Amikacin or vancomycin exposure alters the postnatal serum creatinine dynamics in ELBW neonates

Background: Disentangling adverse drug reactions from confounders remains a major challenge to assess causality and severity in neonates. Vancomycin and amikacin are nephrotoxic but are still often prescribed in neonates. We selected these compounds to assess their impact on creatinine dynamics as sensitive tool to detect a renal impairment signal. Methods: A recently developed dynamical model that characterized serum creatinine concentrations of 217 ELBW neonates (4036 observations) was enhanced with data on vancomycin and/or amikacin exposure to identify a potential effect of antibiotic exposure by nonlinear mixed-effects modelling. Results: Seventy-seven percent of ELBW patients were exposed to either vancomycin or amikacin. Antibiotic exposure resulted in a modest increase in serum creatinine and a transient decrease in creatinine clearance. The serum creatinine increase was dependent on gestational age, illustrated by a decrease with 56% in difference in serum creatinine between a 24 or 32-week old neonate, when exposed in the 3rd week after birth. Conclusions: A previously described model was used to explore and quantify the impact of amikacin or vancomycin exposure on creatinine dynamics. These tools can be used to explore minor changes, or compare minor differences between treatment modalities.


How to recognize the signal in the noise
There are numerous reasons why approval of drugs for use in (pre)term neonates is still lagging behind. A major issue is the more difficult demonstration of efficacy and safety, as this commonly needs assessment tools tailored to this population [1][2][3]. Related to safety and adverse events, this includes assessment of seriousness, causality and severity within the framework of regulatory requirements [1,3]. For both causality as well as severity, disentangling adverse drug reactions from confounders (associated morbidity, multi-pharmacotherapy, maturational changes in laboratory a relevant cohort with 2814 creatinine observations in 148 extreme low birth weight (ELBW) neonates [6]. Besides maturational changes (gestational age, postnatal age), caesarean section and ibuprofen exposure affected these dynamics, in line with previous observations [7,8].
The same dataset and approach can be used to assess the incidence and severity of renal impairment for other commonly used medicines in neonates [9]. As recently described by Al-Turkait et al. in this special issue on drug utilization studies in paediatrics, both aminoglycosides and vancomycin are in the top 10 of most commonly prescribed drugs in neonates, and are nephrotoxic [10]. We therefore selected these compounds to assess their impact on creatinine dynamics as a sensitive tool to detect a signal of renal impairment.

Aminoglycosides and vancomycin related acute kidney injury: developmental toxicology
Toxicity has limited the use of aminoglycosides, but there is a consistently lower rate of acute kidney injury (AKI) in neonates when compared to adults [11]. This suggests either developmental toxicology (less in neonatal life), or poorer performance to recognize the signal in neonates (inaccurate AKI tool), or both. Renal tubular cell toxicity starts with binding to acidic phospholipids of the cell membrane (aminoglycosides are charged negative), followed by cellular uptake, facilitated by megalin and cubulin ligands [11]. These ligands-related uptakes display saturation, explaining the practice of extended time interval dosing of aminoglycosides to avoid toxicity, and improve efficacy (peak concentration related) [9,11,12]. Intracellular accumulation in endosomes and mitochondrial stress subsequently occurs (related to ribosomal inhibition, similar to the bacterial ribosomal inhibition). Because of the maturational expression of these ligands, it seems that aminoglycosides are much better tolerated in neonates compared to adults. In neonates, Nestaas et al. reported an incidence of 8.4% for nephrotoxicity (increased creatinine, urinary aminopeptidase, 50/589 events) in neonates exposed to aminoglycosides [13].
Vancomycin exposure is also associated with a higher risk of AKI in adult critical care, with a relative risk of 2.45 (95% confidence interval, 1.69 to 3.55) [14]. Based on both in vitro and in vivo observations, the assumed mechanism also relates to dose-dependent intracellular accumulation and oxidative damage in the renal tubular cells [15]. In neonates, the evidence on vancomycin nephrotoxicity has been summarized and is suggested to occur in 1-9% of exposed cases [16][17][18].
These incidences are supported by two more recent reports [19,20]. In the Leroux prospective validation study, only 2/190 cases displayed nephrotoxicity, both associated with hemodynamic instability and vasopressor need, and with subsequent normalization [19], while Germovsek et al.
observed nephrotoxicity in 4/77 neonates [20]. Target attainment for vancomycin exposure (AUC24h>400) was common in both recent cohorts, indicating that these findings better reflect contemporary target exposure. This is of relevance, as toxicity relates to the exposure (AUC24h>800) and the duration of vancomycin administration (cumulative exposure, AUCcum) [15,21]. Finally, there However, as the dynamics in serum creatinine are most pronounced in ELBW neonates with an initial physiological increase up to day 3 and a subsequent progressive decrease over the next weeks, these trends are difficult to capture within a 'fixed or proportional increase to former creatinine observation' approach, and perhaps does not provide the granularity needed to detect 'minor changes' [25].
The aim of this study was to quantify the effect of aminoglycoside and/or vancomycin exposure on renal impairment reflected by the change in serum creatinine dynamics during the neonatal period of ELBW neonates. This knowledge may inform and help clinicians to recognize and examine the potential of renal impairment in their vulnerable patients beyond the commonly used AKI definition.

Study population, clinical characteristics and ethics
The dataset consists of a pooled dataset of 4026 serum creatinine observations in 217 ELBW cases in the first 42 days after birth and admitted at the Neonatal Intensive Care Unit of UZ Leuven, as recently published [6]. Serum creatinine was analyzed enzymatically by Roche (Roche Diagnostics, Mannheim, Germany) in all cases, and all measurements were isotope dilution mass spectrometry (IDMS) traceable. All clinical data (demographics, but also information on co-medication like ibuprofen or inotropic agents and type of delivery) were already available in this dataset [6].
For the purpose of this project, all individual electronic medical files were searched for days exposed to vancomycin, amikacin, or both, as this is the established antibiotic regimen for suspected late onset sepsis (LOS) in this NICU [26]. Subsequent adaptations of this empiric treatment were based on the pathogen isolated [27]. Another indication for amikacin in this NICU, of relevance for the current study, is necrotizing enterocolitis (co-treatment with piperacillin-tazobactam). Ethical approval of the current study covered the additional data search (model development and evaluation, S63405) [6]. and was set to start on day 5 (first week of life), day 19 (third week of life) and day 33 (fifth week of life) after birth. Comparison was performed between the predicted serum creatinine and creatinine clearance over a period of six weeks for typical ELBW cases who were either or not exposed to AB.

Refined serum creatinine model
Simulations were performed in a deterministic setting where no inter-individual variability is considered and the predicted median profiles were illustrated as a function of included covariates.
Additionally, stochastic simulations, including inter-individual variability, were carried out to assess the difference in predicted creatinine concentrations between patients being exposed to AB exposure and those who are not. In total, 1000 simulations were performed for each gestational age group.

Population
The median gestational age and birth weight was 27 weeks (26 -28 weeks IQR) and 830 g (720 -910 g IQR), respectively ( Table 2). Of the studied ELBW population, 72% and 71% received (at least one day) vancomycin or amikacin during the first six weeks of life, respectively. Of these patients, 77% were exposed to either vancomycin or amikacin, 66% were simultaneously exposed to vancomycin and amikacin for at least one day ( Table 2). During the first six weeks after birth, patients were exposed on average 6.8 days and 5.6 days to vancomycin or amikacin, respectively. Patients with a lower gestational age were more often exposed to AB, as compared to patients with a higher gestational age (Figure 1).

Refined serum creatinine Model
Antibiotic exposure, defined as receiving either vancomycin or amikacin, resulted in a lower overall creatinine clearance and higher serum creatinine concentrations due to a significant increase of 30% in t50 when exposed to AB (-40.6 points in OFV). The parameter t50 reflects the time point in days where half of the maximum additional achieved clearance is reached. A longer t50 corresponds with lower creatinine clearance values. The combination of vancomycin and amikacin exposure did not indicate a synergistic effect on the clearance parameters. For a typical ELBW patient of 24 weeks with or without being exposed to AB, the t50 was estimated at 31.9 days and 41.7 days, respectively and decreased with approximately 10% for each increased week of gestation (Table 3). Table 3: Parameter estimates of refined creatinine model together with effect size estimates. CV, coefficient of variation; MOD, mode of delivery (1 for C-section and 0 for vaginal delivery); AB, antibiotic exposure present (1 for yes and 0 for no), IBU, ibuprofen exposure (1 for yes and 0 for no).
Median GA was set at 26.56 weeks.

Parameter (unit) Estimates (RSE%) IIV [CV%]
Population parameters Predicted serum creatinine concentration time profiles for four typical ELBW neonates are shown in Figure 2a. An increase in serum creatinine concentrations is observed during AB exposure. The extent of the difference in predicted serum creatinine concentrations during AB exposure at the end of first week of life is less pronounced when focusing on various gestational age groups ( Figure 3). This distinction is more obvious for AB exposure at the end of the third and fifth week of life, with a higher difference in serum creatinine concentrations for the youngest ELBW neonates. For exposure at the end of the first week of life, the difference between being exposed to AB or not was on average 0.056 mg/dL (Table 4). For exposure at the end of the third week, i.e. starting on day 19 until day 21 after birth, the difference in creatinine concentration for a neonate at 24 weeks gestation amounts 0.049 mg/dL, and decreased with 56% for a 32-week-old neonate. For a typical neonate of 24 weeks gestation, the creatinine clearance at day 21 after birth was estimated to be 0.44 ml/min being exposed to AB and increased with 12% to 0.49 ml/min without AB exposure. Similar trend was observed for exposure at the end of the fifth week, i.e. from day 33 to 35 after birth, where the difference in creatinine concentrations between being exposed and not exposed, decreased with 60% between 24 and 32 weeks of gestation (Table 4).

Discussion
This study illustrates that antibiotic exposure, defined as either vancomycin or amikacin treatment, is associated with an increase in serum creatinine concentrations due to a decreased creatinine clearance capacity in the ELBW neonatal population. These changes can be quantified, but the increase in creatinine generally remained below the 0.3 mg/dL threshold of the most recently developed AKI tool (Table 1 vs Table 4). This suggests that more tailored, sensitive tools need to be used to quantify renal side effects of drugs, like amikacin, vancomycin, or the earlier reported ibuprofen, and these may serve for a tailored or comparative pharmacovigilance [6]. Of notice is that the absolute increase in serum creatinine in ELBW cases depends on additional characteristics, further illustrating the difficulty to 'retrieve the signal' based on (fixed) increments only. The increase in serum creatinine due to antibiotic treatment is most pronounced when administered during the first week of life (day 5 to 7 after birth), as a result of limited clearance capacities due to immature kidney function. This increase of serum creatinine concentrations during the first week of life appears to be independent of gestational age. When focusing on the difference in serum creatinine concentrations during the third week after birth and fifth week after birth, a gestational age dependency is observed (Figure 3). This is because a higher increase in serum creatinine is observed for the ELBW patient with lower gestational age. An increase of 0.049 mg/dL or 0.022 mg/dL is observed for an ELBW neonate of 24 weeks gestational or 32 weeks gestation, respectively, when antibiotic exposure occurs from day 19 to day 21 after birth.
With the current dataset it was not possible to characterize the potential synergistic effect of combination therapy of vancomycin and amikacin. This is most likely due to the fact that the days that ELBW neonates are exposed to either vancomycin or amikacin (77%) is almost similar as the days that these neonates are exposed to both antibiotics simultaneously (66%), as the combination is the standard treatment for LOS ( Figure 1) [26]. The estimated effect of receiving either vancomycin or amikacin on t50, one of the parameters characterizing creatinine clearance, increased with 30% when exposed to an antibiotic, resulting in a lower creatinine clearance. This effect did not significantly differ (27%) when assessing the combination therapy or improved the model fit.
The small but quantifiable fluctuations observed in creatinine concentrations and creatinine clearance can be identified as a limitation, but this study showed that the impact of both maturational and non-maturational related effects -like exposure to antibiotics on kidney function -can be recognized, including renal adverse drug reactions. Recognition of such drug toxicity signals in neonates, especially considering the vulnerable ELBW population, should be further developed, as we illustrated its feasibility.

Conclusions
A previously described model on creatinine dynamics has been used to explore and quantify the impact of exposure to amikacin or vancomycin on creatinine dynamics. It seems that such a tool is more sensitive to detect signal of renal impairment beyond the existing AKI grading system. In this way, such tools can be used to explore minor changes, or compare minor differences between treatment modalities.

Non-linear mixed effect modeling description
The population analysis was performed by applying non-linear mixed effect modelling approaches. These models take into account both explained and unexplained variability at inter-and intra-individual levels. Nonlinear mixed effect models are characterized in terms of: (i) Fixed effects: this is the population average of the model parameters θ. These parameters are susceptible to various factors, such as physiological characteristics (gestational age, body weight, etc), genetic characteristics, or drug-drug interactions. These last factors are the fixed effect covariates, zi.
(ii) Random effects: this is the part of the variability that is not explained by the above fixed effect and allows quantification of: • The inter-individual variability (also called the between subject variability), which is the variability between two different individuals. It is expressed by ω 2 , which is the variance of the fixed effect Thereby, three parameters have to be estimated: • the fixed effect vector: • the random effect parameter quantifying the residual unknown variability: 2 • the random effect parameter quantifying the inter-individual variability: Ω.