Modelling of delivery kinetics of gentamicin administered through umbilical long lines

Introduction:Gentamicin is commonly used in the NICU setting and is often administered via long lines, such as umbilical venous catheters, which increases variability in the rate of administration. This variability is not taken into account when considering drug delivery in neonates and recommendations are extrapolated from adult data. We aimed to model drug delivery parameters and use this model to simulate intravenous gentamicin administered via an umbilical venous catheter.

Methods: Data was modelled from infusion simulations of gentamicin delivery using umbilical venous catheters with a background flow rate of 0.5ml/h (carried out by Anita Lala in 2016). Different combinations of dose (2mg, 5mg) were given by bolus injection over 3-5 minutes, followed by a normal saline flush (1ml, 2ml). The amount of gentamicin at the end of the line was measured at 5 minute intervals for an hour via high pressure liquid chromatography.

Phoenix Certara (version 8.1) was used to model the data previously described. An extravascular model with the clearance removed was used to predict the parameters absorption constant (Ka), time lag (Tlag), and “bioavailability” (F). F was used to enable an estimate of the variability in dose administered. Doses used in the model were the actual doses measured in the study rather than the prescribed dose. Different error models were used to ascertain which best described the data.

Results:An extravascular one compartment model with first order absorption and additive error best described the data. The estimates for this model for a dose of 2mg and a 1ml flush were Ka 0.34L/min, Tlag 1.28min, F 0.97, and stdev of 0.14. The -2LL was -41.8, the epsshrinkage was 0.08, and nshrinkage for ka was 0.9999, Tlag 0.25, and nF 0.03. For 2mg with 2ml flush the estimates were Ka 0.86L/min (95% CI 0.76 – 0.95L/min), Tlag 3.01min (95% CI 1.95 – 4.06min), F 0.87 (95% CI 0.79 – 0.95), and stdev of 0.01 (95% CI 0.006 – 0.011). The condition number was 46.41, the-2LL was -381.75, the epsshrinkage was 0.08, and nshrinkage for ka was 0.9999, Tlag 0.01, and nF 0.001. For 5mg with 1ml flush the estimates were Ka 0.48L/min (95% CI 0.41 – 0.54L/min), Tlag 3.13min (95% CI 1.86 – 4.39min), F 1.03 (95% CI 0.90 – 1.17), and stdev of 0.12 (95% CI 0.08 – 0.16). The condition number was 48.48, the-2LL was -36.24, the epsshrinkage was 0.08, and nshrinkage for ka was 0.9999, Tlag 0.07, and nF 0.0001. For 5mg with 2ml flush the estimates were Ka 0.83L/min (95% CI 0.30 – 1.36L/min), Tlag 3.29 min (95% CI 1.40– 5.19 min), F 1.09 (95% CI 0.96 – 1.21), and stdev of 0.02 (95% CI 0.01 – 0.02). The condition number was 63.92, the-2LL was -218.07, the epsshrinkage was 0.09, and nshrinkage for ka was 0.54, Tlag 0.14, and nF 0.09.

Conclusion:This is the first known modelling of gentamicin delivery kinetics. Therefore, there are no previous studies to compare the estimates obtained to. The studies all had high nshrinkage for Ka, therefore the individual estimates of ka may be unreliable. Further studies with a higher number of replicates would provide more favourable data for estimating Ka.


  • Lala, A. C. (2016). Variability in neonatal gentamicin administration influencing drug delivery kinetics (Thesis, Master of Medical Science). University of Otago.