Nick HOLFORD (2),Xiaoyan YANG (1),(2),Zhimei JIANG (3),Hongxin SHEN (3),Jing SHI (1),Xianxiao SHU (1),Yi HUANG (1),Jing ZHAO (1),Jun TANG (1),Dezhi MU (1)
1. Department of Pediatrics, West China Second University Hospital, Sichuan University Chengdu 610041, China; 2. Dept Pharmacology & Clinical Pharmacology, University of Auckland, Auckland 1010, New Zealand; 3. Department of Pharmacology, West China Second University Hospital, Sichuan University Chengdu 610041, China
Objective: Apnea of prematurity (AOP) is defined as an attack of apnea for at least 20 seconds, with bradycardia and cyanosis. It is a common phenomenon in the neonatal intensive care unit. Caffeine is used to suppress or to prevent AOP attacks. The objective of this study is to develop a rational method to predict the dose of caffeine by using the effects to determine the target concentration and provide support for clinicians to determine the dose required to reach the target concentration.
Methods: Caffeine concentrations were measured prospectively in a study of its pharmacokinetics (PK) in premature neonates with AOP. The PK was described with a mixed effect one compartment first order elimination model. The baseline hazard was described by a Gompertz distribution. The pharmacodynamics (PD) caffeine were modelled using a sigmoid Emax model directly on the hazard of an AOP attack. PD covariates included adenosine receptor genotype and various methods of spontaneous breathing ventilation support (LFNC, HFNC, CPAP, BIPAP, BNCPAP) or mechanical ventilation (CMV, HOV). The hazard was set to 0 during mechanical ventilation. The pharmacodynamic (PD) and hazard model was developed using NONMEM 7.41. Visual predictive checks (VPC) was performed for model evaluation.
Results: Caffeine concentrations (1004 measurements) with fixed protocol dosing in 222 premature neonates had a wide distribution (median 8.7 mg/L, 90% within 2.9 and18.1 mg/L after 2 days of treatment). Both fixed (size and post-menstrual age) and random effects on between subject variability in clearance were important. Between occasion variability in clearance was small. The bootstrap hazard and PD parameter estimates are shown in Table 1.
Table 1 Parameters of AOP hazard and pharmacodynamics of caffeine
|
Parameter |
Description |
Units |
Bootstrap average |
|
L_LZ |
Baseline hazard |
1/h |
0.00496 |
|
B_GOM |
Gompertz hazard |
. |
-0.00167 |
|
CAF_EMAX |
Emax for caffeine effect on hazard |
1/h |
0.492 |
|
CAF_C50 |
C50 for caffeine effect on hazard |
mg/L |
1.80 |
|
CAF_HILL |
Hill exponent for conjugate effect on hazard |
. |
2.64 |
The PD parameters were used to simulate the link between caffeine concentration and reduction of AOP hazard. The C50 suggests that a target concentration of 5 mg/L would achieve close to maximum achievable benefit and this was chosen as the target concentration. A Bayesian dose forecasting procedure was developed using NextDose (www.nextdose.org) which clinicians can use to calculate initial doses and then use measured caffeine concentrations to further individualize treatment to achieve the target concentration.
Conclusion: Caffeine approaches its maximum effect and suppresses AOP attacks by about 50% at a concentration of 5 mg/L. The small between occasion variability and wide variation in concentrations with fixed dosing mean that target concentration intervention can be expected to improve outcomes. NextDose provides an easily available tool for clinicians to apply clinical pharmacology to improve patient care.
Funding: The National Key Development Program of Clinical Specialist (neonatologist) (1311200003303); the National Science Foundation of China (81330016); the Foundation of Health and Family Planning Commission of Sichuan Province (150104, 150107).
Reference: PAGE 28 (2019) Abstr 9027 [www.page-meeting.org/?abstract=9027]
Poster: Clinical Applications