Model‐Informed Dose Optimization of Anakinra in Preterm Neonates Implications for Safe and Effective Dosing

Abstract

Anakinra, an interleukin‐1 receptor antagonist, shows promise for reducing inflammatory complications of premature birth. However, the pharmacokinetics (PK) of anakinra in neonates has not been characterized, limiting evidence‐based dosing. This study aimed to develop a population PK model for anakinra to guide dosing in preterm neonates. Data from 25 preterm neonates (median gestational age: 26.3 weeks) from a Phase I/IIa trial of intravenous anakinra (Anakinra Pilot, NCT05280340) were analyzed, with 4–8 samples collected per neonate across 0–20 days postnatal age (PNA). A population PK model was developed and dosing simulations were performed incorporating a pharmacodynamic (PD) direct effect model (I_max, IC₅₀) and target exposure (AUC_0‐24h). The data were best described by a one‐compartment model with linear elimination and a baseline component for endogenous IL‐1Ra. Allometric scaling was applied to clearance (CL) and volume of distribution (V), with a maturation function using PNA on CL to account for early age‐related pharmacokinetic changes. Estimated CL and V were 3.97 L/h/70 kg (0.15 L/h/0.9 kg) and 28.9 L/70 kg (0.37 L/h/0.9 kg). At birth, CL was ~45% of the maximal early postnatal value, achieving ~100% by day 10 PNA. Simulations suggest that an optimal intravenous dosing regimen of 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly from PNA days 2–3 and 1.15 mg/kg 12 hourly day 4 onward. This model delivers the first critical understanding of small protein PK in premature neonates, establishing a foundation for evidence‐based dosing strategies to guide future prospective clinical trials.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Anakinra is widely used in inflammation‐related diseases from infants to adults with emerging indications for use of anakinra in the past 1–2 years. However, its disposition in preterm neonates is unknown, particularly with respect to physiological factors influencing the renal clearance of small therapeutic proteins.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This study characterized the pharmacokinetics of anakinra in preterm neonates through the development of a robust population pharmacokinetic model and optimized dosing regimens informed by the half‐maximal inhibitory concentration and target exposure, thereby supporting an evidence‐based dosing strategy for anakinra in future prospective neonatal clinical studies.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study provides the first robust population pharmacokinetic model of anakinra across early life, incorporating the endogenous IL‐1 receptor antagonist and physiological maturation of clearance. It demonstrates that renal clearance in preterm neonates increases dynamically within the first 21 days after birth, underscoring the need for early age‐dependent dose adjustments to ensure effective and safe drug exposure in this population.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This study is the first to characterize the early maturation of renal clearance of small therapeutic proteins, providing pivotal insights into the elimination of therapeutic proteins via the renal pathway. The population pharmacokinetic model developed provides a robust framework for precision dosing, addressing a long‐standing gap in neonatal pharmacotherapy and supporting the advancement of tailored, evidence‐based treatment strategies in clinical and translational science.

In 2020, an estimated 13.4 million babies were born preterm worldwide. ¹ Preterm birth complications remain the leading cause of death in children under the age of five, accounting for 900,000 deaths annually. ² Survivors are exposed to high rates of inflammation‐related morbidities including cardiopulmonary bronchopulmonary dysplasia (BPD), its severest complication, pulmonary hypertension associated with BPD (BPD‐PH), intestinal necrotizing enterocolitis (NEC), and cerebral diffuse white matter injury (dWMI). ³ , ⁴ , ⁵ The incidence of these conditions increases dramatically if birth occurs before a gestational age (GA) of 32 weeks and is linked to long‐term cardiopulmonary disease and neurodevelopmental impairment. ³ , ⁴ , ⁵ Thus, prevention and treatment of BPD, BPD‐PH, NEC, and dWMI are crucial to improving neonatal health outcomes. Currently, however, there is no safe and effective therapy.

Pre‐clinical studies in rodents and sheep ⁶ , ⁷ , ⁸ , ⁹ have shown that increased abundance of interleukin‐1 (IL‐1), a potent pro‐inflammatory immune mediator, is associated with inflammatory damage to the lung, brain, eye and gut in early life. In vivo treatment with anakinra, a recombinant human IL‐1 receptor antagonist (IL‐1Ra) that competitively inhibits IL‐1α and IL‐1β binding to IL‐1R1, ¹⁰ , ¹¹ prevented the development of BPD, BPD‐PH and dWMI. ⁷ , ⁸ , ¹² With over 20 years of clinical use as an effective anti‐inflammatory therapy with a favorable safety profile, anakinra has emerged a promising candidate for use in preterm neonates to prevent or treat inflammation‐driven morbidities.

Anakinra is typically administered subcutaneously with a high bioavailability ranging from 80.2 to 91.9%. ¹³ As a small therapeutic protein (~17 kDa), anakinra undergoes free glomerular filtration followed by extensive proximal tubular uptake and catabolism, with only 3.2–9.1% recovered in urine. ¹⁴ This “renal clearance” (glomerular filtration then proximal tubular catabolism) accounts for around 80% of anakinra elimination, ¹⁵ with plasma clearance reduced by 70–75% in severe renal insufficiency and end‐stage renal disease. ¹⁴ However, clinical PK data for anakinra in pediatrics younger than 2 years is limited, while in preterm and term neonates is non‐existent. Although several population pharmacokinetic (PPK) models have been published, these models predominantly pertain to infants, children and adults following subcutaneous (SC) or intravenous (IV) administration. ¹⁴ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ Such models cannot be directly extrapolated to preterm neonates, who undergo rapid physiological changes in organ size, perfusion and function that significantly alter drug disposition. ²⁰ , ²¹ , ²² , ²³ , ²⁴ As a result, the PK profile of anakinra in premature neonates remains a gap in current pharmacological knowledge, particularly in light of the growing therapeutic promise of anakinra for this population. ⁸ , ²⁵ , ²⁶ , ²⁷

Anakinra Pilot (AP), a Phase I/IIa clinical trial that evaluated the safety, feasibility, and pharmacokinetics of anakinra in preterm neonates completed recruitment in 2024. ²⁶ The aim of the present study was to (i) develop a PPK model of anakinra in premature neonates during the 21‐day therapy period, to characterize pharmacokinetics associated with unique early postnatal physiological changes and (ii) propose model‐informed optimized dosing strategies for future prospective clinical studies of anakinra as a treatment for inflammation of prematurity.

METHODS

Study and data included

The data for the PPK analyses were collected from the Anakinra Pilot clinical study (AP; NCT05280340), an investigator‐initiated, single‐arm, safety and feasibility Phase I/IIa trial in preterm neonates conducted at Monash Children's Hospital, Melbourne, Australia and Starship Children's Hospital, Auckland, New Zealand. The protocol has been reported. ²⁶ In brief, enrolled neonates born between 24 weeks, 0 days (24⁰) and 27⁶ weeks of GA received anakinra (KINERET®, Swedish Orphan Biovitrum Pty Ltd, Macquarie Park, Australia) IV over the first 21 days after birth in addition to standard care. Enrolment took place between April 7, 2022 and February 21, 2024. In stage 1, 9 neonates received 1.0 mg/kg anakinra every 48 h, while in stage 2, 16 neonates received 0.8 mg/kg daily anakinra.

Neonates were included in the analysis if they received at least one anakinra dose and had at least one measurable anakinra concentration. Due to the known challenges of obtaining sufficient samples and volumes to quantify analytes and determine PK in this population, a sparse PK sampling approach was undertaken with blood collected pre‐dose, at hours 12 and 21–24 post‐dose on days 1, 3, 7, 14, 21, and 0.5–6 hours post‐dose on days 1 and 21.

The concentrations of anakinra were quantified via bead‐based AlphaLISA immunoassay [Revvity, Waltham, MA]. This assay does not differentiate between the endogenous IL‐1Ra and exogenous recombinant IL‐1Ra (anakinra). Thus, all measured concentrations reflect combined endogenous and exogenous IL‐1Ra concentrations. The lower limit of quantification (LLOQ) of the assay was 1.3 pg/mL, with an inter‐ and intra‐assay coefficient of variation of 26% and 7%, respectively (reported from the assay's data sheet).

Population PK model development

The anakinra population pharmacokinetic model was developed using nonlinear mixed‐effects modeling implemented in NONMEM version 7.5.1 (ICON Development Solutions, Ellicott City, MD) and Wings for NONMEM version 751. Parameters were estimated using a first‐order conditional estimation method with interaction.

One‐ and two‐compartment models were evaluated as potential structural models, with the endogenous IL‐1Ra concentration included as a baseline. A baseline and a turnover model were evaluated to characterize both exogenous and endogenous IL‐1Ra. In the baseline model, the baseline of endogenous IL‐1Ra was directly added to the predicted concentrations, while the turnover model incorporated a zero‐order production rate (kin) and a first‐order elimination of endogenous IL‐1Ra, sharing the same elimination process as anakinra.

Following base model development, theory‐based allometric scaling was incorporated a priori to standardize PK parameters with fixed allometric exponents (0.75 on CL and 1 on V), ²⁸ using total body mass (TBM) or normal fat mass (NFM). NFM was calculated from TBM and fat free mass (FFM) with a fraction of fat (F_fat), which describes both the mass and body composition effects on PK parameters (Eq. 1 ). ²⁹ , ³⁰ , ³¹ A standard value for NFM (NFM_std) is a male with a TBM of 70 kg, an FFM of 56.1 kg and a height of 1.76 m (Eq. 2 ). ³²

$$\mathrm{NFM}=\mathrm{FFM}+{\mathrm{F}}_{\mathrm{fat}}\times \left(\mathrm{TBM}-\mathrm{FFM}\right)$$ (1)

$${\mathrm{NFM}}_{\mathrm{std}}=56.1+{\mathrm{F}}_{\mathrm{fat}}\times \left(70-56.1\right)$$ (2)

Due to the limited number of subjects, along with well‐developed population PK models for anakinra in other cohorts, only covariates with known biological plausibility to pharmacokinetic parameters were considered, including age‐related maturation of clearance, renal function and sex. Covariates were evaluated by a stepwise forward inclusion (P < 0.05) and backward exclusion (P < 0.01) process. Missing weight and height data were handled using the standard last observation carried forward (LOCF) method. ³³

Premature neonates have immature kidneys with decreased glomerular filtration, resorption and tubular secretion capacities. The glomerular filtration rate (GFR) increases rapidly during the first weeks of life, ²⁴ , ³⁴ reflecting major changes in blood flow, renal vascular resistance and oxygenation during the transition to extrauterine life. GFR continues to mature over the first 2 years of life, with renal function maturation beyond the initial postnatal weeks more closely linked with postmenstrual age (PMA), rather than postnatal age (PNA). ²⁹ The two key maturation models evaluated were F_mat,PNA based on PNA (Eq. 3 ) and F_mat,PMA based on PMA (Eq. 4 ) to describe the maturation impact on CL. PNA at birth is defined as zero, representing the time elapsed since delivery, while PMA refers to the sum of PNA and GA, capturing the total time from the last menstrual period to the current postnatal time point.

$${\mathrm{F}}_{\mathrm{mat},\mathrm{PNA}}=1-\mathrm{FCL}\times {\mathrm{e}}^{\frac{-\ln 2\times \text{PNAD}}{{\text{PNAT}}_{50}}}$$ (3)

$${\mathrm{F}}_{\mathrm{mat},\mathrm{PMA}}=\frac{1}{1+{\left(\frac{\mathrm{PMA}}{{\text{PMAT}}_{50}}\right)}^{-\text{Hill}}}$$ (4)

Furthermore, assessment of changing renal function via the interpretation of measured serum creatinine in preterm neonates remains challenging, as serum creatinine is generally reflective of maternally transferred creatinine in the first 7–14 days of life. ³⁵ , ³⁶ , ³⁷ As such, no further adjustments using serum creatinine concentrations were used to describe changing renal function beyond maturation.

Model evaluation

Model evaluation was based on the objective function value (OFV), goodness of fit (GOF), and prediction‐corrected visual predictive checks (pcVPCs). A reduction in the OFV of 3.84 was considered significant for two nested models (χ ² < 0.05 for 1 degree of freedom, df). Simulations (n = 1,000) were performed using the final model and parameters to produce the pcVPC. Nonparametric bootstrapping with 1,000 replicates was performed to assess the precision of parameter estimates.

Simulations

At present, the pharmacokinetics/pharmacodynamics (PK/PD) metrics determining the therapeutic response of anakinra in this population are yet to be elucidated. Therefore, dose optimization was guided by utilizing a two‐step process. Step 1, the IC₅₀ (mean IC₅₀: 0.04 mg/L) derived from pre‐clinical and clinical literature ¹⁸ , ³⁷ , ³⁸ was used to produce concentration‐inhibition profiles of anakinra for different dosing regimens, e.g., doses and frequencies. In Step 2, dosing regimens that achieved suitable concentration‐inhibition profiles were evaluated against the target exposure range, i.e., an area under the plasma concentration‐time curve (AUC_0‐24h), for efficacy and safety. Simulations were performed using virtual populations (n = 1,000) with demographic data randomly sampled from neonates in the AP study. See Figure S1 for the summary of the simulation method. All simulations were implemented using the “mrgsolve” package in R. R version 4.4.1 (http://www.r‐project.org) was used for visualization, statistical summaries, and simulations.

Step 1

In detail, to evaluate the relationship between dose, concentration and IC₅₀, a simple I_max or direct effect model (Eq. 5 ) was utilized due to the lack of biomarker data in preterm neonates. In this model, I_max is the maximum % IL‐1Ra inhibitory effect assumed to be 1, C is the anakinra concentration predicted by the final covariate model and IC₅₀, the concentration that produces 50% of the maximum inhibition, was assigned a reference value.

$$\begin{array}{c}\mathrm{IL}-1\mathrm{Ra}\text{inhibition}\left(\%\right)=\frac{{\mathrm{I}}_{\max }\times \mathrm{C}}{{\mathrm{IC}}_{50}+\mathrm{C}},{\mathrm{I}}_{\max }=1, {\mathrm{IC}}_{50}=0.04\mathrm{mg}/\mathrm{L}\hfill \end{array}$$ (5)

The evaluation criteria for success in these simulations were dosing regimen(s) that provided consistent ≥ 50% inhibitions for at least half of the treatment population.

Step 2

For dosing regimens achieving the inhibition criteria in Step 1, a further evaluation against the target exposure range (AUC_0‐24h) for efficacy and safety was undertaken. For efficacy, an AUC_0‐24h of 9.6 mg/L·h was selected to inform dosing design for the AP study, ²⁶ being the exposure sufficient to suppress systemic inflammation ¹⁸ and improve clinical outcomes ³⁶ in children with auto‐immune and auto‐inflammatory diseases. The safety exposure was defined from both pre‐clinical and clinical safety data. In a pre‐clinical toxicokinetic anakinra study cited in the US Food and Drug Administration (FDA) pharmacology review, ³⁹ an SC dose of 10 mg/kg/day with an AUC_0‐∞ of 57.15 mg/L·h on day 28 was identified as the maximum tolerated dose in Rhesus monkeys, based on the minimal injection‐site inflammation. ³⁹ Clinically, adults with end‐stage kidney disease characterized by the lowest anakinra clearance who received anakinra 100 mg daily had AUC_0‐24h values of approximately 30–40 mg/L·h without any serious adverse events. ¹⁴ Given these pre‐clinical and clinical exposure‐safety relationships, the conservative target therapeutic window of anakinra was set as an AUC_0‐24h of 9.6 to 30 mg/L·h.

Any dosing regimens outside the target therapeutic window were returned to Step 1 to further modify the dose and then returned back to Step 2 until a pragmatic final dosing regimen was identified (Figure S1 ).

Ethics statement

The final protocol and informed consent forms were approved by Monash Health Human Research Ethics Committees (RES‐21‐0000‐681A) and the Northern B Health and Disability Ethics Committee (NZ) 2023 FULL 15373.

RESULTS

Demographics

From the 25 preterm neonates who received at least one dose of anakinra in AP, 178 total blood samples were taken over the course of the study, with a median of 8 samples per neonate (range 4–8). The demographic characteristics are summarized in Table 1 . Minimal changes in TBM were observed during the 21‐day therapy period. The distributions of sex, age, and size were comparable between stage 1 and stage 2.

Table 1.

Demographic characteristics

	Total population	Stage 1 (1.0 mg/kg, 48 hourly)	Stage 2 (0.8 mg/kg, 24 hourly)
Number of subjects (n)	25	9	16
Sex (n)
Female	15 (60.0%)	5 (55.6%)	10 (62.5%)
Male	10 (40.0%)	4 (44.4%)	6 (37.5%)
Age
Gestational age (weeks)	26.3 (24.0, 27.9)	27.3 (24.0, 27.9)	26.3 (24.3, 27.7)
Postnatal age (days)a	0–20	0–20	0–20
Baseline size
Total body mass (kg)	0.90 (0.54, 1.30)	0.77 (0.56, 1.25)	0.90 (0.54, 1.30)
Length (cm)	35.0 (28.0, 41.0)	33.0 (30.0, 41.0)	35.0 (28.0, 40.0)
Fat‐free mass (kg)	0.86 (0.51, 1.24)	0.73 (0.53, 1.19)	0.86 (0.51, 1.24)
Size during therapyb
Total body mass (kg)	0.91 (0.53, 1.60)	0.79 (0.53, 1.54)	0.91 (0.54, 1.60)
Length (cm)	35.0 (28.0, 42.0)	35.0 (30.0, 42.0)	35.0 (28.0, 41.0)
Fat‐free mass (kg)	0.86 (0.50, 1.52)	0.75 (0.50, 1.46)	0.87 (0.51, 1.52)
Baseline of endogenous IL‐1Ra (pg/mL)	8,754.9 (294.5, 87,099.7)	8,727.6 (294.5, 32,885.3)	8,754.9 (456.5, 87,099.7)
Study site (n)
Monash Children's Hospital	19 (76%)	9 (100%)	10 (62.5%)
Starship Children's Hospital	6 (24%)	0 (0%)	6 (37.5%)

Demographic and clinical characteristics in 25 preterm neonates. Data are shown as a number of subjects (percentage) for categorical variables or median (min, max) for continuous variables.

^a All preterm neonates received anakinra starting at birth and continued on treatment for 20 days.

^b Data are presented as median (min, max) over a 21‐day period for continuous variables.

Population PK analysis

For base models, a one‐compartment model with linear elimination and proportional error best described the data. Importantly, the population estimates for CL from stage 1 and stage 2 were similar, indicating that CL is dose‐independent within the range administered. Furthermore, the estimated inter‐individual variability (IIV) of V approached zero, and its inclusion in the model did not result in any change in the OFV. While the population estimate for V based on standard errors was estimable, a deterministic identifiability analysis, as part of the model instability assessment, ⁴⁰ further indicated that IIV of V was not identifiable, with a high relative standard error (RSE) of 246%. This low identifiability of IIV on V is likely due to the limited number of blood samples taken during the early post‐infusion period in stage 2. A two‐compartment model was also evaluated. However, it did not improve model fit, as evidenced by a negligible change in OFV compared to the one‐compartment model (Summary of model selections, Table S1 ).

We evaluated the mixed endogenous and exogenous IL‐1Ra, using either a baseline or a turnover model. Compared to the baseline model, the turnover model exhibited instability with the RSE for CL and V close to 0. Therefore, a one‐compartment model incorporating a baseline model was selected as the base model.

In terms of allometric scaling, the low PMA in our study population limited the ability to obtain realistic FFM estimates, thus TBM was used for the body size scaling rather than NFM. Various metrics for maturation were explored (Table S1 ), with the PNA model (Eq. 3 ) best characterizing the maturation effect on anakinra CL. The final models for CL and V are shown in Eqs. 6 and 7 .

$$\mathrm{CL}=3.97\times {\left(\frac{\text{Weight}}{70}\right)}^{0.75}\times \left(1-0.554\times {\mathrm{e}}^{\frac{-\ln 2\times \text{PNAD}}{1.8}}\right)$$ (6)

$$\mathrm{V}=28.9\times {\left(\frac{\text{Weight}}{70}\right)}^1$$ (7)

The final model parameter estimates and bootstrap results are summarized in Table 2 . Of the 1,000 bootstrap runs, 954 (95.4%) ran successfully, with the final parameter estimates being close to the bootstrap median values within the 95% confidence intervals (CIs). The estimate of endogenous IL‐1Ra from the final model was 6,640 pg/mL with a high IIV of 95%, which was comparable to the wide variability reported in preterm neonates, with medians ranging from 310.9 to 7,312.1 pg/mL (IQR: 149.3–16,131.1). ⁴¹

Table 2.

Parameter estimates of anakinra pharmacokinetics

Parameter	Final model		Bootstrap resultsa
	Estimate	RSE (%)	Median	95% CI
CL, L/h/70 kg	3.97	14.3	4.0	2.75–6.78
V, L/70 kg	28.9	15.7	27.9	20.1–38.9
BSL, pg/mL	6,640	27.0	6,905	3,668–12,317
WT on CL	0.75 (Fixed)
WT on V	1 (Fixed)
FCL	0.55	13.4	0.58	0.34–0.87
PNAT50, day	1.80	24.3	1.76	1.01–20.8
IIV CL	0.28	18.7	0.27	0.14–0.42
IIV BSL	0.95	25.7	0.86	0.40–1.52
Proportional error	0.77	8.3	0.74	0.65–0.84

BSL, baseline of endogenous IL‐1Ra; CI, confidence interval; CL, clearance; FCL, fractional difference in CL at birth; IIV, inter‐individual variability; PNAT₅₀, half time required to achieve 50% of the postnatal change; RSE, relative standard error; V, volume of distribution.

^a Bootstrap based on 954 successful runs out of 1,000.

The pcVPC plot for the final population PK model (Figure 1 ) indicates that the observed concentrations were well described by the model predictions. Additionally, the GOF plots demonstrated adequate performance of the final model (Figure S2 ).

Figure 1.

Prediction‐corrected visual predictive check (pcVPC) for the final pharmacokinetic model with a log‐scale y‐axis. The 20%, median, and 80% percentiles of the distribution of the observations are in black, and predictions are in red and blue. The 95% confidence intervals for the prediction percentiles are indicated by the shaded areas.

The fractional changes in CL predicted from the model (Figure S3 ) show that anakinra clearance increased with increasing PNA and gradually approached the maximal predicted clearance around PNA day 10. At birth, the clearance of anakinra was 44.6% of the maximal postnatal age‐associated maturation from the start CL value of 1.77 L/h/70 kg to the value of 3.97 L/h/70 kg at the last day of treatment. This rapid increase of CL could be driven by the increased renal blood flow, with a rapid increase in GFR in neonates during the first 2 weeks of life, as well as the tubular function postnatal development. ²⁹ , ³⁰

For exposure (AUC_0‐24) results, the AUC_0‐24 in stage 1 was lower (Figure 2 ) than the nominal target of 9.6 mg/L·h throughout the 21 days. This was expected due to the original study design ²⁶ i.e., the conservative dose and frequency selected for stage 1. In comparison, the anakinra exposure for stage 2 was on target with an AUC_0‐24 of 9.8 mg/L·h at birth, then declined rapidly and plateaued at ~4.5 mg/L·h by PNA day 10, which was ~54% lower than the target exposure. The reason for this observed change is the rapid anakinra clearance increase in the first 10 days of treatment (life) described above (Figure 1 ). On this basis, a revised dosing schedule is necessary to achieve 9.6 mg/L·h across the 21‐day period.

Figure 2.

Whisker plots of anakinra exposure stratified by stage. Red boxplots represent stage 1 (1.0 mg/kg, IV, 48 hourly) and teal boxplots represent stage 2 (0.8 mg/kg, IV, 24 hourly). The red dashed line represents the target exposure (9.6 mg/L·h) suppressing the inflammation.

Simulation results

Step 1 simulation results

Simulations of the concentration‐inhibition profiles when linked to the immediate effect PD model identified two key findings: (i) with 0.8 mg/kg 24 hourly, as per stage 2 (Figure 3 a ), the estimated inhibition rapidly dropped below the 50% IL‐1Ra inhibitory effect across the 24 hour dosing interval (Figure 3 b ), while (ii) the same total daily dose administered 12 hourly, i.e., 0.4 mg/kg every 12 hours, mitigated the peak‐to‐trough fluctuations, with trough concentrations (C_trough) exceeding the IC₅₀ target during the first 48 hours of therapy (Figure 3 c ) and achieving ≥ 50% inhibition until 60 hours post‐start of treatment (Figure 3 d ). However, beyond PNA day 3, both concentration and inhibition gradually declined (Figure S4 ). These results indicate the need for increasing doses over time in combination with twice‐daily dosing to account for the rapid change of anakinra clearance with increasing PNA.

Figure 3.

Simulated concentration‐inhibition profiles of anakinra vs time for IV administration following (a, b) 0.8 mg/kg, 24 hourly, or (c, d) 0.4 mg/kg, 12 hourly. The green dashed line represents the IC₅₀ of 0.04 mg/L. The red dashed line represents the 50% inhibition. The 95th prediction percentiles are indicated by the gray areas.

After adjusting doses to account for PNA‐related maturation of anakinra clearance, the dosing regimen (R1) of 0.4 mg/kg 12 hourly at birth and on PNA day 1, followed by 0.6 mg/kg 12 hourly on PNA day 2, 1.2 mg/kg 12 hourly on days 3 to 5 and 1.8 mg/kg 12 hourly from PNA days 6 to 20, resulted in sustained inhibition above 50% with median C_trough stabilizing at ~0.04 mg/L over 21 days (Figure S5 ).

Step 2 simulation results

AUC‐based exposure simulations from the R1 derived from Step 1 showed that approximately 50% of the population exhibited exposures below the target AUC_0‐24h of 9.6 mg/L·h on the first 3 days of life, while ~20% exceeded the safety exposure of 30 mg/L·h from day 6 onwards, potentially increasing the safety risks for neonates (Figure 4 a ). To balance efficacy and safety, further optimization was undertaken, resulting in a new dosing regimen (R2) of 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on PNA days 2 to 3 and 1.3 mg/kg 12 hourly from PNA days 4 to 20. Using this regimen, ~95% of the simulated population achieved exposure ≥ 9.6 mg/L·h, with 4% exceeding the safety limit (Figure 4 b ). To further reduce excessive exposure risk, the dose from PNA day 4 onward was then adjusted from 1.3 mg/kg to 1.15 mg/kg 12 hourly, yielding regimen R3. This dosing regimen resulted in a 10‐fold reduction in safety risk, with only a 5% decrease in target attainment (Figure 4 c ).

Figure 4.

Simulated exposure (AUC_0‐24h) for 1,000 premature neonates based on the final model. (a) Dosing regimen (R1): 0.4 mg/kg 12 hourly at birth and on PNA day 1, 0.6 mg/kg 12 hourly on PNA day 2, 1.2 mg/kg 12 hourly on PNA days 3 to 5 and 1.8 mg/kg 12 hourly from PNA days 6 to 20 anakinra, IV. (b) Dosing regimen (R2): 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on PNA days 2 to 3 and 1.3 mg/kg 12 hourly from PNA days 4 to 20 anakinra, IV. (c) Dosing regimen (R3): 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on PNA days 2 to 3 and 1.15 mg/kg 12 hourly from PNA days 4 to 20 anakinra, IV. The red dashed line indicates the target exposure of 9.6 mg/L·h, while the blue dashed line indicates the safety exposure of 30 mg/L·h.

Importantly, taking both R2 and R3 back to Step 1, the concentration‐inhibition profile demonstrated ≥ 50% inhibition in over 50% of the population with C_trough exceeding 0.04 mg/L in the first week under both regimens (Figure 5 ). During the second and third weeks, the proportion of individuals exhibiting 50% inhibition decreased slightly, with minimal differences between R2 and R3 (45% vs. 42%). Overall, by integrating concentration, inhibition, and AUC metrics, we identified an optimal and pragmatic dosing regimen (R3) that maximizes efficacy while minimizing safety risk under a conservative safety exposure in our target population.

Figure 5.

Simulated concentration‐inhibition profiles of anakinra for 3 weeks. (a) Concentration‐time course. (b) Inhibition‐time course. Dosing regimen (R2): 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on PNA days 2 to 3 and 1.3 mg/kg 12 hourly from PNA days 4 to 20 anakinra, IV. (c) Concentration‐time course. (d) Inhibition‐time course. Dosing regimen (R3): 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on PNA days 2 to 3 and 1.15 mg/kg 12 hourly from PNA days 4 to 20 anakinra, IV. The green dashed line represents the IC₅₀ of 0.04 mg/L. The red dashed line represents the 50% inhibition. The 95th prediction percentiles are indicated by the gray areas.

DISCUSSION

Over the past 1 to 2 years, interest in repurposing anakinra for high‐risk neonatal pathologies is growing, including BPD, ⁸ perinatal asphyxia, ²⁵ and other inflammatory conditions. ²⁷ Importantly, all these cases rely on empiric dosing, carrying a risk of either compromised efficacy or unexpected toxicity.

Additionally, a recent systematic review of anakinra clinical PK and PD ⁴² demonstrated that SC dosing in infants or older children (0.3–16.8 years) ranged from 1 to 11 mg/kg/day, whereas no IV administration data were identified in these pediatric populations, highlighting a notable lack of IV PK profiles in pediatrics. Most notably, no pharmacokinetic studies or population PK models were identified in the neonatal or premature neonatal populations.

The present study identified that a one‐compartment model with TBM and PNA effectively captured the PK profile of anakinra in premature neonates. The estimated CL per 70 kg was 3.97 L/h, which was approximately half that from adult studies with IV administration (range 7.3–10.8 L/h), while the V of 28.9 L/70 kg was 2–4 fold higher compared to adults (range 6.29–10.4 L) from the non‐compartment analysis results. ¹³

The model developed in this study incorporated age‐ and birth‐related changes in anakinra disposition, well accounting for the unique developmental ontogeny of preterm neonates compared with older children and adults, characterized by rapid changes in organ maturation. Furthermore, our model integrates a component describing endogenous IL‐1Ra concentrations, a feature that has not been included in previously published pediatric PPK models. ⁴²

The model‐informed covariate simulation platform used to design this Phase I/IIa study assumed a gradual postnatal maturation of anakinra CL. ²⁶ However, the raw data and subsequent analysis demonstrated that the early postnatal increase in CL was substantially steeper than these previously published maturation models. ²⁸ , ²⁹ , ⁴³ , ⁴⁴ As illustrated in Figure 2 , target anakinra exposure was achieved during the first day of therapy but was not sustained thereafter. To describe this “rapid maturation” a covariate model using postnatal age best described this phenomenon, which demonstrated that anakinra CL was approximately 50% of its maximal early postnatal maturation value at birth, followed by a rapid increase over the first 10 days of life. This is physiologically plausible given the rapid increases in renal blood flow and GFR after birth driven by decreasing renal vascular resistance and increasing cardiac output. ⁴⁵ To our knowledge, such early maturation of clearance has not been previously characterized for small proteins. Indeed, a similar time‐frame for this early change in clearance has been reported previously for gentamicin, a small molecule which is highly filtered (fu = 0.9), with a PNAT₅₀ of 1.7 days in neonates (GA 23.3–42.1 weeks), ⁴⁶ which further provides evidence to support it being increasing GFR for anakinra. Importantly, our model captures only the short‐term changes in anakinra CL during the first 21 days after birth in preterm neonates (24–28 weeks GA), whereas full GFR maturation is expected to continue throughout the first 2 years of life. ⁴³ , ⁴⁴

The final population PK model was highly effective in estimating the fixed estimates of CL and V (RSE < 20%). However, due to limited blood samples taken immediately post‐dose, the inter‐individual variation of V was challenging to estimate. This was expected, as the primary design of the AP study was to obtain accurate estimates of fixed and inter‐individual variation of CL, ⁴⁷ which controls overall exposure. This study yielded precise estimates of CL, including accurate quantification of the early maturation of CL in the immediate days post‐birth. Future studies with earlier sampling times, i.e., immediately post‐infusion, will help support more precise estimates of V.

Technically, NFM is an effective descriptor for body size, incorporating both total body mass and fat‐free mass. When applied for allometric scaling, the F_fat of CL and V was near 0, indicating that FFM represented the majority of the body size, consistent with the physiological features of preterm neonates. However, plausible FFM values could only be achieved when including individuals with higher PMA, making TBM a more practical descriptor for the current population. Furthermore, we have described maturation using PNA with a “half‐life of maturation” rather than a sigmoid PMA model or a combined PNA and PMA model. ²⁸ , ²⁹ , ⁴⁸ This is due to the limited PMA range (24.0 to 30.9 weeks) in our study population from recruitment into the trial to completion of the trial. A limitation from this is that the current PNA model may not be able to predict clearance across the entire GA period to complete maturation. However, until further data from a broader neonatal age are collected, the current model adequately describes CL and its maturation in the current population, i.e., 24–28 weeks' GA at birth.

For simulations, dosing optimization was informed comprehensively by a two‐step process, (i) a direct effect model incorporating IC₅₀ and (ii) target exposure (AUC_0‐24h) range for efficacy and safety. Our simulation results have demonstrated that the dose range (1.2–2.3 mg/kg/day) provides similar exposure across the 21‐day dosing period, i.e., AUCs to the dose range (1–2 mg/kg/day) given for neonates with deficiency of IL‐1Ra or treatment of neonatal‐onset multisystem inflammatory disease. ⁴⁹ However, the dosing regimen in the immediate neonatal period is more complex as it needs to account for the rapid maturation of clearance in the early postnatal period due to rapid increase in GFR. Furthermore, while certain doses in the optimized dosing regimen are higher than those in stage 1 and stage 2 could raise concerns of nonlinear elimination, they remain within the wider 1 to 10 mg/kg/day range shown to exhibit a linear elimination profile in a recently published systematic review. ⁴²

Given that the target exposure and IC₅₀ were derived from literature in non‐neonatal populations and have not been robustly validated, ⁴² the simulated dosing schedules serve as an exploratory, model‐informed foundation for future clinical trials rather than actionable bedside recommendations.

While the optimal dosing regimen (R3) achieved the target exposure (~10 mg/L·h) in 90% of the population whilst remaining exposure within clinically observed ranges, the proportion of individuals exhibiting ≥ 50% inhibition dropped during the second and third weeks, attributable to C_trough falling below the IC₅₀. These findings suggest that transitioning from the IV route to the SC route after the first or second week might be an alternative strategy to achieve ≥ 50% inhibition in over 50% of the population, due to delayed absorption from the injection site, causing a prolonged half‐life, thus increasing C_trough and maintaining the exposure. Further clinical studies are required to verify this approach.

One limitation of this study was the use of a static baseline model in the final estimates to describe endogenous IL‐1Ra, which did not account for the production and clearance kinetics. In addition, given the available data, the model was not able to characterize additional elimination pathways. Such additional pathways could be expected for anakinra, which is a therapeutic protein with potential non‐renal clearance such as lysosomal degradation in extrarenal tissues. Considering these limitations, further clinical trial data are required in the target population to refine the population model, including the use of a turnover model describing the endogenous IL‐1Ra vs. exogenously administered anakinra, and further exploration of the maturation effect on the distribution and the quantification of renal and non‐renal clearance.

A key limitation of the current study is the reliance on the assumption of PD target concentration, i.e., IC₅₀, similarity extrapolated from non‐neonatal data. Consequently, the direct relationship across the continuum of PK, PD, and responses remains to be confirmed. Thus, additional biomarker data reflecting PD metrics, such as C‐reactive protein, cytokines, and immune makers, alongside clinical efficacy and safety outcomes in preterm neonates are required to facilitate dose optimization in this population. We believe the current PPK model established could potentially be applied to other small therapeutic proteins that undergo glomerular filtration and proximal tubular catabolism. However, this will require further validation, given the fractional clearance of small therapeutic proteins at the glomerulus differs depending upon size (from freely filtered below 20 kDa, then progressively more limited until negligible filtration by 60–70 kDa). As for extrapolation of anakinra PK to neonates of higher GA at birth, there is already a sound basis, noting that our pre‐trial mechanistic model‐based extrapolation—using size and GFR maturation—was accurate in predicting anakinra clearance in the current cohort beyond the neonatal transition period in the initial week. This supports application of the same mechanism‐based clearance estimates in dosing term or mildly preterm neonates with anakinra. Nevertheless, noting the relatively small cohort of extremely preterm neonates (i.e., 24–28 weeks GA) on which the current model is based, further clinical study with external validation will increase confidence in such extrapolation.

In summary, we present the development of a population PK model for anakinra in preterm neonates, and evaluate it as fit for purpose. We recommend a dosing regimen of IV anakinra of 0.6 mg/kg 12 hourly at birth, 0.8 mg/kg 12 hourly on PNA day 1, 1.0 mg/kg 12 hourly on days 2 to 3 and 1.15 mg/kg 12 hourly from PNA days 4 to 20, to reach the target exposure of ~10 mg/L·h with at least half of the treatment population attaining ≥ 50% inhibition. This study presents the first robust PPK model for anakinra that comprehensively accounts for endogenous IL‐1Ra and physiological maturation in early life. Our findings underscore the potentially transformative role of pharmacometric modeling and mechanism‐based extrapolation and simulation in neonatal drug development. By enabling precise dose optimization within established safety margins, these techniques, long regarded as the gold standard in industry, address a critical gap in academic research. The demonstrated success in premature neonates highlights their potential to accelerate evidence‐based dosing strategies across pediatric therapeutics, aligning with global calls for broader adoption in investigator‐initiated trials.

FUNDING

This work was supported by an Australian Government Research Training Program Scholarship to J.L. and J.C.O., a Cerebral Palsy Alliance PhD Scholarship to E.G. and Mid Career Fellowship to R.G. (ERG02123), a Catalyst Seeding grant by the Royal Society of New Zealand (CSG‐UOA2209) to G.T., NHMRC Investigator Grants APP1175634 to C.T.R., APP1173731 to G.R.P., APP 2008793 to A.M. and APP1173584 and APP 2033196 to C.A.N.‐P., the Fielding Fellowship 2017 to M.F.N. A.J.G. is supported by the Health Research Council of New Zealand (Grant 22/559). This study was supported by a Catalyst: Seeding General (NZ) grant (22‐UOA‐071‐CSG) to G.T.

CONFLICTS OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

D.K.M., E.G., G.T., R.G., R.A., A.J.G., P.J.B., A.V., R.W.H., C.A.N.‐P., M.F.N. and C.M.J.K. designed the research; E.G., G.T., M.C., R.A., I.R., J.C.O., S.K.D., S.C.F., A.Maj., G.R.P., M.F., P.P., A.Mal., C.T.R., S.H.S., A.S., K.T. and M.F.N. performed the research; J.L., D.K.M., S.X.C. and C.M.J.K. analyzed the data; J.L., D.K.M., M.C., M.F.N. and C.M.J.K. wrote the manuscript.

Supporting information

Data S1