unique developmental pattern, three distinct developmental patterns have been proposed, largely based on in vitro data from human livers (figure 1). Individual enzymes show activity either mainly prenatally, postnatally or throughout development.11 First, in the prenatal pattern, enzyme activity in fetal liver is high before birth and shortly thereafter; examples of these enzymes include CYP3A7, FMO1, SULT1A3/4, SULT1E1. Second, in the postnatal pattern, enzyme activity is low before birth, but increases to adult levels in the first weeks or months after birth; examples are: ADH1C, ADH1B, CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, FMO3 and SULT2A1. Finally, in the constant pattern, activity remains stable from early fetal life through adulthood; examples are: CYP3A5 and SULT1A1. These distinct developmental patterns likely reflect the physiological significance of these enzymes across the development continuum.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Schematic depiction of drug-metabolising enzyme ontogeny (3). CYP, cytochrome P450; TPMT, thiopurine S-methyltransferase; UGT, UDP-glucuronosyltransferases; SULT, sulfotransferases; FMO, flavin-containing mono-oxygenase.

Age-related changes in drug metabolism impact drug disposition and effect. Logically, the most profound effect is observed in the neonatal and infant periods, when typically the largest changes in DME activities occur. The grey-baby syndrome is one of the first observations of immature drug metabolism. In the late 50 s, neonates who received the antibiotic, chloramphenicol for sepsis, developed a ‘grey’ appearance with multiple-organ failure resulting in death due to toxic chloramphenicol levels.12 Only recently, chloramphenicol metabolism was attributed to UGT2B7, which is indeed immature at birth, as in vitro studies and pharmacokinetic studies on morphine, also a UGT2B7 substrate, have shown.13 ,14 Neonates <10 days of age, need about 25% of body-weight corrected morphine doses than infants of 1 year of age to reach similar plasma concentrations.14

The observation that toddlers require higher per-kilogram doses than adults for many drugs subject to drug metabolism suggests that DME activity is higher in toddlers than adults. This phenomenon has been attributed to the ratio of total liver mass to total body mass, which reaches a maximum between 2 years and 4 years of age.15 Nevertheless, strong evidence to support this hypothesis is still lacking. Moreover, in vitro studies have not consistently shown increased DME activity. Anderson and Holford make indeed a compelling argument, namely, that once DME capacity has matured, the increased clearance of many drugs in toddlers is related to size and reflects overall body metabolism.16 Although body surface area may be a surrogate for size and is frequently used to dose oncology drugs, it does not accurately reflect size. Instead, metabolic weight (kg^0.75) is recommended for extrapolating adult doses to children of different ages (and consequently different sizes). It should be kept in mind that this approach may be valid for older children, but not for neonates. In neonates, metabolic clearance of individual drugs matures at different rates, depending on the drug metabolism pathways involved.

An underappreciation of the higher clearance of metabolised drugs in children has led to underexposure in this population. Probably, the most striking example is the underdosing of HIV medication as a result of bodyweight-based extrapolation of adult doses to children. Forty percent of evafirenz levels were subtherapeutic in children who received the drug according to the initial WHO guidelines.17 The guidelines were revised not until 5 years later, after pharmacokinetic data had become available. Meanwhile, hundreds to thousands of children had received inadequate HIV therapy for a prolonged period of time.

Maturation of drug metabolism may impact disposition and effect of a drug by its direct effect on systemic exposure of the parent drug and the relative impact of active metabolites on drug effect may change as metabolic pathways mature. Case reports suggest that neonates are less vulnerable to paracetamol-induced liver toxicity than are older children and adults.18 Fewer toxic metabolites may be formed due to immaturity of the CYP enzymes involved (mainly CYP2E1).

Another example is that in young renal transplant patients, other sirolimus metabolites are formed than in adults.19 The clinical significance has not been studied yet, but if these metabolites in children are more toxic than the metabolites in adults, children may be at risk of unexpected adverse events. The anticancer drug, ifosfamide, is nephrotoxic, and younger children are at higher risk of developing nephrotoxicity, supposedly due to increased activity of renal CYP3A, resulting in high local formation of the nephrotoxic metabolite chloroacetaldehyde.20

Sophisticated software has enabled the use of population-based and physiology-based pharmacokinetic modelling to translate our knowledge on the maturation of DMEs into paediatric dosing guidelines. Using a model drug for a specific pathway, population pharmacokinetics and physiologically-based pharmacokinetic models have been developed describing the maturation of the studied pathway, for example, morphine for UGT2B7 and midazolam for CYP3A and have been used to develop age-appropriate dosing guidelines.21 ,22 Subsequently, these maturation models have been used to predict the age-related clearance of drugs metabolised by the same enzyme, such as zidovudine for UGT2B7 and alfentanil for CYP3A.23 ,24

Drug–drug interactions

Drug–drug interactions may also result in significant variation in drug metabolism. Several drugs are known to either induce or inhibit enzyme activity. Consequently, when such a drug is given with a drug metabolised by the same enzyme, the blood levels of the drug and metabolites may change accordingly. A few classes of drugs are notorious in this respect, such as azole fungals and macrolide antibiotics, well-known drug inhibitors. Well-known inducers are phenobarbital, rifampicin and dexamethasone (table 2).

Drug–drug interactions have been extensively studied in adults, often as part of registration requirements. By contrast, prospective research in children is lacking and most of our knowledge results from case reports. For example, in a recent systematic review on the safety of fluconazole in children, a fluconazole drug–drug interaction was identified in only five reported cases.25

Children's vulnerability to metabolic drug–drug interactions will alter as DMEs mature and disposition pathways change.26 If DME activity is developmentally low, the impact of a DME inhibitor will probably not significantly impact specific DME activity. In other words, a drug–drug interaction that occurs in adults may not be evident in children, especially very young children, if the DME of interest is not expressed, for two reasons. First, the DME may not contribute substantially to the disposition of the victim drug and second, there is little DME activity to inhibit. On the other hand, due to immature drug metabolism, the disposition pathway of a specific drug may change to preferential renal excretion, as is, for example, the case for caffeine in neonates. Hence, theoretically, a potentially nephrotoxic drug, for example, an aminoglycoside, may impact clearance of caffeine, while this would not be a likely drug–drug interaction in adults.

Disease and drug metabolism

Several disease states and treatment modalities may importantly impact drug metabolism and, consequently, drug absorption. An example is intestinal disease.27 Diarrhoea has been associated with toxic plasma levels of immunosuppressants in paediatric solid organ transplant recipients. The underlying mechanism appears to be reduced intestinal drug metabolising and transporter activity due to inflammatory downregulation or outright disappearance of functional villi.28 Consequently, a lesser amount of drug will be subject to first-pass metabolism—resulting in higher systemic exposure. This may seem counterintuitive, as one might expect a diseased intestine to work less well and absorb less drug—consequently resulting in lower plasma levels.

Critical illness may also contribute to variation in hepatic drug metabolism. Animal and adult studies have demonstrated that several cytokines, including interleukin 6 (IL-6), can downregulate the expression of several DMEs.29 IL6 levels were higher and CYP450 drug metabolism was much lower in paediatric intensive care patients with multiple organ failure than in those without organ failure.30 In line with these findings, clearance of the CYP3A substrate midazolam was significantly lower in paediatric intensive care patients than in relatively healthy children.31 The clinical significance of these observations still needs further study, as no relationship between midazolam clearance and sedation was found in these patients.29 Moreover, inflammation may also affect drug transporters and receptors and, consequently, drug effect.

Next, therapeutic interventions like hypothermia, cardiopulmonary bypass and extracorporeal membrane oxygenation (ECMO) may also impact drug metabolism.32 ,33 In both animal and human studies, hypothermia significantly reduced drug metabolism. With the recent introduction of hypothermia as standard of care in neonatal asphyxia and postcardiac arrest in older children, this observation is increasingly relevant for drug dosing of these critically ill patients. In a randomised trial designed to study the impact of hypothermia for traumatic brain injury, children in the hypothermia group showed a trend towards higher phenytoin levels and lower metabolic clearance during the rewarming phase than normothermic children.34

Extracorporeal membrane oxygenation elicits a systemic inflammatory response and release of cytokines and other inflammatory mediators that theoretically could affect drug metabolism. It is difficult to ascertain the relative contribution of ECMO treatment to changes in drug clearance, as other factors such as absorption to the ECMO system, ongoing disease, altered liver blood flow or iatrogenic infections may impact drug metabolism as the same time. For example, the increased CYP3A-mediated midazolam clearance after start of neonatal ECMO may be explained by maturation, but also by increased liver flow due to circulatory support (as midazolam is a medium-extraction drug).35

Pharmacogenetics

Genetic variation in DME activity has been associated with clinically relevant differences in drug disposition and/or efficacy and safety in children.

For example, CYP3A5 expression appeared to be related to clearance of tacrolimus given after paediatric renal or heart transplantation: CYP3A5-expressers needed higher doses from transplant onwards.36 ,37 Interestingly, age was also an independent predictor of tacrolimus disposition, hence the impact of age and CYP3A5 genotype was additive. In adult kidney recipients, CYP3A5 genotype-based dosing resulted in a shorter time to reach therapeutic tacrolimus concentrations early after transplant than did standard dosing.38 By contrast, clinical outcomes were not different, which can be partially explained by the start of tacrolimus at day 7 post-transplantation, as some of the outcomes, such as delayed graft function are diagnosed in the first week after transplantation. A similar study is now underway in children.39

Pharmacogenetic studies are increasingly focusing on the impact on drug outcomes. Very recently, in two studies, the TPMT genotype was associated with the risk to develop cisplatin-ototoxicity in paediatric oncology patients, while another could not replicate the results. These disparate results may be explained by a difference between the patient groups in exposure to craniospinal radiation and the otoprotectant amifostine.40 Also, patients with paediatric asthma using inhalation fluticasone and carrying the CYP3A4*22 allele showed improved asthma control compared with wild-type patients.41 Because CYP3A4*22 is associated with reduced CYP3A4 activity, its presence may increase systemic fluticasone exposure and indirectly lead to better asthma control.

Still, the uptake of genetic testing to improve patient care in children has been very slow so far. Only two drugs, atomoxetine and pimozide, have a US Food and Drug Administration genotype-based dose advice for children42 ,43 (table 3).

Gene-association studies in children are scarcer and usually smaller in size. Moreover, similar to underlying disease and co-medication, the interplay of age and genetics may prevent the extrapolation of adult pharmacogenetic data to children.44 For example, in children, age and body size are more important contributors to variation in warfarin disposition than are CYP2C9 and VKORC1 genetic variation.45 ,46 To reduce the risk of underdosing or adverse event, genotype-based dosing guidelines of warfarin in children should take the impact of age into account.47

Nevertheless, awareness that genetic variation may contribute to drug failure or toxicity may ultimately improve the care of individual patients. Psychiatry may be an area where pharmacogenetics may improve efficacy and safety in individual patients when other reasons for therapy failure or adverse events have been excluded. Especially in adolescents, drug metabolism capacity appears to be similar to adult levels, and dosing recommendations based on data derived from adults may be useful. For this reason, paediatricians are recommended to take into account age-related differences in pharmacogenetics when consulting information sources on pharmacogenetics. The most comprehensive source is (http://www.pharmgkb.org), which presents up-to-date pharmacogenetic data for individual drugs, including clinical practice guidelines for adult patients, when available.