3.1. Demographics

The most frequently considered covariates for phenobarbital pharmacokinetics in neonates were actual body weight (ABW), gestational, and postnatal age (Table 1 and Table 2).

Some studies have indicated these demographic descriptors as significant covariates for phenobarbital CL [26,32,34]. In addition, Touw et al. also described an association of height and body surface area with Vd and CL, respectively [26]. However, other studies have indicated rather inconsistent data, making conclusions on valid covariates for the drug dosing difficult. We have previously noticed an upward relationship between Vd and ABW, height, and body surface area, whereas CL was not associated with either demographic or clinical features [33]. Pitlick et al. observed no correlation between Vd and gestational age, while CL increased with postnatal age during the first month [35]. Grasela et al. showed that neither Vd nor CL was affected by gestational age [37]. Gilman et al. found no correlation between half-life and either gestational or postnatal age [21]. The study of Völler et al. presented birthweight and postnatal age as the best predictors for maturation of phenobarbital CL and ABW as a predictor for Vd [38]. Moffett et al. showed that significant covariates included fat-free mass (FFM) and postmenstrual age on CL, and FFM and postnatal age on Vd across the pediatric age populations [39]. Back et al. proposed a population nonlinear mixed effect pharmacokinetic, modeling size and maturation functions as covariates of phenobarbital dispositions [51]. In neonates and young infants, both size and maturation functions application was more effective for pharmacokinetic analysis than when only size function was considered. Similar methods and findings have been shown by Thibault et al., where ABW and postnatal age were found as covariates of CL, while ABW predicted phenobarbital Vd [36]. However, this mixed effect approach is relatively exacting and therefore is unlikely to find an application in daily practice.

3.2. Laboratory Parameters

No relationships have been observed between laboratory markers of liver functions (total bilirubin, aspartate aminotransferase, alanine aminotransferase, international normalized ratio) and phenobarbital disposition [33,34,39]. In contrast, a recent study using a nonlinear mixed effect approach has described that albumin increases phenobarbital Vd [36]. From the laboratory markers of renal functions, levels of serum urea, serum creatinine, and blood urea nitrogen (BUN) were tested. Urea level was not found to be a significant covariate for the phenobarbital CL. Additionally, creatinine was also not found to be a significant descriptor of the phenobarbital CL variability [33,45,50], while Moffett et al. described a relationship between serum creatinine and phenobarbital CL [39]. As inflammation can influence CYP450 enzyme activity, the C-reactive protein (CRP) level was also tested as a predictor of phenobarbital CL in the critically-ill neonates undergoing ECMO (47), but no relation was found.

3.3. Asphyxia

The impact of asphyxia has also been studied, but with contradictory results. Gal et al. reported CL reduction in asphyxiated neonates [40,52], while Grasela et al. noticed no effect on CL, while increased Vd was noted in the presence of asphyxia [37]. Pokorna et al. presented severity of asphyxia as a covariate of phenobarbital CL in patients undergoing therapeutic hypothermia [43], but no effect of asphyxia and its severity on the drug pharmacokinetics was shown in a relatively similar patient population using a population pharmacokinetic modeling approach [48].

3.4. Therapeutic Modalities

Therapeutic modalities potentially affecting phenobarbital pharmacokinetics that have been studied are therapeutic hypothermia, renal replacement therapy (RRT), and extracorporeal membrane oxygenation (ECMO). Shellhaas et al., van den Broek et al., and Favie et al. have not identified any impact of hypothermia on the phenobarbital disposition [34,41,42]. Although Filippi et al. stated that phenobarbital administered to newborns under whole body hypothermia resulted in higher plasma concentrations and longer half-lives than expected in normothermic newborns, this study did not contain any normothermic control group, making comparison of the pharmacokinetics between the hypo- and normo-thermic neonates impossible [30]. Thus, the effect of therapeutic hypothermia does not seem to be clinically relevant for phenobarbital dosing. A recent study has found that interaction of severity of asphyxia and hypothermia is associated with a clinically relevant reduction of phenobarbital CL, suggesting the potential relevance of disease characteristics beyond hypothermia itself [43].

Pokorna et al. observed increased phenobarbital CL in neonates receiving ECMO support, while Vd was not significantly different compared to neonates without ECMO [44]. These observations are consistent with the necessity of higher doses in ECMO patients described by Dillman et al. [53]. Thibault et al. reported the effect of ECMO therapy on phenobarbital Vd in neonates after congenital heart surgery, resulting in the need for a higher loading dose, but the drug CL was not affected [36]. In another study, the same research group found a 6-fold increase of phenobarbital CL in neonates and infants undergoing continuous veno-venous hemodiafiltration (CVVHDF) compared to the neonates and infants without CVVHDF. Additionally, the authors found no impact of ECMO on phenobarbital Vd. When analyzed phenobarbital levels before, during, and after ECMO, Michaličková et al. observed that phenobarbital CL linearly increased with time during the ECMO phase, while in the post-ECMO phase, CL initially decreased and subsequently increased slowly, which was likely driven by maturation [45]. Moreover, the authors found no impact of ECMO on phenobarbital Vd. Thus, data indicated that higher phenobarbital doses are needed during ECMO, however, the particular dosing recommendation varied.

3.5. Drug Interactions and Genetic Polymorphisms

Impact of the co-medication of several drugs on phenobarbital pharmacokinetics has been investigated repeatedly [37,41,42,46,53]. No significant effect of co-administered dopamine, dobutamine, norepinephrine, phenytoin, sufentanil, midazolam, tramadol, or furosemide was observed in short-term concomitant treatment [46]. Michaličková et al. also did not observe any effect of diuretics and inotrope use on phenobarbital CL in critically-ill neonates undergoing ECMO [41]. Although Moffett et al. concluded that midazolam, phenytoin, and pantoprazole significantly affected phenobarbital CL [39], this finding has been questioned as a chance finding only [54,55]. The cytochrome P450 2C19 genotype did not also affect phenobarbital pharmacokinetics in neonates and infants [47].

4. Covariate-Based Phenobarbital Dosing

The routinely used phenobarbital dosing in neonates is based on body weight and consists of loading dose of 15–20 mg/kg followed by maintenance dose of 3–5 mg/kg per day. Several body weight-based dosing regimens or nomograms have been described. The intersection of these findings well corresponds to the above-mentioned dosing routines. Furthermore, doses at the lower limit of the suggested range should be preferred in patients with severe asphyxia, while the upper limit of the range should be targeted in neonates receiving ECMO support.