Hyperinflammation Reduces Midazolam Metabolism in Critically Ill Adults with COVID-19

Current guidelines for the management of pain, agitation, delirium, immobility and sleep (PADIS) in critically ill patients are widely endorsed and applied at Intensive Care Units (ICUs) worldwide. They include aiming for a comfortably awake state, avoiding benzodiazepine-based continuous sedation and other measures in order to avoid or mitigate delirium [1]. However, titrating analgesia and sedation towards wakefulness has been challenging in acute respiratory distress syndrome (ARDS) related to coronavirus disease 2019 (COVID-19) due to frequent application of neuromuscular blockade and prone positioning, requiring higher levels of sedation [2]. This has translated into high incidence of benzodiazepine-related delirium and an increased number of days spent in coma [3]. It seems likely that benzodiazepine drug dosing may have been too high in many patients in circumstances where deep sedation was perceived as essential to avoid patient-ventilator dyssynchrony and ensure patient comfort during proning.

Midazolam, the most used benzodiazepine for continuous intravenous (iv) sedation, undergoes extensive metabolism by hydroxylation via CYP3A enzymes to form mainly the metabolite 1-OH-midazolam and a very small amount of 4-OH-midazolam. After hydroxylation, hydroxymidazolam is subsequently metabolized to chiefly 1-OH-midazolam-glucuronide as its major metabolite and renally excreted. Both an in vitro study and a study in healthy volunteers showed that the metabolites 1-OH-midazolam and 1-OH-midazolam-glucuronide have a sedative potency compared to midazolam of 60-80% and 10%, respectively [4, 5]. Due to the CYP3A-mediated metabolism of midazolam to 1-OH-midazolam, midazolam is a known indicator for human CYP3A4 activity. CYP3A can be down regulated by cytokine release during inflammation resulting in decreased drug clearance [6‐9]. In an in vitro study, it was found that CYP3A4 activity is reduced by 50% (EC₅₀) at an interleukin 6 (IL-6) concentration of 17.1 pg/mL [10]. Furthermore, Vet et al. demonstrated reduced CYP3A-mediated clearance of midazolam in critically ill children with inflammation and organ failure [11].

Since hyperinflammation plays an important role in COVID-19, altered midazolam pharmacokinetics seem likely. Therefore, we aimed to develop a population pharmacokinetic model for midazolam and to assess the effect of inflammation, reflected by IL-6 and C-reactive protein (CRP), on the pharmacokinetics of midazolam administered as continuous infusion to adult intensive care patients with COVID-19 infection. We hypothesized that a strong proinflammatory response reduces midazolam clearance.

We developed a population pharmacokinetic model of intravenous midazolam provided as a continuous infusion in critically ill COVID-19 patients. The aim of our study was to determine the effect of inflammation on the pharmacokinetics of midazolam. We found that incremental levels of IL-6, an important mediator of inflammation in COVID-19, were significantly associated with decreased midazolam clearance and predicted decreased clearance more significant than CRP. Our finding of a negative correlation between IL-6 levels and midazolam clearance is in line with the study of Vet and colleagues in a pediatric ICU population [11].

We also showed that decreased renal function was significantly associated with decreased 1-OH-midazolam-glucuronide clearance. Similarly, a previous study observed accumulation of 1-OH-midazolam-glucuronide in patients with renal failure [5].

In contrast to previous studies albumin was not found to be correlated to midazolam clearance (midazolam is 94–97% protein bound) [14‐16]. Although midazolam is highly protein bound, it seems unlikely that midazolam clearance is reduced during hypoalbuminemia. You would expect that a decrease in albumin would increase the unbound midazolam concentration and therefore result in increased clearance. Given the found reduced midazolam clearance, other proteins may also play a role in protein binding of midazolam or as suggested by the authors of Franken et al. the reduced clearance may be due to an inflammatory response and that hypoalbuminemia can be an expression of inflammation and critical illness [14]. However, there may be a difference in patients between our population and the studied terminally ill patients by Franken et al. (43–93 years; median 71 years, albumin 13–39 g/L; median 25 g/L and CRP 1 to 625 mg/L; median 92 mg/L).

The pharmacokinetics of midazolam established in the current study differ from those reported by Swart et al [12]. Swart and colleagues reported midazolam pharmacokinetics in critically ill adults (21–84 years; mean 57 years and APACHE II 6 to 34; median 26) with and without alcohol abuse. Patients with alcohol abuse showed a lower midazolam clearance, which probably is the result of reduced CYP3A activity due to alcoholic cirrhosis. The simulated midazolam pharmacokinetics with normal IL-6 concentration established in the current study are similar to the study of Swart et al. in the patient group without alcohol abuse and so without possible reduced CYP3A4 activity [12]. However, when increased IL-6 levels are taken into account a clearly reduced midazolam clearance was observed.

We found a median midazolam plasma concentration of 1446 µg/L. This appears to be high compared to previous investigations [17‐19]. Our median plasma concentration is comparable to the finding by Nies et al. They observed a median midazolam concentration of roughly 1500 µg/L for RASS -5 [20]. The difference in median concentrations are probably due to the fact that on the one hand, the study by Bremer et al. included post-operative ICU patients and the study by Oldenhof et al. included only 17 patients, and on the other hand that the study by Nies et al. included more long term severe ICU patients.

There appears to be a good correlation between both median midazolam and 1-OH-midazolam plasma concentrations with the degree of sedation [19]. However, since there is a large interpatient variability in plasma concentrations of midazolam and metabolites, it is difficult to formulate a target attainment for a particular degree of sedation [19]. Nonetheless, when we simulated an IL-6 concentration of 116 pg/mL in our model, a steady-state midazolam concentration of almost 3500 µg/L was reached. Similar to our study, other studies observed also relatively high midazolam dose requirements in patients with COVID-19-related ARDS [21, 22]. Interestingly, Tapaskar et al. found a higher total median dose of midazolam in COVID-19 ARDS patients compared to non-COVID-19 ARDS patients. However, no difference in median RASS scores were found between their COVID-19 ARDS group and non-COVID-19 ARDS group, suggesting the higher midazolam dose in COVID-19 ARDS patients was required to achieve equivalent levels of sedation targeted in non-COVID-19 ARDS patients. [22] The high sedation requirements in COVID-19 patients are possibly related to patient factors such as agitation, increased patient-ventilator dyssynchrony, challenging mechanical ventilation in prone position and ARDS, and hyperinflammation, which has previously been linked to opioid tolerance [2, 3, 23]. Additionally, the binding of midazolam to the GABA receptor is saturable [24]. Therefore, a so-called ‘ceiling effect’ may occur, due to which greater doses will not produce greater pharmacological effects. However, limited information is known about the effect of COVID-19 and hyperinflammation on the functioning of GABA receptors and more research is needed on the possible saturation of the receptors.

Our results may have implications for more personalized titration of midazolam in COVID-19, mainly to avoid oversedation and subsequently increased length of ICU stay. Our findings indicate that IL-6 is an important biomarker indicative of reduced midazolam clearance. This theoretically contributes to prolonged coma, increased delirium rates and longer ICU stay, which seems to be supported by recent work [3]. Additionally, serum creatinine is an important biomarker for decreased 1-OH-midazolam-glucuronide elimination. Despite the potency of only 10% for 1-OH-midazolam-glucuronide, plasma concentration up to 39,000 µg/L as found in our study can contribute significantly to prolonged sedation effect in patients with kidney injury.

Therefore, increasing levels of IL-6 and renal failure should lead to incentives to decrease midazolam dosage as much as feasible to mitigate adverse consequences of oversedation.

Gordon et al. recently showed that in critically ill patients with COVID-19, an IL-6 receptor antagonist improved outcomes, including survival and becoming standard care of treatment [25]. In our cohort, none of the patients received an IL-6 receptor antagonist. Theoretically, IL-6 inhibition could neutralize the downregulation of CYP3A enzymes and hence also the reduced clearance of midazolam. Therefore, we suggest that the effect of IL-6 receptor antagonists on midazolam clearance should be further investigated.

In this study, we did not relate the pharmacokinetics to the pharmacodynamics of midazolam. Model-based Therapeutic Drug Monitoring (TDM) might be a valuable tool to relate the pharmacokinetics to the pharmacodynamics of midazolam and to identify patients at risk. Model-informed precision dosing has already been used for antibiotics in critically ill patients, but for sedative agents such models are scarce [26, 27]. For the individual patient, model-based TDM may help earlier dose reduction of midazolam or a switch to other sedative drugs, which may contribute to shorter duration of mechanical ventilation and mitigate days in coma and with delirium.

Several limitations of the study should be noted. The data was limited with respect to the number of patients and samples. The pharmacokinetic model showed a tendency towards under prediction of midazolam concentrations in the first week after the start of treatment, which could not be described in the pharmacokinetic model due to the limited number of available samples within this interval. In addition, especially for observations after withdrawing infusion, the model tends to an under prediction of the lower concentrations of midazolam and metabolites. Furthermore, depending on how concentrations of midazolam and metabolites behave above the upper limit of quantification (ULQ), the extrapolated samples above the ULQ may influence the model performance with more variability or bias.

We did not correlate the effect of the reduced midazolam clearance with the previous reported risk of delirium with midazolam and other risks, such as prolonged coma and increased ICU length of stay [3]. We also did not elucidate the reasons for the observed high-dose midazolam administration. Therefore, further studies on the pharmacodynamics in this population are needed before any firm conclusions can be drawn on dose adjustments. For future research, we propose a larger study population with more observations and development of a population pharmacodynamic model based on RASS scores. Another possible limitation of our study was that we did not identify patients’ genetic polymorphisms of CYP3A4/5. When expressed, it could potentially affect the metabolism of midazolam [28].

This study shows that inflammation, reflected by high IL-6, significantly reduces midazolam clearance in critically ill patients with COVID-19. This knowledge may help to identify COVID-19 patients at increased risk for prolonged sedation due to oversedation. Further studies are warranted to relate the pharmacokinetics to the pharmacodynamics of midazolam.