Paracetamol and morphine for infant and neonatal pain; still a long way to go?

ABSTRACT

Introduction: Pharmacologic pain management in newborns and infants is often based on limited scientific data. To close the knowledge gap, drug-related research in this population is increasingly supported by the authorities, but remains very challenging. This review summarizes the challenges of analgesic studies in newborns and infants on morphine and paracetamol (acetaminophen).

Areas covered: Aspects such as the definition and multimodal character of pain are reflected to newborn infants. Specific problems addressed include defining pharmacodynamic endpoints, performing clinical trials in this population and assessing developmental changes in both pharmacokinetics and pharmacodynamics.

Expert commentary: Neonatal and infant pain management research faces two major challenges: lack of clear biomarkers and very heterogeneous pharmacokinetics and pharmacodynamics of analgesics. There is a clear call for integral research addressing the multimodality of pain in this population and further developing population pharmacokinetic models towards physiology-based models.

KEYWORDS:

• Pain
• pediatrics
• neonates
• infants
• analgesia
• pharmacokinetics
• pharmacodynamics
• opioids
• paracetamol
• acetaminophen

1. Introduction

Only few years ago, doses of commonly used drugs such as morphine and paracetamol (acetaminophen) in newborns and infants were based on extrapolated data from adults or older children. Sufficient knowledge about the efficacy, appropriate dosing, and safety of these drugs derived from properly designed studies has been lacking for years. However, dose validation and the beginning of formulation of evidence-based guidelines for these drugs have now been based on a number of clinical trials, population PK/PD studies, and research on pain assessment and long-term outcomes [1,2,3,4]. To understand the challenges in designing and conducting clinical investigations with analgesic medications in newborns and infants, it is important to learn from the previous successes and failures in this area. This review will describe and discuss the challenges of analgesic research in this population on different levels. We first discuss the pain definition and next address the issue of adequate pain assessment. With paracetamol and morphine as model drugs, we provide an overview of clinical trials so far, discuss the clinical pharmacology of these drugs as well as their short- and long-term effects. Special attention will be paid to premature neonates, as research in this population is even more challenging.

2. Pain

2.1. Definition of pain and its multimodality

The International Association for the Study of Pain (IASP) has provided the following definition of pain: ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage.’ They further state that ‘pain is always subjective. Each individual learns the application of the word through experiences related to injury in early life.’ However, as this is not applicable to neonates and infants [5], they revised this statement and added: ‘The inability to communicate verbally does not negate the possibility that an individual is experiencing pain and is in need of appropriate pain-relieving treatment.’

This definition aims to cover the multimodality of pain, as pain is more than nociception and/or perception of (possible) noxious stimuli alone. Loeser proposed a pain model which includes four components: nociception, pain, suffering, and pain behavior [6] (see Figure 1). Nociception refers to the detection of tissue damage by transducers connected to Aδ and C nerve fibers. The IASP defines this as ‘the neural process of encoding noxious stimuli.’ Pain occurs at the moment of perception, once the signal from the peripheral nerve system reaches the central nervous system (CNS). Pain can also occur without input from the peripheral nerve system, for example after CNS damage. Pain usually leads to suffering as described by Cassel [7]: a ‘consequence of a physical or psychological threat to the integrity of the human being,’ which has similarity with ‘actual or potential tissue damage.’ The fourth component is pain behavior, which includes the whole spectrum from small moans in acute short-term pain to frequent doctor visits in cases of chronic pain. This behavior can be interpreted by others as having pain. Loeser’s model makes clear that pain is a biopsychosocial phenomenon, not merely a biological one.

Figure 1. Depiction of pain multimodality according to Loeser. Adapted with permission from: Loeser JD. Pain and suffering. Clin J Pain 2000; 16(2 Suppl): p. S2-6.

2.2. Types of pain

In neonatal and infant critical care, four different types of pain need to be distinguished:

• Procedural pain, acute pain caused by a short-term procedure such as venipuncture, heel lancing, or chest drain insertion/removal.

• Postoperative pain, defined as the pain experienced in the first 24–48 h after surgery.

• ‘Prolonged’ pain, a term increasingly used for pain with a duration >72 h and specifically for the neonatal population [8,9].

• Chronic pain, defined as pain persisting beyond the expected tissue healing time. However, expected healing times are not clearly delineated. Thus, chronic pain was assumed to persist for time periods varying from 1 to 6 months [10], but in general practice a duration >3 months is being used.

These types of pain are important when implementing pain assessment tools both in research and clinical practice. It is important to know for what type of pain a tool is validated.

3. Pain assessment and PD markers

3.1. Behavioral assessment tools

Taking the definition of pain into account, it follows that pain can be only reliably assessed by self-report. In preverbal infants, this is not possible and then we have to rely on the interpretation of pain behavior.

A great variety of pain assessment tools has been developed over the past decades and to date, more than 40 different tools are available just for pain assessment in neonates and infants [11].

Table 1 lists the validated observational pain assessment tools for the use in preverbal children, including their indication and age category for which it has been validated. These are recommended by several international guidelines [12,13].

Table 1. Validated pain assessment tools according to age and their indication.

Pain assessment tool	Indication	Age category	Type of tool
Premature Infant Pain Profile (PIPP) [14–16]	Procedural and postoperative pain	Premature infants >24 weeks GA	Behavioral, contextual, physiological parameters
Revised Premature Infant Pain Profile (PIPP-Revised) [17,18]	Procedural pain	Premature infants >28 weeks GA	Behavioral and physiological parameters
Neonatal Pain, Agitation and Sedation Scale(N-PASS) [19]	Procedural and prolonged pain Sedation level	Premature infants >23 weeks GA	Behavioral and physiological parameters
COMFORTneo scale [20]	Prolonged pain Sedation level	Premature infants >24 weeks	Behavioral parameters
COMFORT scale [21,22]	Postoperative pain Sedation level	Children 0–18 years	Behavioral and physiological parameters
COMFORT behavior scale [23]	Postoperative pain Sedation level	Children 0–3 years Children 0–18 years	Behavioral parameters
Faces, Legs, Arms, Cry, Consolability (FLACC) scale [24,25]	Postoperative pain	Children 2 months–7 years	Behavioral parameters
Multidimensional Assessment of Pain Scale (MAPS) [26]	Postoperative pain	Infants 0–31 months	Behavioral and physiological parameters

GA: gestational age.

3.2. Physiology-based pharmacodynamic markers

As behavioral assessment tools are subjective to a certain degree, research efforts have been directed at identifying objective pharmacodynamic markers for the estimation of pain. Changes in vital signs, such as heart rate and blood pressure, do not serve as a link of pain neurobiology to pain behavior, as these autonomous responses to pain may be absent after a noxious stimulus [23]. Therefore, physiology-based markers such as near-infrared spectroscopy (NIRS), heart rate variability (HRV), skin conductance or pupillary reflex dilatation (PRD) have been studied but are not yet sufficiently validated. An overview is given in Table 2.

Table 2. Potential pharmacodynamic markers in neonates and infants.

Potential outcome measures	Advantages	Limitations
NIRS [27,28]	Noninvasive, continuously monitoring	Measures only cortical response
aEEG [29]	Continuous monitoring	Not validated for pain in children <2 years
Skin conductance [30–33]	Noninvasive, continuously monitoring	Sympathetic activity may be caused by anxiety and/or distress as well
SSEP [34,35]	Noninvasive, continuously monitoring (when repetitive stimuli are being given)	Only reflects sensory response, not validated in infants
Pupillary reflex dilatation [36,37]	Promising bedside application in older children	Sympathetic activity may be caused by anxiety and/or distress, practical difficulties in infants
HRV (including ANI) [38–50]	Noninvasive, continuously monitoring	Sympathetic activity may be caused by anxiety and/or distress
fMRI [51,52]	Good overlap between findings in infants and adults, insight in pain beyond the sensory cortex	Not suitable for clinical application
Salivary cortisol levels	Noninvasive collection of sample	Delay in laboratory results
Plasma cortisol levels [53–55]	Suitable for both short- and long-term pain	Delay in laboratory results, sampling restricted in neonates
Plasma adrenalin levels [53,56]		Delay in laboratory results, sampling restricted in neonates, less sensitive than noradrenalin
Plasma noradrenalin levels [53,56]	Significantly reduced by analgesics	Delay in laboratory results, sampling restricted in neonates

aEEG: amplitude-integrated electroencephalography; ANI: Analgesia Nociception Index; fMRI: functional magnetic resonance imaging; HRV: heart rate variability; NIRS: near-infrared spectroscopy; SSEP: somatosensory evoked potentials.

3.3. Limitations of current pain assessment methods

On a critical note, the currently available pain assessment tools have a number of limitations. First, they cannot satisfactorily distinguish pain from anxiety, stress, or other emotional states [57]. Second, application of a particular tool in different contexts and circumstances, such as severity of illness and diagnosis, can be problematic. For example, lethargy, stiff limbs, minimal movement, and grunting all predict severity of illness, [58] but may significantly affect the total score. Third, they may be subject to a certain degree of subjectivity from the observer, who may or may not know how the child usually reacts to pain and may interpret certain behavior such as less movement as reflecting being comfortable, when in fact the child holds still because of pain.

Physiology-based assessment methods such as NIRS, skin conductance, and HRV also have their limitations. Overall, they measure either the sympathetic nervous system activity or a cortical brain response to a stimulus. Sympathetic activity may reflect pain, but is also associated with stress, anxiety, and delirium. Also, vasopressor agents may influence sympathetic tone. A cortical brain response to a stimulus may be indicative of nociception, but does not tell us directly whether a stimulus is perceived as painful. For example, in a study using NIRS, [27] cortical brain responses were not altered by the administration of oral sucrose solution, whereas observational pain scores decreased significantly.

Practical aspects of performing such measurements also impede the use of physiological tools. As an example, pupillometry can only be done in subjects with eyes opened. Infants are unable to collaborate and measuring the pupil diameter could become another stressful event. Last, amplitude-integrated electroencephalography (aEEG) measurement has been used for research purposes but has not yet been shown a useful marker for pain and analgesia in newborns and young infants [59–62].

3.4. Item Response Theory

Pain assessment tools include several behavioral and/or physiological items. However, the items may not be indicative of pain to the same extent. Item Response Theory, an advanced statistical technique, could give insight in the informativeness of each separate item: the highest grade of intensity in one item may be more indicative of pain than the highest grade of another item. A recent study applied this technique to both the COMFORT scale and the Premature Infant Pain Profile (PIPP) in term and preterm neonates [63]. The behavioral items corresponded best with pain; physiological items did less. A similar pattern was previously reported, [21] and now advanced statistics show that high ratings of some behavioral items corresponded better with high pain levels than other items. This should be taken into account when using such a scale in new clinical trials, or when developing new observational assessment tools.

Despite numerous efforts to quantify pain, finding the optimal PD marker in infant pain studies remains a challenge [64]. We still have to rely on surrogate end points in neonatal and infant pain research. Beecher posed the problem of scientists’ and clinicians’ wishes to express subjective outcomes in objective measures research already 50 years ago and this problem has not yet been mitigated in infant and neonatal pain research [65,66].

4. Clinical trials

Clinical trials to evaluate dosing, efficacy, and safety of paracetamol and morphine in infants and newborns have been performed rather unconventionally compared to the scientifically desired drug development process known from newly introduced drugs. Both drugs are not new at all and were both given to newborns and young infants long before the first trials were performed. Several research groups started to evaluate the efficacy of analgesia during surgery [67] and ventilation [68] in the early 1990s. The first pharmacokinetic studies were also published [69]. Figure 2 presents a schematic overview of steps toward evidence-based pharmacotherapy (the desired way is illustrated by the black boxes) and illustrates that clinical research with morphine and paracetamol has followed the opposite way so far (as the white boxes indicate). The gray boxes indicate factors that influence the white and black boxes. These factors require further research to optimize analgesic research in neonates and infants.

Figure 2. Schematic diagram of clinical research wi th morphine and paracetamol (white boxes) in practice. Black boxes represent new drugs to be studied. Dose finding and population PK/PD modelling with both ion into clinical practice will be possible. RCT: randomized controlled trial; pop PK/PD: population pharmacokinetics/pharmacodynamics; PCM: paracetamol.

Before those trials, it was generally believed that neonates were not capable of experiencing pain, therefore neonatal surgery often was performed without any analgesia [70]. Anand et al. [53] showed significantly lower hormonal stress responses in an analgesia group of operated newborns and better neurological outcome compared to placebo-treated neonates.

These findings raised public and scientific interest in neonatal and infant pain research [71]. Consequently, several analgesic trials were performed since. These trials compared for instance the short-term outcomes of continuous and intermittent postoperative morphine in newborns and infants, [72] postoperative rectal paracetamol vs. morphine, [73] and routine morphine with placebo during endotracheal ventilation in preterm newborns [74]. See Tables 3a and 3b for the randomized controlled trials (RCTs) performed. Studies involving neonatal abstinence syndrome or without analgesic end point (for example endotracheal intubation facilitation) were excluded.

Table 3a. Randomized controlled trials with morphine in the young pediatric population.

Authors	Year	Sample size	RCT study design	Outcome	Major limitation(s)
Barker et al. [75]	1995	27	Low vs. high loading dose diamorphine	Adequate analgesia in both groups, high loading dose produced greater respiratory depression	Pain was not the primary end point and assessed by hormonal stress response
Wood et al. [76]	1998	88	Morphine vs. diamorphine in ventilated preterm neonates	Reduced stress response in both groups, more hypotension in morphine group	High morphine dose: 200 mcg/kg in 2 h and 25 mcg/kg.hr in preterm neonates Pain assessed by hormonal stress response, no placebo control group
Anand et al. [77]	1999	67	Morphine vs. midazolam vs. placebo in ventilated preterm neonates (NOPAIN pilot trial)	Morphine and midazolam reduced PIPP scores, only morphine improved neurological outcome	High midazolam loading dose and high morphine maintenance doses
Lynn et al. [78]	2000	83	Morphine continuous targeting predefined plasma level vs. intermittent bolus in infants postsurgery	Continuous morphine provided better analgesia, morphine dosing was higher	Power calculation performed on ventilatory end point
Van Dijk et al. [72]	2002	181	Morphine continuous vs. intermittent postsurgery aged 0–3 year	No difference in pain scores or morphine consumption	No power calculation performed
Simons et al. [74]	2003	150	Morphine continuous vs. placebo in ventilated preterm neonates	No beneficial effect on pain, less IVH in morphine group, comparable composite outcome	Rescue morphine not based on pain assessments
Anand et al. [79]	2004	898	Morphine continuous vs. placebo in ventilated preterm neonates (NEOPAIN trial)	No beneficial effect on neurological outcome measures	High morphine dose, no baseline for primary composite outcome
Carbajal et al. [80]	2005	42	Morphine boluses vs. placebo for procedural pain	No adequate analgesia in morphine group	Continuous morphine used for procedure
Taddio et al. [81]	2006	132	Morphine vs. placebo vs. tetracaine for central catheter insertion in premature neonates	Beneficial effect of both morphine and tetracaine	Morphine infusion allowed during intervention

PIPP: Preterm Infant Pain Profile.

Table 3b. Randomized controlled trials with paracetamol in the young pediatric population.

Authors	Year	Sample size	RCT study design	Outcome	Major limitation(s)
Howard et al. [82]	1994	44	Paracetamol vs. placebo in neonates undergoing circumcision	Minimal effect of paracetamol	Sweet solutions (both PCM and placebo) may have biased the results Crying time as pain end point
Shah et al. [83]	1998	75	Paracetamol vs. placebo 90 min before heel lance	No effect of paracetamol	No validated pain assessment tool
Van der Marel et al. [73]	2007	54	Morphine vs. rectal paracetamol in postsurgery infants	No effect of paracetamol	High continuous morphine doses could have masked a clean paracetamol effect
Van Lingen et al. [84]	2001	122	Rectal paracetamol vs. placebo after vacuum extraction	No analgesic effect of paracetamol	No validated pain assessment tool, rectal administration unreliable in neonates
Bonetto et al.	2008	76	Oral glucose vs. oral paracetamol vs. topical EMLA vs. placebo	No effect of paracetamol	No comment on interobserver variability of pain assessment tools
Badiee et al. [85]	2009	72	Oral paracetamol vs. placebo 90 min before heel lance	No effect of paracetamol	Time of administration could possibly be too early: no clear PK data on start of study
Manjunatha et al [86]	2009	18	Oral paracetamol vs. oral morphine vs. placebo 60 min before ROP screening	No significant effect of morphine or paracetamol	Underpowered (calculated sample size was n = 63)
Tinner et al [87]	2013	123	Rectal paracetamol vs. placebo after forceps or vacuum extraction delivery	No effect of paracetamol	Only 2 doses of paracetamol: no steady state
Seifi et al [88]	2013	120	Oral acetaminophen vs. oral sucrose vs. placebo	No effect of paracetamol	Paracetamol administered 30 min before ROP screening
Ceelie et al [3]	2013	71	Morphine vs. IV paracetamol in postsurgery infants	Significant reduction of morphine consumption	Single-center study

IV: intravenous; ROP: retinopathy of prematurity.

In the earlier trials, dosing regimens were based on scarce neonatal pharmacokinetic data [69,89]. Later trials used dosages based on population PK/PD models derived from these early trials [1,3]. Besides optimizing dosing regimens in clinical practice, this evidence-based dosing improved the quality of analgesic clinical trials. In the Ceelie et al. trial [3] for example, a fairer comparison between paracetamol and morphine could be made, as morphine plasma levels across all ages were the same. In the Van der Marel et al. trial [73], which found no beneficial effect of paracetamol, the analgesic effect of paracetamol could have been masked by relatively high morphine plasma levels as their study population was very young.

5. Pharmacokinetics/pharmacodynamics

One of the major challenges in neonatal and infant drug research is the rapidly changing pharmacology in this age group. Due to the rapid developmental changes in both the pharmacokinetics and the pharmacodynamics of a drug, a very heterogeneous population exists. These developmental changes will to a large extent determine the safety and efficacy of the studied drugs. Below, we will discuss the PK and PD of morphine and paracetamol as model drugs, to illustrate the importance of the changes throughout the first phase of life.

5.1. Morphine pharmacokinetics

5.1.1. Morphine metabolism

Drug metabolizing enzymes are classified by the reactions they catalyze: phase I reactions including oxidation, reduction, or hydrolysis and phase II reactions including glucuronidation, sulfation, methylation, or acetylation. Traditionally, phase I enzymes such as the cytochrome P450 system have received more attention in pharmacological research than phase II enzymes such as the uridine diphosphate glucuronosyltransferase (UGT) isoenzymes [90]. Morphine is glucuronidated by the UGT isoenzyme 2B7 into two active metabolites, morphine-3-glucuronide (M3G) and morphine-6-glucuronide (M6G) [91].

5.1.2. Maturation of morphine metabolism

Maturation of glucuronidation significantly influences the clearance of morphine in neonates, and most of the maturation process takes place in the first few months. A first indication came from the work of Lynn et al. [92] in a group of post cardiac surgery newborns and infants. Morphine clearance (expressed per kg bodyweight) was reduced in the first month of life but then increased to above adult levels. This pattern has been confirmed by two models derived from nonlinear mixed-effects modeling [93] and clinical research, respectively. Bouwmeester et al. [94] described that neonates in the first week of life required less morphine and had higher morphine plasma concentrations than thereafter, at the same mg/kg dosing regimen. These findings led to the development of a pharmacokinetic model for children up to the age of 3 years, including preterm neonates [2]. This model shows that the morphine glucuronidation capacity and the clearance of the glucuronides are influenced by bodyweight in a nonlinear manner (bodyweight-based power equation with an exponential scaling factor of 1.44). Furthermore, before the postnatal age of 10 days, clearance and glucuronidation capacity was approximately 50% lower than thereafter. The resulting new model-based dosing advice recommended significantly lower doses, particularly for the youngest neonates, than the generally recommended 10–40 mcg/kg/h [1]. It is thought therefore that neonates, especially aged <10 days, frequently may have been overdosed worldwide because of their low glucuronidation capacity. Figure 3 illustrates the plasma levels of morphine, M3G and M6G with the old vs. the new dosing regimen.

5.1.3. Model-based dosing

As this increase in clearance at day 10 after birth may be considered an arbitrary cut-off, Wang et al. further evaluated morphine pharmacokinetics and, using a wider population, developed a bodyweight-dependent exponent (BDE) model. The BDE model predicted clearance across the entire pediatric age range better than the model with a fixed exponent of either 0.75, the ‘classical’ allometric scaling exponent, or the age-dependent exponent of 1.44 [95]. While for neonates and infants below 1 year, the dosing schedule hardly differs between these two models, the next step in pediatric morphine research is evaluating this morphine dosing regimen based on the BDE model for other indications than postoperative pain after major noncardiac surgery [1]. An observational study, for example, showed that higher dosages for NEC are required, most likely because this is a very painful condition [96]. These studies are important because so far the model-based dosing guidelines [1,95] are only corrected for differences in PK and not for type of pain or severity of illness.

5.1.4. Factors contributing to variability

Despite this progress in optimizing exposure to morphine in infants by correcting for structural pharmacokinetic differences as a result of developmental changes, there is still a large random variability in morphine clearance,[97,98] with the highest variability in critically ill neonates. This may perhaps be attributed to variability in hepatic and renal function and hepatic blood flow, which in turn are influenced by positive pressure ventilation [99]. Other factors such as therapeutic hypothermia, [100] extracorporeal membrane oxygenation (ECMO) treatment [101] or type of surgery [92] may also be influential.

Furthermore, it cannot be excluded that genetic differences play a role here. As an example, single-nucleotide polymorphisms (SNPs) in the gene encoding for UGT2B7 have been shown to alter the pharmacokinetic parameters in adolescents [102]. The drug transporter P-glycoprotein, also known as MDR1 or ABCB1, may alter pharmacokinetics of morphine, [103] as well as organic cation transporter 1 (OCT1) and ABCC3 [104]. So far, however, the effects of pharmacogenetics on pharmacokinetics in children have been rarely studied. Large sample sizes are needed to demonstrate a significant contribution of certain SNPs, because some SNPs occur only in 1–5% of the population. Sufficiently large sample sizes can most likely be achieved only in (international) multicenter studies.

5.1.5. Plasma concentrations vs. CNS concentrations

While to date most developmental changes in plasma pharmacokinetics of morphine have been characterized, another source of variability could be the distribution into the CNS. As the blood–brain barrier (BBB) prevents a 1:1 concentration ratio of many substances between brain interstitial fluid and plasma, targeting certain plasma concentrations may not adequately reflect desired CNS concentrations. Therefore, insight in the transport of morphine and its active metabolites across the BBB will contribute to individualized dosing. So far, only a few pharmacokinetic studies have considered concentrations in the cerebrospinal fluid (CSF) in humans, and such studies in children are rare, and completely missing in preterm newborns.

One of the studies in adults showed an increase of the CSF:plasma morphine concentration ratio from 0.2 to 0.6 over a two-hour period [105]. Another study in patients with chronic use of oral morphine reported a ratio of 0.9, [106] suggesting that reaching a steady-state balance between both compartments takes some time. In children, only one study so far has linked serum and CSF concentrations of morphine [107] after a single infusion. The plasma:CSF concentration ratio was nearly 1:1 after 2 h. For M6G, this ratio remained about 10:1, as this metabolite is less lipophilic. This finding could be relevant, as it is being debated whether morphine itself or M6G is the most important pain relieving substance [108,109]. M3G, which circulates in substantially higher concentrations than morphine and M6G, is thought to lead to adverse effects such as hyperalgesia, particularly upon prolonged use when this metabolite accumulates [110,111]. Studies have shown that M3G accumulates in critically ill patients, even after a 33% dose reduction [111,112]. Morphine plasma levels decreased, but M3G levels remained the same after this dose reduction, indicating that M3G could be highly responsible for side effects in this specific population. This is an important finding and the role of M3G in morphine safety should be studied further.

Even though attempts have been made to describe the CNS pharmacokinetics of morphine, the full picture is not yet clear. The BBB is changing throughout childhood, and the fact that the P-glycoprotein drug transporter is less readily available in the neonatal brain could mean an increase in diffusion of morphine into the brain [103,113]. Unfortunately, no human data are available to confirm this. Animal models using microdialysis have shown a higher morphine influx in premature sheep than in adult sheep [114]. Also, morphine efflux out of the brain is reduced in premature rats compared to adult rats [115]. In pigs, the plasma:CSF transfer ratio decreased from 0.7 to 0.5 during the first 6 weeks of life, which was not statistically significant [116]. Nevertheless, data of animal models cannot easily be extrapolated to humans, notably in view of the considerable differences in BBBs across species [117]. Moreover, it is very likely that the BBB and related morphine diffusion have undergone changes in CNS disorders such as meningitis and encephalitis.

5.2. Morphine pharmacodynamics

5.2.1. PK–PD relationship

The pharmacokinetic research described above centralized the plasma concentrations of morphine, but unfortunately the pharmacodynamic effects of morphine greatly vary between infants, even at similar plasma concentrations. It is known that analgesic needs in general depend on the type, severity and duration of pain. In relation to postoperative pain, morphine requirements depend on duration and severity of surgery but also on the type of surgery, such as cardiac or abdominal surgery [3]. Newborns operated for NEC need much higher morphine dosages postoperatively than newborns operated on for other conditions [96].

Efforts to establish a minimal effective plasma concentration or a therapeutic window have not led to a clear target [118] and a concentration–response curve is lacking [119]. Future research should aim for specific plasma targets for different types of pain or procedures.

5.2.2. Development of morphine sensitivity

Postmenstrual age may play a role in morphine sensitivity, due to maturation of nociceptive pathways. Morphine exerts its effects mainly on the mu-opioid receptor and to a lesser extent on the kappa and delta opioid receptors [120]. Sensitivity to morphine seems to be higher at neonatal age, although this has been suggested merely in rat models [121]. However, human neonates [3] below the age of 10 days needed significantly less rescue morphine than older neonates despite similar morphine plasma levels [1]. This could be due to less capability of pain expression, or to a higher BBB permeability for morphine. Another explanation may lie in the postnatal reorganization of opioid receptor expression. In rats, [121] during the first 3 weeks of rat life, the mu-opioid receptor expression is downregulated in the A fibers, but remains unchanged in the C fibers. Also, in this period, the central terminals of the A fibers are found in the superficial dorsal horn, whereas at adult age only C fibers project into the superficial dorsal horn and the A fibers project in the deeper lamina, suggesting a higher morphine sensitivity in the early weeks of life.

5.2.3. Pharmacogenetics

Pharmacogenetics may also play a role in morphine sensitivity. In human neonates, a combination of SNPs in two different genes, OPRM1 and COMT, was found associated with postoperative morphine consumption [122]. The same polymorphisms are associated with the severity of neonatal abstinence syndrome [123]. While OPRM1 is the coding gene for the mu-opioid receptor, COMT is a regulatory gene of mu-opioid receptor expression [124]. If both expression and function of the mu-opioid receptor are being disrupted by gene mutations, this could diminish the response to opioids.

5.3. Paracetamol pharmacokinetics

Paracetamol is different from morphine with respect to safety and efficacy aspects. Paracetamol is a weaker analgesic, but on the other hand has a more favorable side effect profile. Still, the only side effect to take into account is probably more lethal than the opioid-induced respiratory depression, namely acute liver failure. In Western countries, paracetamol overdose is the most common cause of acute liver failure in adults and children [125,126]. However, when kept within the therapeutic range, paracetamol provides analgesia as well as antipyrexia.

5.3.1. Paracetamol metabolism

Paracetamol is mainly metabolized by phase II enzymes to paracetamol-glucuronide and paracetamol-sulfate. Only a small fraction (1–4%) is excreted unchanged by the kidneys. The remainder is being metabolized to the hepatotoxic metabolite NAPQI through the action of cytochrome P450 (CYP) enzymes such as CYP2E1, CYP1A2, CYP3A4, and CYP2A6 [127]. Under normal circumstances, reduced glutathione (GSH) neutralizes NAPQI very rapidly and the inactive cysteine and mercaptopuric metabolites are being formed and excreted renally. See Figure 4 for the metabolic pathways of paracetamol.

Figure 3. (a, d) Morphine concentrations, (b, e) morphine-3-glucuronide (M3G) concentrations, and (c,f) morphine-6-glucuronide (M6G) concentrations predicted in model-based simulations in children weighing 0.5, 1, 2, 2.5 and 4 kg and a postnatal age of <10 days (dotted lines) and children weighing 0.5, 1, 2, 2.5,4, 10 and 17 kg and a postnatal age of >10 days (solid lines) based on (a–c) a dosing regimen with a loading dose of 100 mg/kg and maintenance dose of 10 mg/kg/h and (d–f) a regimen with a loading dose of 100 mg/kg followed by an infusion of 10 mg/kg1.5/h with a 50% reduction in the maintenance dose for children with a postnatal age <10 days. Reprinted with permission from: Knibbe, C.A, et al., Morphine glucuronidation in preterm neonates, infants and children younger than 3 years. Clin Pharmacokinet, 2009. 48(6): p. 371–85.

Figure 4. Metabolic pathways of paracetamol. Reprinted with permission from: Roofthooft, D.M.E., Paracetamol and Preterm Infants: a painless liaison? PhD thesis, 2015, Erasmus University.

5.3.2. Maturation of paracetamol metabolism

Glucuronidation of paracetamol is very low in preterm infants and matures during early childhood, thus simultaneously increasing the relative contribution to paracetamol elimination [128] whereas the sulfation route remains fairly constant [129]. CYP enzymes mature during early infancy as well, but data on NAPQI formation in neonates are lacking. One modeling study in 47 patients could not attribute any clearance to oxidative pathways [128] However, as renal clearance is lower in neonates compared to adults, renal metabolite clearance may be reduced [130]. Expression of CYP2E1, the main isoenzyme responsible for NAPQI formation, increases during the first three postnatal months [131]. Whether NAPQI formation is reduced in this period remains unclear, but clinically there are no clues of NAPQI formation in neonates leading to hepatotoxicity, even at higher doses [132].

The pharmacokinetics of paracetamol in children for different administration routes is well described.[133–138] Population PK studies showed that weight is the most important predictor of paracetamol clearance in neonates [130,139]. In a BDE model, this relationship was found to be nonlinear between children within the neonatal range and adults [4].

5.3.3. Factors contributing to variability

There is, however, a great interpatient variability in paracetamol clearance [129]. For instance, clearance is lower in preterm neonates than in term neonates [140]. Paracetamol is being metabolized by many enzymes, thus the current role of pharmacogenetics is small. SNPs and mutations in metabolizing enzymes have been described, but often have not been studied in relationship to paracetamol [141]. Whether genetic variability influences the PK of paracetamol is hard to tell. If one or two metabolizing enzymes lose function due to genetic polymorphisms, other enzymes may take over. Differences in activity of transporters, epigenetic phenomena, and organ-specific activity of metabolizing enzymes probably all play a role. System biology-based modeling strategies may contribute to insight in the complex interactions between ontogeny, metabolic functions, and genetics [142].

5.4. Paracetamol pharmacodynamics

5.4.1. Unknown site of action

While the pharmacokinetic basis for an evidence-based paracetamol dosing regimen has been established,[133–138] much groundwork remains to be done on the pharmacodynamics. For one thing, the exact mechanism of action remains unclear, even after comprehensive research [127, 143]. Its primary target has been suggested to lie within the CNS, as CSF concentrations corresponded better with analgesic response than did plasma concentrations [144]. In contrast, another study showed that analgesic response occurred earlier than changes in CSF plasma level [145]. A delay of the onset of action of the drug, a phenomenon called hysteresis, could perhaps explain why it is difficult to relate actual plasma and/or CSF levels to a pharmacodynamics end point such as temperature or pain score. Therefore, Gibb and Anderson recommend to use indirect-response models to describe paracetamol PK and PD [146].

5.4.2. PK–PD relationship

As the mechanism of action is unclear, factors influencing the pharmacodynamic effects of paracetamol in children are hard to establish. The administration route, though, certainly has an impact on its effectiveness [3,73,137,147] because of more favorable pharmacokinetics of the oral and intravenous routes compared to the rectal route.

A dose-dependent response in children has been suggested when paracetamol was administered as treatment for both pain and fever [148,149]. In general pediatrics, an adequate analgesic target plasma level of 10 mg/L is suggested [147]. It is unclear whether this target fits neonates and infants. Validation of these targets using validated assessment scales is necessary. Moreover, the question remains why there is a great variability in response. Whether this can be attributed fully to the action of paracetamol or to other factors such as the interpretation of the pain assessment tool needs to be studied further, especially as most trials studying the morphine-sparing effect of acetaminophen have been performed in older children [150–152].

Pharmacogenetics also does not help explain variability in response. This is mainly due to the lack of knowledge on the mechanism of action, impairing the search for relevant genes.

6. Short-term and long-term side effects

6.1. Morphine

6.1.1. Short-term effects

Morphine has several side effects, observed not only in adults but even more so in children and in neonates, such as respiratory depression, nausea, vomiting, constipation, urinary retention, and hypotension. Hypotension in neonates, who are more vulnerable to changes in blood pressure, [79] may have severe consequences such as intraventricular hemorrhage (IVH) or periventricular leukomalacia (PVL). On the other hand, pain leads to an increased blood pressure which could also contribute to these sequelae. In a large RCT, the incidences of PVL and IVH were significantly higher in the placebo group compared to the morphine group [79]. In another RCT using lower morphine doses, IVH was also less frequent [74]. These data implicate that adequate analgesia may protect the brain on the short term. Both clinical trials concluded, however, that routine administration of morphine in ventilated newborns should not be recommended.

An important challenge is detecting side effects of analgesics in neonates and infants, as these can overlap with symptoms of their underlying disease. Structural proactive screening for side effects should take place in analgesic trials in the neonatal and infant population.

6.1.2. Long-term effects

Drugs administered when the child’s brain is still developing possibly interfere with neurodevelopment [153–155]. This may especially be applicable to opioids, which directly act on the CNS. Moreover, in premature neonates, this interference may be even larger as the third trimester of gestation is important in CNS maturation. The current literature is not consistent regarding the long-term outcome of neonates receiving opioids. Some studies suggest alterations in neurological anatomy or structure, [153,154,156] but neuropsychological outcomes seem to be not affected by morphine exposure at neonatal age [154, 157,158]. Follow-up trials cannot include a control group, as adequate analgesia is ethically obligatory and keeping neonates without opioids is highly undesirable. In the vulnerable preterms admitted to NICU who still receive a large amount of morphine, [159] morphine should not be considered the only causative factor for abnormal neurodevelopmental outcome [156,160].

Schuurmans et al. suggest a more standardized approach with large patient samples to detect small outcome differences [161]. Data about analgesics and long-term effects from such large studies are not yet available. Standardization of long-term follow-up should include clinically relevant end points, preferably neuropsychological outcomes, especially executive function skills, as they determine a person’s functioning in daily life [158,162].

6.2. Paracetamol

6.2.1. Short-term effects

Paracetamol is regarded very safe on the short term. The major short-term side effect is acute liver failure, but no major hepatotoxic events have been reported so far in neonates [132].

6.2.2. Long-term effects

Recent studies on atopy development associated with paracetamol exposure in early childhood are reason for concern [163–165] although they have important methodological limitations [166,167]. Prospective studies should be performed to establish this potential causal link [168]. Such trials should at least incorporate pharmacogenetic assessment, [169,170] although fundamental research on the mechanism(s) of action should pave the way for targeted pharmacogenetic research.

6.3. Long-term effects of pain

An important other challenge regarding the long-term effects of analgesics, is the long-term effect of pain itself. Pain exposure during early infancy also leads to alterations in brain structure [171,172] and affects pain sensitivity in later life [173]. Finding an optimal balance between accepting long-term opioid effects, on which the literature is still divided, [154,174,175] and the long-term effects of pain itself, is not yet feasible. Finding this balance could be supported by either more optimal dosing strategies, such as always starting with the lowest dose possible and up-titrate on the effect, taking into account interindividual variability, or introducing alternative analgesics. This should be the aim of neonatal pain research in the near future, including larger sample size studies.

7. Premature neonates

Special attention should be paid to (extreme) premature neonates when it comes to both pain and developmental clinical pharmacology. In the vulnerable premature neonate, pain is not only ‘an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage,’ but is harmful on both the short and long term [171–173] These children daily undergo many potentially painful procedures, [176–178] while it is still unclear how to prevent or treat this pain optimally. What is more, the developing brain of the preterm infant is probably extremely vulnerable for the toxic effects of pharmacological agents, such as morphine [153].

Evidence from animal and human studies suggests an inverse relationship between postconceptional age and pain sensitivity, as measured by the reflex withdrawal test using von Frey hairs. However, whether this also means an increased pain sensation is doubtful as cortical pain responses seem to increase with increasing gestational age [179]. Nevertheless, little is known about the premature cortical pain processing and consequently the statement that (extreme) premature neonates feel more pain than their term peers is hardly based on solid methodological evaluation. At present, the techniques to investigate these phenomena are far from ideal. Therefore, investigators developed pain assessment tools based on the heel lance as a standardized painful procedure. Heel lancing is associated with an acute pain response that differs from more prolonged continuous distress and pain responses that are associated with preterm neonatal care.

The pain experienced by neonates should be treated with adequate analgesics and preferably be prevented using the concept of preemptive analgesia. Morphine is recommended in several guidelines for the treatment of severe pain in premature neonates [12,180]. The question arises whether the clearance of morphine differs between premature neonates and term neonates and if recommended doses can be used in extreme premature neonates. A recent study in extreme low birth weight neonates [181] compared five population PK models applied to their own prospective dataset of PK samples. The model based on data from extreme low birth weight infants [182] fitted best with their own dataset. The authors concluded that not only bodyweight but also maturation (including hepatic and renal function) contributes significantly to clearance, independently of bodyweight.

For paracetamol, clearance changes nonlinearly with bodyweight [4]. In premature neonates, the paracetamol clearance matures more slowly than morphine clearance, and this could be attributed to the complex metabolism of paracetamol [139]. However, robust PK data in extreme preterm neonates are scarce [183] and in this population, more research is warranted on PK and its relation with both the analgesic effects and short- and long-term neurodevelopmental outcomes.

8. Expert commentary

Neonatal and infant pain management and research have taken an extensive scientific journey so far. However, there is still a long way to go. The complex multimodality of pain comes along with major challenges in research. The first challenge is the assessment of pain. While current pain assessment tools merely reflect the outer circle of Loeser’s pain model, i.e. pain behavior, opioids and paracetamol act mainly on the nociception circle or the pain circle. This means that end points in pain research in children can only be surrogate end points. To measure analgesic effectivity, we should aim at biomarkers reflecting the direct effect of analgesics in the CNS. Attempts have been made with pupillometry, skin conductance, and HRV. These physiological, objective markers reflect sympathetic nervous system activity and are closer to the CNS. NIRS, and somatosensory evoked potentials reflect the perception of a stimulus by the CNS, and are therefore promising in pain research. Multimodal studies, such as performed by Slater et al. [27] and Hartley et al. [184] demonstrate lack of correlation between nociceptive brain activity and behavior at least in an experimental procedural pain study, which is a clear call for more clinical research.

The other major challenge is optimal pharmacological pain treatment. Healthcare professionals need to take two important steps: choosing the right drug and defining the right dose. In order to choose the right drug, it is necessary to define type of pain first. From the clinical trials performed so far in neonates and infants, it seems that morphine is not the best choice for procedural and chronic pain, but has proven effectiveness for postoperative pain. Many clinics use fentanyl as an alternative to morphine and fentanyl or alternatively one of the synthetic derivatives is proven effective for procedural pain. Paracetamol (Table 3b) seems to have no effect on procedural pain and a slight effect on postoperative pain. It is unknown whether it is effective for chronic pain. However, again it should be taken into account that these trials all have used surrogate end points and there is still a long way to go to determine true effectivity for each type of pain.

The choice of drugs is also heavily influenced by the safety profile, and both morphine and paracetamol have been ascribed long-term negative effects without convincing evidence [156,160]. These speculations call for well-designed trials with long-term outcomes as primary end points. Studies have been performed in different patient groups but with relatively small sample sizes and therefore underpowered to detect small differences in neurological outcome [154,185,186].

Dosing is also a major challenge due to both the rapid changing pharmacokinetics and pharmacodynamics in neonates and infants. Many population-based pharmacokinetic models have been developed for this population and have provided important data on the maturation of drug clearance. Still, there is no clear PK–PD relationship for morphine. PK is mainly focused on plasma levels, but PD effects are the result of morphine levels in the CNS. Further insight in the plasma–CNS relationship is necessary to define target plasma levels. The same holds for paracetamol, for which the PK is well described for term neonates and infants but a PK–PD relationship also has great variability.

9. Five-year view

Pain assessment could possibly be improved by techniques reflecting CNS activity such as NIRS [27,28] or aEEG, [29] although the analyses and interpretations of these techniques are yet far from optimal to recommend their clinical use as pain measurement instruments. Functional MRI seems a promising method to look for specific brain areas involved in pain processing and the role of analgesics on these pain areas. Still, it should be kept in mind that activity in these areas does not always reflect pain, as these areas are also responding to stimuli in patients not capable of experiencing pain [187]. This technique should be further developed and applied in order to compare the ‘true’ effect of different analgesics, and comparative studies with both behavioral assessment tools, fMRI and/or NIRS/SSEP/aEEG may provide further insight in how to optimize pain assessment. Meanwhile, behavioral pain assessment tools are not to be forgotten. Application of the Item Response Theory [63] may identify the most pain-specific items of pain assessment scales and further improve bedside pain assessment.

Furthermore, interactions between analgesics, for example paracetamol and opioids, deserve attention. A study already found that the use of paracetamol could reduce infants’ opioid consumption by 66% after major noncardiac surgery [3]. In our center, a similar trial is ongoing in cardiac surgery patients (the PACS trial; Dutch Trial Registry ID NTR5448). Trials like these could be a great opportunity to address the issue of long-term effects of both analgesics.

The increasing knowledge of the ontogeny of drug metabolizing enzymes and drug transporters provides a good basis for a system-based approach of pediatric pharmacokinetics. This approach enhances the development of both individualized evidence-based pharmacotherapy of currently existing drugs and new analgesic drugs for children. Knowledge gained in the pharmacokinetic modeling of one drug could be applicable to other drugs as well. When specific properties such as logP and pKa of other drugs are applied to such models, clearance of these drugs can be predicted. Building such physiologically based pharmacokinetic models may save the tremendous effort of describing PK of all drugs separately [188].

Other analgesics are also being introduced to neonatal and infant pain management. This review focused on morphine and paracetamol as model drugs, but fentanyl is also often used in the NICU [12,189,190]. Current use of fentanyl is based on little evidence, and its PK is highly variable in preterm neonates [191]. Studies performed with fentanyl are small [192–194]. This is a problem in most clinical trials involving neonates, especially in the NICU population. Therefore, to improve clinical research in this population, multicenter studies, if possible in established international consortia, could increase sample sizes. If this is not feasible, the required sample size may be reduced with the use of comparative effectiveness studies rather than superiority trials [195].

Another opioid which seems promising, especially for application in procedural pain, is remifentanil [196–198]. This very-short-acting opioid is being metabolized by plasma esterases, independent of organ function or age. However, it does not automatically ‘do away with’ the dosing problem as its side effects such as chest wall rigidity can be age dependent [197]. Caution is required with the clinical application of remifentanil.

Last, the focus should be set on long-term effects. We do not know yet which analgesic is most harmful in the long term, but must not forget pain is harmful anyway. Finding the optimal balance remains challenging [199] and calls for standardized long-term follow-up in neonatal pain trials.

Key issues

• Pharmacotherapy in neonates and infants is often not evidence-based. Finding the optimal design for clinical studies remains challenging.

• Current pediatric drug research starts with clinical practice and ends up in more fundamental research. The other way round would be more conducive to evidence-based pharmacotherapy.

• Pain is a biopsychosocial and subjective phenomenon and is therefore hard to assess.

• The currently available pain assessment tools, both behavioral and physiological, for infants and neonates do not adequately differentiate between nociception and/or pain sensation on the one hand and stress or distress on the other hand.

• Pharmacokinetics of morphine and paracetamol have been well described, even though there is still a great random variability in pharmacokinetics. The pharmacodynamics of morphine and paracetamol deserve further study with inclusion of ontogenic, pharmacological and genetic aspects.

• Most literature focuses on plasma levels of analgesics, but insight in central nervous system pharmacokinetics is lacking.

• Physiology-based pharmacokinetic models may serve as predictors of clearance which can be applied to multiple drugs.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Acknowledgments

The authors thank Ko Hagoort, Department of Pediatric Surgery, Eramus MC- Sophia Children’s Hospital, Rotterdam, The Netherlands for editorial assistance.

Additional information

Funding

This paper was not funded.