Larger Dose Reductions of Vancomycin Required in Neonates with Patent Ductus Arteriosus Receiving Indomethacin versus Ibuprofen

Ibuprofen and indomethacin are commonly used to induce ductus arteriosus closure in preterm neonates. Our group previously reported that ibuprofen decreased vancomycin clearance by 16%. In this study, we quantified the impact of indomethacin coadministration on vancomycin clearance by extending our vancomycin population pharmacokinetic model with a data set containing vancomycin concentrations measured in preterm neonates comedicated with indomethacin.

ABSTRACT

Ibuprofen and indomethacin are commonly used to induce ductus arteriosus closure in preterm neonates. Our group previously reported that ibuprofen decreased vancomycin clearance by 16%. In this study, we quantified the impact of indomethacin coadministration on vancomycin clearance by extending our vancomycin population pharmacokinetic model with a data set containing vancomycin concentrations measured in preterm neonates comedicated with indomethacin. The modeling data set includes concentration-time data of vancomycin administered alone or in combination with either ibuprofen or indomethacin collected in the neonatal intensive care units of UZ Leuven (Leuven, Belgium) and São Francisco Xavier Hospital (Lisbon, Portugal). The derived vancomycin pharmacokinetic model was subsequently used to propose dose adjustments that yield effective vancomycin exposure (i.e., area under the concentration-time curve from 0 to 24 h [AUC_0–24] between 300 to 550 mg·h/liter, with a probability of <0.1 of subtherapeutic exposure) in preterm neonates with patent ductus arteriosus. We found that indomethacin coadministration reduced vancomycin clearance by 55%. Model simulations showed that the most recent vancomycin dosing regimen, which was based on an externally validated model, requires 20% and 60% decreases of the loading and maintenance doses of vancomycin, respectively, when aiming for optimized exposure in the neonatal population. By analyzing vancomycin data from preterm neonates comedicated with indomethacin, we found a substantial decrease in vancomycin clearance of 55% versus a previously reported 16% for ibuprofen. This decrease in clearance impacts vancomycin dosing, and we anticipate that other drugs eliminated by glomerular filtration are likely to be affected to a similar extent as vancomycin.

INTRODUCTION

Vancomycin is frequently used in neonates as therapy for late onset infections with coagulase-negative Staphylococcus or as an alternative therapy for methicillin-resistant Staphylococcus aureus (1). Recently, Janssen et al. proposed a vancomycin dosing regimen for both preterm and term neonates, based on an externally validated population pharmacokinetic (PK) model yielding effective and safe vancomycin exposure (i.e., an area under the curve [AUC] of around 400 mg·h/liter) from the start of treatment (2).

Comedication given to preterm neonates with a patent (symptomatic) ductus arteriosus (PDA) includes ibuprofen and indomethacin, which have been proven to effectively induce PDA constriction and closure (3). Both nonsteroidal anti-inflammatory drugs (NSAIDs) are known to have renal side effects, as they suppress the vasodilatory effects of prostaglandins leading to vasoconstrictive renal hypoperfusion, even though exact quantification is incomplete (3, 4). Vancomycin clearance (CL) was shown to decrease by 16% when coadministered with ibuprofen (5), upon which it was proposed to decrease the vancomycin dosage for neonates with PDA comedicated with ibuprofen (2). Less is known about the impact of indomethacin on vancomycin CL. Upon quantifying the influence of indomethacin on vancomycin CL, we could improve vancomycin dosing in this special population. Additionally, since vancomycin CL is mainly eliminated by glomerular filtration, a reduction in CL of vancomycin as a result of coadministration with ibuprofen or indomethacin may also imply a reduction in CL for other drugs such as aminoglycosides (5, 6) cleared by the same pathway.

In the present analysis, our goal was to quantify the impact of indomethacin coadministration on vancomycin CL in neonates with PDA, in addition to the previously quantified impact of ibuprofen on vancomycin CL in this population. For this, vancomycin PK data collected during routine therapeutic drug monitoring (TDM) in preterm patients pharmacologically treated for PDA with indomethacin (7) were analyzed within the context of a previously published population pharmacokinetic model for vancomycin and vancomycin coadministered with ibuprofen (5). This model has been externally validated and used to propose dosing guidelines for vancomycin in neonates (2). Model-based simulations were subsequently used to evaluate available dosing regimens (2, 8–10) for vancomycin in preterm neonates with PDA comedicated with ibuprofen or indomethacin and to propose dose adjustments.

RESULTS

Population pharmacokinetic model.

Our analysis showed that indomethacin reduced vancomycin clearance by 55% (see Table S1 in the supplemental material) (fraction of 0.447; relative standard errors [RSE%] of 14%), while the reduction for ibuprofen was 16% (5). Adding indomethacin coadministration as a covariate on the central distribution volume (V₁) did not lead to statistically significant improvement of the model.

Figure 1 illustrates these findings, showing the relationship between individual vancomycin CL values and body weights of patients in the overall data set in the presence or absence of either ibuprofen or indomethacin. Besides the systematic difference in vancomycin CL values between the three groups, a relatively high overall interindividual variability of 33.6% in vancomycin clearance was estimated (Fig. 1; Table S1).

FIG 1.

Vancomycin individual clearance values versus body weight in the overall studied neonatal population (semilog scale). Light gray circles, vancomycin clearance in neonates without NSAID coadministration; blue circles, vancomycin clearance in preterm neonates with PDA with indomethacin coadministration; orange circles, vancomycin clearance in preterm neonates with ibuprofen coadministration.

The model described the data with good accuracy, as confirmed by the goodness-of-fit plots, for all three patient groups (no NSAID, ibuprofen, and indomethacin) (see Fig. S1), while the normalized prediction distribution error (NPDE) analysis confirmed accurate predictions (see Fig. S2 and S3). Estimated PK parameters had acceptable precision, as indicated by the RSE% of the estimates being less than 20%. The bootstrap analysis confirms the robustness of the model (Table S1).

Vancomycin dosing optimization.

Simulations showed that, to maintain an effective vancomycin exposure (i.e., area under the concentration-time curve from 0 to 24 h [AUC_0–24] within 300 to 550 mg·h/liter) when NSAIDs are coadministered in preterm neonates with PDA, different dose adjustments should be made for ibuprofen and indomethacin to compensate for the differences in decreases in vancomycin CL. Table 1 displays how the vancomycin dosing regimen proposed by Janssen et al. (2) for neonates without coadministration of NSAIDS should be adapted when NSAIDs are coadministered, i.e., a decrease of the maintenance dose by 20% for ibuprofen and decreases in both the loading and the maintenance doses by 20% and 60%, respectively, for indomethacin (Table 1).

TABLE 1.

Vancomycin dosing regimen according to Janssen et al. (2) and proposed vancomycin doses for ibuprofen and indomethacin coadministration resulting from model-based simulations with the final model, aiming for a target of AUC_0–24 of between 300 and 550 mg·h/liter

PNA (days)	BW (g)	Vancomycin dosing (2)a		Vancomycin dosing with coadministration of:
				Ibuprofen		Indomethacin
		Loading dose (mg/kg)	Maintenance dose (mg/kg/day [no. of doses])	Loading dose (mg/kg)	Maintenance dose (mg/kg/day [no. of doses]) (20% reduction)	Loading dose (mg/kg) (20% reduction)	Maintenance dose (mg/kg/day [no. of doses]) (60% reduction)
0–7	≤700	16	15 (3)	16	12 (3)	13	9 (3)
	700–1,000		21 (3)		17 (3)		13 (3)
	1,000–1,500		27 (3)		22 (3)		16 (3)
	1,500–2,500		30 (4)		24 (4)		18 (4)
8–14	≤700	20	21 (3)	20	17 (3)	16	13 (3)
	700–1,000		27 (3)		22 (3)		16 (3)
	1,000–1,500		36 (3)		29 (3)		22 (3)
	1,500–2,500		40 (4)		32 (4)		24 (4)
14–28	≤700	23	24 (3)	23	19 (3)	18	19 (3)
	700–1,000		42 (3)		34 (3)		25 (3)
	1,000–1,500		45 (3)		36 (3)		27 (3)
	1,500–2,500		52 (4)		42 (4)		31 (4)
21–28	≤700	26	24 (3)	26	19 (3)	21	19 (3)
	700–1,000		42 (3)		34 (3)		25 (3)
	1,000–1,500		45 (3)		36 (3)		27 (3)
	1,500–2,500		52 (4)		42 (4)		31 (4)

^aJanssen et al. (2) proposed decreases of 2 mg/kg/dose for both the maintenance and loading doses with ibuprofen coadministration.

Monte Carlo simulations in virtual preterm neonates pharmacologically treated for PDA.

Figure 2 shows the probabilities of attaining vancomycin exposure within, above, or below the predefined target range of 300 to 550 mg·h/liter following the dosing guidelines of Janssen et al. (2) (with and without dose reduction of 2 mg/kg/dose for ibuprofen coadministration) and our proposed dose adjustments for coadministration with ibuprofen or indomethacin (see Table 1) in virtual patients resampled from the available PDA patient group.

FIG 2.

Probability of target attainment for AUC_0–24 (first day of treatment) between 300 and 550 mg·h/liter for vancomycin for different dosing regimens, derived from Monte Carlo simulations in virtual preterm neonates with PDA. (Left) Results in preterm neonates with PDA after vancomycin coadministered with ibuprofen. (Right) Results for preterm neonates with PDA after vancomycin coadministered with indomethacin. Each bar represents the results obtained with one dosing regimen (see Table 1 for detailed descriptions the dosing regimens).

The proposed dose reduction when ibuprofen is coadministered decreases the probability of underdosing, especially in the smallest children (Fig. 2 and 3, left). Using vancomycin dosing regimens with no adjustments or with the same adjustment for both NSAIDs would lead to major differences in vancomycin target attainment (Fig. 3) and particularly increase the probability for overexposure and, thereby, the risk of experiencing side effects.

FIG 3.

Vancomycin AUC_0–24 values on the first day of treatment obtained following stochastic simulations for each dosing regimen in hypothetical individuals with birth body weights of 770 g, 1,050 g, and 1,250 g and postnatal ages of 6, 9, and 12 days, respectively. Each color represents one dosing regimen (see Table 1 and Table S2 in the supplemental material for details of each dosing regimen) and the colors intensify with increasing birth body weight. (Left) Results in preterm neonates with PDA after vancomycin coadministered with ibuprofen. (Right) Results for neonates with PDA after vancomycin coadministered with indomethacin. The red dashed lines represent the target AUC_0–24 range of 300 to 550 mg·h/liter.

Stochastic simulations in hypothetical preterm neonates pharmacologically treated for PDA.

Figure 3 shows results of stochastic simulations in representative hypothetical patients with pharmacologically treated PDA illustrating how variability in vancomycin CL is reflected in AUC_0–24 values following vancomycin administration with our proposed dosing (Table 1) and published dosing guidelines (Table S2), with adjustments for comedication when available (3–6). Remaining variability in these plots results from random interindividual variability in vancomycin CL, for which TDM remains necessary.

Figure 3 illustrates that large variability in exposure may be expected depending on both the selected dosing regimen and the birth body weight of the neonate as well as the NSAID involved (Fig. 3).

DISCUSSION

In preterm neonates treated concomitantly with ibuprofen for PDA and with vancomycin for suspected or confirmed late onset sepsis, a 16% decrease in vancomycin clearance was reported previously (5). In the present study, we found a 55% decrease in vancomycin clearance when PDA is treated with indomethacin. Based on these findings, we propose dose adjustments to ensure a safe and effective vancomycin treatment for this special population, i.e., a decrease of the vancomycin maintenance dose by 20% when ibuprofen is coadministered and decreases of the loading and the maintenance doses of vancomycin by 20% and 60%, respectively, when indomethacin is coadministered.

In the model-based simulations, AUC_0–24 values (between 300 and 550 mg·h/liter) were defined as targets, as proposed in recent publications (2, 11). However, vancomycin trough concentrations taken at the end of the first day of treatment are still commonly used as surrogate markers for vancomycin exposure. In adults, trough concentrations of >15 mg/liter are associated with an effective vancomycin exposure of around 400 mg·h/liter. However, Neely et al. showed, using Bayesian modeling, that 60% of adult patients with a vancomycin AUC of at least 400 mg·h/liter, had a trough concentration of <15 mg/liter (12). For neonates, Frymoyer et al. showed that trough levels ranging between 7 and 10 mg/liter were highly predictive of an AUC_0–24 of >400 mg·h/liter (11). Both these studies suggest that guiding dose individualization based on a trough concentration of 15 mg/liter could lead to overexposure and increased risk of adverse events. In addition, when correlating trough concentrations with AUC_0–24, vancomycin dosing frequency should be accounted for (13).

To ensure an efficacious vancomycin treatment, a target AUC_0–24 of around 400 mg·h/liter for a pathogen MIC of 1 mg/liter should be attained from the start of therapy, as this was correlated with a better treatment outcome and a shorter time to reach steady state (14). Therefore, we decided to aim for a therapeutic window of 300 to 550 mg·h/liter. U.S. guidelines recommend an AUC_0–24 of around 700 mg·h/liter for efficiency, when the MIC is >1.5 mg/liter. A higher pathogen MIC indicates development of bacterial resistance and would justify the use of a higher therapeutic target (15) or an alternative drug. When aiming for an (median) AUC of 700 mg·h/liter, the dosing advice in Table 1 should be adjusted by 700/400.

Previously, Janssen et al. proposed to decrease the vancomycin dose by 2 mg/kg/dose when coadministered with ibuprofen (2). This recommendation was shown to have a relatively larger impact in small neonates (see Fig. 3), who receive lower doses on average, tending toward underexposure. This limitation has been considered in the present proposal by decreasing the dose proportionally to the decrease in CL (Table 1).

Even though both ibuprofen and indomethacin belong to the same drug class (NSAIDs) and are used for the same therapeutic indication, the extent to which they influence vancomycin clearance is >3-fold different. While it is unknown whether this results from the drug itself or from a nonequivalent dose compared to this side effect, it seems that a specific dose adjustment for each NSAID should be applied for the best vancomycin treatment outcome. Ibuprofen is associated with a decreased risk of necrotizing enterocolitis and transient renal insufficiency compared to that with indomethacin (16). There are no reviews comparing how different dosing regimens or modes of administration of the different NSAIDs used to treat PDA affect the treatment outcome or the risk for side effects (17). From these results, it also seems that dose adjustments might be required for other drugs with similar physicochemical properties to vancomycin that are coadministered with NSAIDs and are eliminated by glomerular filtration (5). The proposed dosing regimens should be prospectively validated before applying them in clinical practice.

Figure S4A in the supplemental material shows the probability of target attainment for an AUC_0–24 of between 300 and 500 mg·h/liter derived from Monte Carlo following various currently advised vancomycin dosing regimens without dose adjustments in patients without NSAID coadministration. Dosing according to the Dutch Children’s Formulary, British National Formulary, and Neofax (meningitis) guidelines results in considerable underexposure in neonates with neither PDA nor cotherapy with NSAIDs; therefore, it is important that these dosing guidelines are not further reduced using our proposal.

The results of our stochastic simulations show how the relatively high interindividual variability in vancomycin CL is carried over to the yielded exposure, as this variability in CL cannot be accounted for a priori (Fig. 3). The high interindividual variability in vancomycin CL in all neonates makes dosing challenging. Therefore, even though the proposed adjustments improve the vancomycin target attainment in the population as a whole, TDM is still required to individualize dosing in clinical practice.

Conclusions.

In preterm neonates with suspected or confirmed late onset sepsis and pharmacologically treated for PDA, vancomycin CL is reduced by 16% and 55% when coadministered with ibuprofen and indomethacin, respectively. To reach the same exposures as in patients without PDA and with coadministration with NSAIDs, we propose dosing adjustments of 20% in the maintenance dose when ibuprofen is coadministered and reductions of 20% and 60% in loading dose and maintenance dose, respectively, when indomethacin is coadministered, compared to previously reported neonatal dosing guidelines (2). Therapeutic drug monitoring is still required due to the remaining random variability on vancomycin CL that can yield high exposures which increase the risk of adverse events. PK of drugs with similar properties to vancomycin that are also eliminated by glomerular filtration may be affected to a similar extent by NSAID coadministration.

MATERIALS AND METHODS

Data exploration.

For this analysis, we used vancomycin PK data collected during routine TDM at two neonatal intensive care units: University Hospitals Leuven (Leuven, Belgium; here referred to as UZ Leuven) and São Francisco Xavier Hospital (Lisbon, Portugal; here referred to as HSFX). All preterm neonates diagnosed with PDA received either ibuprofen (UZ Leuven) or indomethacin (HSFX) together with vancomycin. Data on vancomycin without comedication from neonates without PDA were all collected at UZ Leuven. Findings from both sets of data have been published separately before by De Cock et al. (5) (UZ Leuven) and Silva et al. (7) (HSFX). The combined data set was used for model development in the present analysis. A summary of the demographics of the patients included in this analysis is provided in Table 2, which shows a large degree of similarity regarding age- and weight-related demographics in these preterm neonates.

TABLE 2.

Summary of demographic characteristics of the patients included in this analysis

Characteristic	Value (mean [range]) for population treated with vancomycin:
	Alone (5) (N = 263)	With ibuprofen (5) (N = 23)	With indomethacin (7) (N = 33)
Postmenstrual age (wks)	31 (24–38)	28 (24–33)	29 (26–35)
Gestational age (wks)	29 (23–34)	27 (24–33)	28 (25–34)
Postnatal age (days)	14 (1–28)	7 (2–12)	11 (4–30)
Birth body weight (g)	1,150 (385–2,550)	832 (415–1,930)	1,000 (570–1,960)
Current body weighta (g)	1,256 (485–2,630)	810 (415–1,930)	981 (628–1,850)

^aThe patient’s body weight at the start of the treatment.

Model development.

The previously published population PK model, developed with the data collected at UZ Leuven to characterize vancomycin disposition and quantify the impact of ibuprofen on vancomycin CL (5), was used as a basis for the present analysis. Briefly, this model concerns a two-compartment model that includes birth body weight (BW), postnatal age (PNA), and ibuprofen coadministration as covariates on CL and current body weight (CW) as a covariate on the central and peripheral distribution volumes (V₁ and V₂) (5). This model was externally validated in a previous study (2). In the present analysis, all population parameters describing vancomycin disposition and the influence of ibuprofen on CL were fixed to the estimates reported by De Cock et al. (5). The combined data set, including the data from both UZ Leuven and HSFX (7), was used to quantify the influence of indomethacin coadministration as a covariate (F_indo) on CL and V₁.

Model selection was based on numerical and graphical criteria (e.g., decrease in objective function value of >3.84 with one more degree of freedom [P < 0.05], relative standard errors below 30%, and unbiased goodness-of-fit plots).

Model validation.

The robustness of the parameter estimates of the final model was assessed by a nonparametric bootstrap. For this, the extended data set was resampled with replacement 1,000 times and stratified on vancomycin comedication (i.e., vancomycin without comedication, vancomycin with ibuprofen, or vancomycin with indomethacin). The resampled data sets were subsequently fitted with the final model, after which median and 95% confidence intervals of the parameters were obtained.

The predictive properties of the model were assessed by a normalized prediction distribution error (NPDE) (18) analysis using the NPDE package in R v3.3.2. Each observed concentration was compared to 1,000 simulated values for that observation to calculate the prediction error (18). The results of the NPDE were also stratified by comedication.

Vancomycin dosing optimization.

The final vancomycin PK model was used for Monte Carlo simulations and stochastic simulations to guide dose adjustments upon coadministration with either ibuprofen or indomethacin. For this purpose, we defined a safe and effective vancomycin target exposure, i.e., an AUC in the first 24 h (AUC_0–24) ranging between 300 and 550 mg·h/liter, which should lead to a median AUC/MIC of 400 mg·h/liter for a MIC of 1 mg/ml. For the recommended dose adjustments, we aimed for a probability of reaching subtherapeutic exposures (AUC_0–24 of <300 mg·h/liter) of <0.1.

As basis for our proposed vancomycin dosing adjustments, we used a recently published dosing regimen for vancomycin (2) (Table 1) that reaches and maintains the vancomycin target AUC_0–24 in children, including preterm neonates. This dosing regimen was based on an externally validated population PK model and proposed a fixed dose reduction of 2 mg/kg/dose for both the loading and the maintenance doses, upon coadministration with ibuprofen, to account for the reduced vancomycin CL. This regimen was evaluated together with other dosing guidelines for vancomycin that are currently in clinical use but that have not been optimized for scenarios with coadministration of NSAIDs (see Table S1 in the supplemental material) (Dutch Children’s Formulary [10], British National Formulary [9], and Neofax [8]).

Monte Carlo simulations in virtual preterm neonates pharmacologically treated for PDA.

For the Monte Carlo simulations, a virtual patient population was created by resampling with replacement 1,000 patients from our original sample of patients with PDA. The final model was used to simulate individual vancomycin concentration-time profiles following dosing with the different guidelines and to calculate AUC_0–24 values for each of the virtual patients. The results are presented as probabilities of exposure attainment within, above, or below the predefined AUC_0–24 target range.

Stochastic simulations in hypothetical preterm neonates pharmacologically treated for PDA.

For the stochastic simulations, three individuals with birth body weights representing the 1st quartile (BW = 770 g), median (BW = 1050 g), or 3rd quartile (BW = 1250 g) and postnatal ages (PNA) of 6, 9, and 12 days, respectively, were derived from the sample of patients with PDA.

For each of these individuals, 1,000 stochastic simulations were performed with the final model taking interindividual variability of the model parameters into account. Simulated individual concentration-time profiles obtained after dosing vancomycin following different guidelines were used to calculate AUC_0–24 for each hypothetical individual.