Pharmacokinetic model‐guided enoxaparin dosing in the Neonatal ICU: Retrospective cohort study to plan for prospective feasibility trial

Abstract

Traditional milligram per kilogram (mg/kg) dosing of enoxaparin in neonates frequently fails to achieve target anti‐Xa levels promptly, necessitating repeated laboratory monitoring and dose adjustments. This study investigated whether a personalized dosing strategy based on predicted individual clearance and volume of distribution could improve outcomes, comparing standard‐of‐care (SOC) mg/kg dosing to pharmacokinetic (PK) model‐informed precision dosing (MIPD). A retrospective analysis was conducted on hospitalized neonates treated with enoxaparin at less than 44 weeks postmenstrual age from 2019 to 2022. Data on demographics, drug dosing, PK model covariates, and clinical outcomes were extracted from electronic health records and analyzed using the Pumas‐AI Lyv dosing tool. The primary focus was on comparing the initial SOC dose to the MIPD‐recommended dose. The secondary outcome measured was the time required to achieve therapeutic anti‐Xa levels. The study included 168 neonates with a median postnatal age of 15 days (range 1–149) and a median dosing weight of 3.1 kg (range: 0.82–5.2). MIPD‐recommended initial doses were 20%–60% higher than SOC doses in 32% of the cases and over 60% higher in 11% of cases. Neonates who received SOC doses that were much lower than the MIPD recommendation showed the longest delays in reaching therapeutic anti‐Xa levels. The results indicate that PK model‐informed of enoxaparin dosing leads to higher initial dosages than SOC in neonates, potentially reducing the time to therapeutic anti‐Xa levels. These findings are being utilized to define dosing limits for a prospective trial of MIPD in neonatal intensive care settings.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Initial dosing of enoxaparin based on the CHEST and ASH mg/kg guidelines is the standard‐of‐care (SOC) for the treatment of arterial and venous thrombus in neonates; however, it often fails to achieve targeted anti‐Xa levels, thereby requiring dose titration and close monitoring. Despite clinical recognition of the need for higher first doses, an alternative standard has yet to be established. Model‐informed precision dosing (MIPD), which integrates complex mathematical models of pharmacokinetic data with individual demographic and clinical patient characteristics is a tool for drug dosage optimization. MIPD has been successfully implemented into the care of neonates for vancomycin dosing.

• WHAT QUESTION DID THIS STUDY ADDRESS?

How does the SOC weight‐based first dose of enoxaparin in neonates compare with the recommended dose generated using an MIPD tool? Is there evidence that using MIPD can decrease the time to therapeutic pharmacodynamic end point?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

The dose recommendations made by the MIPD tool showed that if individual PK parameters of predicted clearance and volume of distribution are considered, neonates would receive a higher first dose of enoxaparin compared with standard weight‐based dosing. Furthermore, infants who were the most underdosed with SOC had the longest time to reach a therapeutic anti‐Xa level.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

In the era of precision medicine, MIPD is a tool offering a more personalized approach to dosing that may lead to improved patient outcomes in contrast to standard weight‐based guidelines.

INTRODUCTION

Critically ill neonates are a complex population for whom drug therapy options and dose optimization strategies are lacking. New drug development is hampered by the practical and ethical challenges of studying neonates, and as a result, off‐label drug use in the neonatal intensive care unit (NICU) is common. ¹ Without the robust trial data needed for drug approval, drug dosage for neonates is typically derived using weight‐based (allometric) scaling of effective doses from other populations. ² This approach ignores less visible, potentially impactful, physiological differences that can lead to highly variable drug exposure, ³ , ⁴ therapeutic failure, ⁵ and significant toxicity. ⁶

Enoxaparin is the low molecular weight heparin (LMWH) formulation most commonly used in neonates for the prevention and treatment of arterial and venous thrombotic events. However, neonates, among all pediatric patients with venous thrombi, have the lowest chance of complete thrombi resolution (42.9%) and a median time to complete resolution of 34 days. ⁷ This may be due, in part, to delayed identification of therapeutic doses associated with the current weight‐based neonatal dosing guidelines. One study found that, in neonates less than 2 months, an average of 2.8 dose titrations were needed to achieve a therapeutic anti‐Xa activity level between 0.5 and 1 IU/mL. ⁸ Despite knowing that current neonatal dosing guidelines rarely facilitate rapid target attainment, clinicians in the NICU continue to rely on weight‐based dosing paradigms.

Initial doses are titrated to response using therapeutic drug monitoring (TDM) over a period of days or weeks. Blood samples are usually collected after the first two drug doses and used to assess current anti‐Xa activity and adequacy of enoxaparin dosing. If anti‐Xa activity is outside the established clinical therapeutic range of 0.5–1 IU/mL, the patient's dose is adjusted accordingly. Often, this process must be repeated multiple times before a clinically effective dose is identified and during that time the infant is subjected to multiple painful blood draws and an extended period of suboptimal therapy. There are limited data suggesting that prolonged time to therapeutic anti‐Xa is associated with poorer clinical outcomes. In a study of pediatric enoxaparin dosing protocols, ⁹ successful resolution of thrombus was directly associated with obtaining a therapeutic anti‐Xa concentration upon the first measurement. Thus, a tool that can improve dosing paradigms by quickly identifying optimal dosage regimens is urgently needed in this population.

In the last 5 years, model‐informed precision dosing (MIPD) techniques have gained popularity with the emergence of software platforms designed for precision dosing of narrow therapeutic drugs. ¹⁰ , ¹¹ , ¹² Briefly, MIPD leverages patient‐specific covariates (e.g., body weight and serum creatinine) and pharmacokinetic (PK) modeling to provide individualized predictions of drug clearance (CL) and volume of distribution (Vd). Once known, these individual PK parameters can be used to optimize dosing in real time. Unfortunately, there is very little experience with MIPD in the NICU, which prompted this retrospective comparison of weight‐based, standard‐of‐care (SOC), dosing to MIPD in preparation for a prospective MIPD feasibility trial.

METHODS

This retrospective cohort study was approved by the Hospital for Sick Children (SickKids) Research Ethics Board prior to data collection. The Hospital for Sick Children (Toronto, Canada), electronic health record (EHR) system was queried for hospitalized infants of any gestational age, but less than 44 weeks postmenstrual age (PMA), treated with enoxaparin between January 2019 and December 2022. Infants that received enoxaparin for thrombus prophylaxis were excluded from consideration. Data collection included demographics (gestational age, postnatal age, sex, and birthweight), dose administration records, and laboratory values (serum creatinine, hematocrit, platelet count, and anti‐Xa activity).

Comparator arms: standard‐of‐care dose versus MIPD dose

This study is focused on the first dose of enoxaparin received by patients in the study cohort and how it compared with (1) the dose prescribed at the first therapeutic anti‐Xa level and (2) a post hoc MIPD recommendation for initiating therapy (i.e., the dose that would have been prescribed if MIPD was available). If the model‐informed starting dose is closer to the dose at first therapeutic anti‐Xa compared with the actual starting dose, that would provide strong support for conducting a prospective feasibility study.

The SickKids formulary includes SOC dosing guidelines for enoxaparin in neonates. All infants were prescribed a starting dosage regimen of 1.75 mg/kg rounded to the nearest 0.5 mg every 12 h (q12h). This local SOC dosing is based on pharmacy literature review of multiple sources. Subsequent doses were guided by TDM with the first anti‐Xa activity level collected 4 h after the second dose. Dose adjustments were based on an internal nomogram and duration of therapy varied by indication. The MIPD platform, Lyv, developed by Pumas‐AI, was used to calculate a post hoc initial dosage recommendation for each patient. Lyv relies on a Bayesian population PK model developed in 159 infants and validated using data from 69 infants receiving enoxaparin. ¹³ , ¹⁴ Final parameter estimates and their associated distributions from the model developed by Moffett et al. were used as prior information when developing the Lyv model. Body weight, PMA, serum creatinine, and hematocrit (hct) were retained as significant covariates on CL, while body weight and platelet count were retained on Vd.

Statistical analysis

The primary outcome of the study was the comparison of the initial dose of enoxaparin between SOC dosing and MIPD. The secondary outcome focused on the time required to reach the first therapeutic anti‐Xa level, analyzed according to the variance between SOC and MIPD initial dosing recommendations. For descriptive analysis, percentage change from SOC dose was categorized as within ±20%, 20%–40% above, 40%–60% above, or greater than 60% above. Continuous data were summarized using means with standard deviations or medians with ranges, while categorical data were presented as frequencies and percentages. The Wilcoxon singed‐rank test was used to determine whether a statistically significant difference exists between the therapeutic SOC doses achieved after clinical titration, and the starting doses recommended by the MIPD tool (α = 0.05).

RESULTS

There were 168 neonates included in the study. Demographic and laboratory information are described in Table 1. The median gestational age, birthweight, and postnatal age at enoxaparin dosing were 38 weeks, 2.8 kg, and 15 days, respectively. The median (IQR) SOC dosage received in the retrospective cohort was 1.74 (1.62, 1.83) mg/kg every 12 h. Among the infants who received SOC dosing, the median (IQR) time to achieve a therapeutic anti‐Xa was 40.1 (13.5, 171.9) h. A median number of four dose adjustments were required to achieve therapeutic anti‐Xa. The median (IQR) MIPD starting dosage for the cohort was 1.99 (1.79, 2.3) mg/kg dosed every 12 h.

TABLE 1.

Demographic characteristics and PK Model Variables for the Retrospective Cohort.

Demographic variables	Median	Minimum	Maximum
Gestational age at birth (weeks)	38	23	41
Birthweight (kg)	2.8	0.47	5.0
Postnatal age at dosing (days)	15	1	149
Postmenstrual age at dosing (weeks)	40	29	45
Body weight at dosing (kg)	3.1	0.82	5.2
PK model covariates
Serum creatinine (μmol/L)	0.35	0.15	3.5
Hematocrit (%)	36.5	20.1	57.9
Platelets (×109/L)	212	37	648

When comparing SOC dosing to MIPD‐recommended dosing (referred to as “MIPD‐dose” for rest of manuscript) in each infant, the MIPD dose was within ±20% of the SOC dose in 57% of infants, 20%–40% above the SOC dose in 18%, 40%–60% above the SOC dose in 14%, and greater than 60% above the SOC dose in 11% (Figure 1, panel a). Patient‐specific covariate values were stratified by dose type (MIPD dose, or SOC) and summarized in Table S1. Infants who were most underdosed with SOC dosing (as compared to MIPD‐dose) had lower gestational age at birth and birthweight, lower hematocrit at the time of enoxaparin dosing (Table S1).

FIGURE 1.

Enoxaparin dosing and time to therapeutic outcomes. (a) Enoxaparin standard‐of‐care (SOC) dosing (mg/kg q12 h) received by infants versus what model‐informed precision dosing (MIPD) enoxaparin dose would have been. (b) Time to therapeutic anti‐Xa level by relative SOC versus MIPD dosing. (c) Infant final (titrated) enoxaparin dose (SOC treatment) versus initial MIPD dose. For (b), the lines displayed are a violin plot representing density plots where the width of the band corresponds to the frequency of the numerical value in that region of the plot.

The median time to therapeutic anti‐Xa was 39.7, 40.6, 40.1, and 75.0 h for the within 20%, 20%–40%, 40%–60%, and >60% above groups, respectively (Figure 1, panel b). These data show that the infants who were the most underdosed with SOC (as compared to MIPD‐dose) took the longest time to achieve therapeutic enoxaparin levels. The median (IQR) final SOC dose (after multiple adjustments) was 2.1 mg/kg (0.9, 4.4) and was similar to the recommended initial MIPD dose of 2.1 mg/kg (1.2, 3.0). The comparison included 154 paired observations (Figure 1, Panel c), with no significant difference between the two distributions (p = 0.202). This confirms that using MIPD from the start of enoxaparin anticoagulation therapy can lead to therapeutic anti‐Xa levels without the need for multiple up‐titrations.

DISCUSSION

Model‐informed precision dosing is a novel approach that has been used successfully in neonates to improve target attainment for anti‐infective drugs, such as vancomycin ¹⁵ and ganciclovir. ¹⁶ However, logistical barriers and a lack of clearly defined exposure targets have limited widespread adoption of this technique in other therapeutic areas. ¹⁷ In the NICU, for example, weight‐based dosing has been used for decades, despite evidence that body size alone is a poor descriptor of inter‐subject variability and can lead to variable, sometimes negative, outcomes.

Weight‐based dosing assumes that weight‐related changes in CL and Vd are exactly proportional, despite evidence to the contrary from many pediatric PK studies. While allometric scaling of PK parameters is both a common, and useful method of explaining inter‐subject variability, covariate analyses should not stop with body size. For example, in the PK model used in this study CL and Vd are affected by one or more of the following: PMA, serum creatinine, hematocrit, and platelet count. These covariates, in addition to body weight, play a significant role in determining drug exposure and should be used to optimize enoxaparin dosing.

This study used retrospective patient data from NICU patients <44 weeks PMA to assess whether MIPD recommendations would provide a better starting point for enoxaparin therapy. We found that 43% of all infants in our cohort would have received a starting dose >20% above the SOC if the MIPD tool were used. This relative increase in starting dose aligns with prior publications in which neonates required multiple dose titrations to achieve therapeutic anti‐Xa levels. This prolonged time to achieve doses within the therapeutic range would have corresponded to multiple, painful, blood draws, and an increased duration of therapy during which the infant's thrombus was inadequately treated.

The range of enoxaparin doses received as SOC in this cohort confirms what has been published in other neonatal cohorts. In a review of multiple studies reporting enoxaparin dosing, the mean maintenance dose ranged from 1.48 to 2.27 mg/kg q12 h. ¹⁸ In a retrospective cohort of Canadian neonates, children less than 3 months received a median enoxaparin dose of 1.83 mg/kg (range 1.71–2.14), and those who received an increased initial enoxaparin dose had faster attainment of therapeutic anti‐factor Xa levels requiring significantly fewer venipunctures. In preterm infants, doses of 2.06 ± 0.61 mg/kg were required for therapeutic anti‐Xa attainement. ¹⁹

Although we did not relate time to therapeutic dose with clinical outcomes in our cohort, there is literature to suggest that time to therapeutic level is associated with outcomes. In a study published in 2021 on a different LMWH (nadroparin) in children, ²⁰ it was found that six neonates had recurrent venous thrombo‐emboli (VTE). Five of these six neonates had subtherapeutic anti‐Xa (<0.5 IU/mL), which might reflect the importance of achieving therapeutic levels as soon as possible. In a study of pediatric enoxaparin dosing protocols, ⁹ successful resolution of thrombus was directly associated with obtaining a therapeutic anti‐Xa concentration upon the first measurement. Specifically, for subtherapeutic, therapeutic, and supratherapeutic FIRST anti‐Xa values, thrombus resolution occurred in 50%, 60%, and 73% of patients, respectively. This trend was consistent in <60 days and >60 days cohorts.

Interestingly, infants in our cohort with the lowest Hct were the most underdosed with SOC compared with the MIPD‐dose. The physiologic reason for this is unknown. During model development, hematocrit was found to be a significant covariate on clearance, similar to the work reported by Moffett et al. ¹³ They hypothesized that low sample volume leading to an increase in hematocrit may have caused this covariate to appear as a significant covariate on clearance. The result being that a unit decrease in hematocrit compared with the reference value (32.9%) corresponds to a 1% increase in enoxaparin clearance. Thus, during the dose optimization process, Lyv recommended a higher starting dose compared with SOC.

When considering MIPD of enoxaparin in neonates, the important topics of efficacy and safety must be considered. While MIPD may improve time to therapeutic anti‐Xa by increasing starting dose, it may also lead to an elevated anti‐Xa level and increase the risk of bleeding adverse events. To our knowledge, there is no widely accepted anti‐Xa threshold that corresponds to a dramatic increase in bleed risk outside of the clinically accepted therapeutic range of 0.5–1 IU/mL, used to titrate dosing to desired trough level., that the bleeding risk may be magnified in very early gestation infants who are within the first week of life and at risk for intraventricular hemorrhage (IVH). Studies of enoxaparin dosing show low incidence of bleeding, even with doses up to 3 mg/kg q12h ⁷ , ⁸ , ²¹ ; however, a prospective evaluation of MIPD for enoxaparin should include robust pharmacovigilance and highlight the associated risks and benefits as part of the consent, data collection and reporting process.

This study compared hospital‐administered SOC dosing to post hoc MIPD recommendations and found clear evidence that the latter may help optimize enoxaparin dosing in the NICU. The range of SOC doses observed in this study and the range of MIPD doses will be used to establish dosing bounds for an NIH‐funded prospective feasibility trial of enoxaparin MIPD in the NICU at The Hospital for Sick Children. The MIPD tool will be embedded into the local EPIC EHR and prescribers will access it in real‐time when prescribing the first dose of enoxaparin.

In summary, the results of this study suggest a higher starting dose of enoxaparin is needed when patient‐specific factors known to affect enoxaparin PK are taken into consideration. This finding corroborates ongoing clinical observations and published literature. With a paradigm shift toward MIPD in drug prescribing, we hope to decrease variability in drug exposure and achieve improved therapeutic outcomes for patients. The clinical feasibility and benefit of this approach will be tested in prospective feasibility trial which will have Health Canada oversight and will include close monitoring for potential adverse events.

AUTHOR CONTRIBUTIONS

HB, CS, LB, VI, and TL wrote the manuscript. HB, VI, and TL designed the research. HB, CS, and TL performed the research and analyzed the data. HB and VI contributed to the new analytical tools.

FUNDING INFORMATION

This study was funded by the NICHD MPRINT Opportunity Pool (5P30HD106451–02) and the SickKids Sellers Endowed Chair in Pharmacology and Pharmacogenetics.

CONFLICT OF INTEREST STATEMENT

The authors Vijay Ivaturi and Haden Bunn work at Pumas‐AI, Inc. All other authors declared no competing interests for this work.

Supporting information

Table S1.

Bunn H, Schentag C, Brandão LR, Ivaturi V, Lewis T. Pharmacokinetic model‐guided enoxaparin dosing in the Neonatal ICU: Retrospective cohort study to plan for prospective feasibility trial. Clin Transl Sci. 2024;17:e70040. doi: 10.1111/cts.70040