Inflammation and cardiovascular status impact midazolam pharmacokinetics in critically ill children: An observational, prospective, controlled study

Abstract Altered physiology caused by critical illness may change midazolam pharmacokinetics and thereby result in adverse reactions and outcomes in this vulnerable patient population. This study set out to determine which critical illness‐related factors impact midazolam pharmacokinetics in children using population modeling. This was an observational, prospective, controlled study of children receiving IV midazolam as part of routine care. Children recruited into the study were either critically‐ill receiving continuous infusions of midazolam or otherwise well, admitted for elective day‐case surgery (control) who received a single IV bolus dose of midazolam. The primary outcome was to determine the population pharmacokinetics and identify covariates that influence midazolam disposition during critical illness. Thirty‐five patients were recruited into the critically ill arm of the study, and 54 children into the control arm. Blood samples for assessing midazolam and 1‐OH‐midazolam concentrations were collected opportunistically (critically ill arm) and in pre‐set time windows (control arm). Pharmacokinetic modeling demonstrated a significant change in midazolam clearance with acute inflammation (measured using C‐Reactive Protein), cardio‐vascular status, and weight. Simulations predict that elevated C‐Reactive Protein and compromised cardiovascular function in critically ill children result in midazolam concentrations up to 10‐fold higher than in healthy children. The extremely high concentrations of midazolam observed in some critically‐ill children indicate that the current therapeutic dosing regimen for midazolam can lead to over‐dosing. Clinicians should be aware of this risk and intensify monitoring for oversedation in such patients.


| INTRODUC TI ON
Midazolam is a benzodiazepine with sedative, amnesic, and antiepileptic properties. An intravenous bolus dose of the drug alleviates symptoms in a matter of minutes and, when administered prior to a short-lived unpleasant procedure, prevents significant patient distress. 1 Continuous IV infusion midazolam is commonly used in pediatric critical care to provide sustained patient sedation. 2 While therapeutically very useful, prolonged midazolam administration often results in drug tolerance and severe adverse reactions including respiratory depression and long-lived neuro-psychiatric disturbances on drug withdrawal. [3][4][5][6][7][8] The frequency of adverse reactions reports suggests that midazolam dosing is not optimal for this population. [9][10][11] Presently, IV midazolam doses recommended for children are based on weight-based scaling of doses used in adults. 12 Conceivably, therefore, personalizing pediatric IV midazolam dosing recommendations, e.g., according to patient genetics or physiology, could optimize its effectiveness and lead to improved long-term outcomes.
Midazolam acts by potentiating the effects of gammaaminobutyric acid (GABA) on GABA A receptors. 13 It is metabolized in the liver by cytochrome P450 3A4 (CYP3A4) enzyme to an active metabolite (1-hydroxy midazolam). Glucuronidation of 1-hydroxy midazolam by UDP-glucuronosyltransferases (UGTs) generates inactive metabolites that are excreted in the urine. Polymorphisms in CYP3A4 and UGT genes may account for differences in pharmacokinetics in healthy and critically-ill individuals. 14 Physiological changes caused by critical illness and treatment interventions could further alter the midazolam exposure-response relationship. 15 These include changes in GABA A signal transduction, diminished hepatic blood flow, and altered liver enzyme metabolizing capacity. The latter two factors are potential modifiers of midazolam clearance in the critically ill. 16,17 This prospective, observational, pharmacokinetic (PK) study explored midazolam disposition in otherwise healthy children undergoing elective surgery (control group) and critically ill patients requiring mechanical ventilation in intensive care. Contemporaneous recruitment of a control group should enable critical illness-related covariates influencing PK parameters to be identified with greater certainty.

| Study design and study population
This was a single-center, observational, prospective study of IV midazolam pharmacokinetics in two groups of children; clinically well children (control group) receiving IV midazolam prior to elective day case surgery and ill, intubated, and ventilated children receiving IV midazolam in intensive care (critically ill group). Children were aged between 1 month (corrected gestational age) and less than 16 years and admitted either for planned surgical procedures requiring general anesthesia or to the pediatric intensive care unit (PICU). Although there was no sample size calculation, a minimum of 50 children were to be recruited with at least 150 PK samples to enable a robust population PK model to be developed. In the control group, midazolam was administered as a single IV 'bolus' dose (25-50 mcg kg −1 ) as part of general anesthetic induction.
In the critically ill group, infants and children received an initial 25-50 mcg kg −1 bolus dose and were then initiated on a continuous infusion of 50-200 mcg kg −1 h −1 . The target sedation level was assessed and reviewed at least twice daily. The continuous infusion rate was altered according to the unit's algorithm for dose adjustment and titrated to the desired sedation score. Similarly, additional bolus doses were occasionally administered as necessary to achieve and maintain the desired sedation level.

| Assessments and endpoints
In the control group, initial PK samples (1 and 2) were taken at various time points, starting from the pre-dose sample before surgery, during surgery, and after surgery was complete, and surgical drapes covering the child were removed. Subsequent samples (3, 4, and 5) were taken either in the recovery suite or on the wards. PK blood samples were thus obtained for up to 6 h post-dose.
In the critically ill group, blood samples for PK were either scavenged (from laboratory samples obtained for monitoring patients) or opportunistic (at times when blood samples were being taken for clinical reasons). PK blood samples were obtained for the duration of the period the child was on midazolam infusion and up to 96 h after treatment had stopped. CRP, liver, and kidney function assessments were determined at least once daily. To measure cardio-vascular status, a scoring system was developed for this study utilizing data available in the PICU cohort (Table 1). Each variable was scored between 1 and 3 and added together to generate a final score ('CV score'). Increasing CV score implies worsening cardiovascular function. A secondary objective of the study was to validate the volumetric absorptive microsampling system of PK sampling as an alternative to the collection of conventional blood samples in tubes. This volumetric system involved using Mitra® microsampling devices based on VAMS® technology to collect blood for analysis as a dried sample. In both groups of patients, each PK sample was collected as a duplicate: a whole blood wet sample (for centrifugation and processing) and a 10 μl dry sample using the Mitra device with a VAMS system. The validation data for the VAMS system met internationally accepted guideline criteria. 18

| Patients and data
Thirty-five patients contributed PK observations from the control group and 54 patients contributed from the critically ill group.
Baseline covariates for both groups of patients are listed in Table 2.
The median (range) age was 22 months (1 month to 15 years). In the control group, 80% of children were over 2 years of age, whereas in the critically ill group the age was skewed toward the younger age group with 67% less than 2 years of age. The body weight ranged from 2.9 to 78.4 kg (median, 13.70 kg) and the BMI ranged from 9.8 to 21.8 kg m −2 (median 15.8 kg m −2 ) (  CRP was assumed to be normal (≤3 mg L −1 ) in the control group. The mean (range) CV score was 7 (4-12) in the critically ill group. Control group patients did not have cardiovascular dysfunction or any requirement for inotropic support and therefore a fixed score of 4 was imputed during model development.

| PK model development
A two-compartmental structural model for midazolam, with a single additional compartment for 1-hydroxy midazolam, described the observed PK data satisfactorily. Due to structural non-identifiability of the volume of distribution of the 1-hydroxymidazolam compartment, it was set equal to that of the midazolam central compartment.   Table 3. GOF plots (Appendix S2, Figures S10 to S15) and VPC (Appendix S2, Figures S16 and S17) from the final model were satisfactory. A full description of the model results can be found in Appendix S2.

| Simulations
The

| DISCUSS ION
It is well established that altered physiology caused by critical illness alters the PK and pharmacodynamics of many drugs as a result of disrupted one or more 'ADME' processes. 23,24 In this prospective population PK study, we show that markers of systemic inflammation (CRP) and cardiovascular function (CV score) are associated with reduced midazolam clearance in critically ill children. Similar findings have been reported previously but unlike previous studies, the inclusion of healthy children in our study increases the certainty of the findings. 16,17 C-reactive protein (CRP) is an acute phase reactant that increases during periods of inflammation, e.g., due to sepsis or trauma. Its concentration in blood correlates with concentrations of interleukin-6 (IL-6) a pro-inflammatory cytokine that is known to induce CRP. The PK model developed herein estimates that critically ill children with a CRP > 100 ng ml −1 have almost a 50% reduction in midazolam clearance (Cl mid ) and those with a CRP > 200 ng ml −1 Cl mid reduces around 70% compared to otherwise healthy children with a CRP <3 ng ml −1 .
Reduced urine output, treatment to support blood pressure (inotropic drugs such as dopamine and boluses of fluid), and increased acid (negative base excess) production are all features of depressed cardiovascular function. In this study, these variables were used to develop a score to provide an integrated measure (CV score, Table 1) of cardiovascular dysfunction. Children with a CV score of 12 are estimated to have an almost 75% reduction in midazolam clearance compared to those with a normal (=4) CV score. Model simulations predict that a combination of a CRP > 200 ng ml −1 and CV score = 12 will result in median midazolam concentrations between 1000 and 3000 ng ml −1 (higher in children than infants) when administered continuous IV midazolam infusion at recommended dosing rates.
Although no pharmacodynamics assessments were included in our score). In a subsequent larger cohort study of 83 critically ill children aged 1 day to 7 years, a CRP of 300 mg L −1 was associated with a 64.5% lower clearance than a CRP of 10 mg L −1 and three failing organs were associated with a 35% lower clearance compared with one failing organ. 17 The investigators interpreted the effect of failing organs as largely the result of altered hepatic blood flow.
In healthy individuals, midazolam has a low/intermediate hepatic extraction ratio. Consequently, midazolam clearance in health is largely dependent on liver CYP3A4 activity and less influenced by changes in liver blood flow. 28 Hence, given the observed negative correlation between CRP and midazolam clearance, reduced CYP3A4 enzyme activity secondary to inflammation could account for reduced midazolam clearance in pediatric critical illness. More specifically, interleukin-6 (IL-6) is known to promote CRP synthesis and, in contrast, strongly inhibits hepatic CYP3A4 activity. [29][30][31][32][33] Whether CYP3A4 inhibition wholly accounts for reduced midazolam clearance in pediatric critical illness is unclear. The negative correlation between CV score and midazolam clearance (a threefold increase in midazolam concentrations associated with an increase in CV score from 4 to 12 is consistent) suggests that hepatic vein blood flow is also a critical factor.
Ischemia-induced hepatocyte injury could directly reduce CYP3A4 capacity or as a consequence of IL-6 release indirectly reduce CYP3A4 activity. Reduced substrate delivery to hepatocytes due to reduced hepatic blood flow is a more likely explanation, as no patient in this study had clinical evidence of severe liver damage.
The clinical implication here is that intensivists need to be cog-

ACK N OWLED G M ENTS
The authors would like to thank the patients, parents/carers for participating and the surgical and intensive care teams at University Hospitals of Leicester for their assistance. Bioanalytical support was provided by GlaxoSmithKline, Stevenage, UK.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding authors.

E TH I C S A PPROVA L
The study was approved by the East Midlands-Derby Research Ethics Committee in England (14/EM/1261).