Conor J. O'HAnlon

Accurately labeling outcomes in real‐world data for machine learning is challenging due to data sparsity and imbalances. This study developed and evaluated a pharmacokinetic–pharmacodynamic (PKPD)‐informed labeling strategy to enhance the risk prediction of docetaxel‐induced neutropenia. Machine learning models were trained on real‐world data from 4,248 patients using two approaches for comparison. The “naive” labeling method used only neutrophil observations, while the “PKPD‐informed” method used simulations from a semi‐mechanistic model to determine the neutrophil nadir for each treatment cycle. Three machine learning models (logistic regression, XGBoost, TabPFN) were trained with baseline laboratory data to predict severe neutropenia (neutrophil count 9 /L) prior to the first docetaxel dose. The PKPD labeling approach enabled the labeling of 3.4 times more patient instances (7,719 vs. 2,283) than the naive method. Across all machine learning architectures, models trained with PKPD‐informed labels demonstrated significantly superior predictive performance (AUC‐ROC and AUC‐PR) compared to those trained with naive labels. This advantage was maintained even when training set sizes were matched. PKPD‐informed labeling overcomes limitations of sparse real‐world data, increasing both the quantity and apparent quality of labels for machine learning model training. This methodology enhances the performance of machine learning models for predicting severe neutropenia and represents a robust, generalizable framework for improving clinical outcome prediction.

Cardiopulmonary bypass (CPB) can alter pharmacokinetic (PK) parameters and the drug may adsorb to the CPB device, altering exposure. Cefazolin is a beta‐lactam antibiotic used for antimicrobial prophylaxis during cardiac surgery supported by CPB. Adsorption of cefazolin could result in therapeutic failure. An ex vivo study was undertaken using CPB devices primed and then dosed with cefazolin and samples were obtained over 1 hour of recirculation. Twelve experimental runs were conducted using different CPB device sizes (neonate, infant, child, and adult), device coatings (Xcoating™, Rheoparin®, PH.I.S.I.O), and priming solutions. The time course of saturable binding, using B_max (binding capacity), K_d (dissociation constant), and T2_off (half‐time of dissociation), described cefazolin adsorption. B_max estimates for the device sizes were neonate 40.0 mg (95% CI 24.3, 67.4), infant 48.6 mg (95% CI 5.97, 80.2), child 77.8 mg (95% CI 54.9, 103), and adult 196 mg (95% CI 191, 199). The Xcoating™ K_d estimate was 139 mg/L (95% CI 27.0, 283) and the T2_off estimate was 98.4 min (95% CI 66.8, 129). The Rheoparin® and PH.I.S.I.O coatings had similar binding parameters with K_d and T2_off estimates of 0.169 mg/L (95% CI 0.01, 1.99) and 4.94 min (95% CI 0.17, 59.4). The B_max was small (< 10%) relative to a typical total patient dose during cardiac surgery supported by CPB. A dose adjustment for cefazolin based solely on drug adsorption is not required. This framework could be extended to other PK studies involving CPB.

Background and Objectives Milrinone is an inotrope and vasodilator used for prophylaxis or treatment of low cardiac output syndrome after weaning from cardiopulmonary bypass (CPB). It is renally eliminated and has an acceptable therapeutic range of 100–300 μg/L, but weight-based dosing alone is associated with poor target attainment. We aimed to develop a population pharmacokinetic model for milrinone from premature neonates to adolescents, and to evaluate how age, renal function and recovery from CPB may impact dose selection. Methods Fifty paediatric patients (aged 4 days to 16 years) were studied after undergoing cardiac surgery supported by CPB. Data from 29 premature neonates (23–28 weeks’ postmenstrual age) treated for prophylaxis of low systemic blood flow were available for a pooled pharmacokinetic analysis. Population parameters were estimated using non-linear mixed effects modelling (NONMEM 7.5.1). Results There were 369 milrinone measurements available for analysis. A one-compartment model with zero-order input and first-order elimination was used to describe milrinone disposition. Population parameters were clearance 17.8 L/70 kg [95% CI 15.8–19.9] and volume 20.4 L/h/70 kg [95% CI 17.8–22.1]. Covariates included size, postmenstrual age and renal function for clearance, and size and postnatal age for volume. Milrinone clearance is reduced by 39.5% [95% CI 24.0–53.7] immediately after bypass, and recovers to baseline clearance with a half-time of 12.0 h [95% CI 9.7–15.2]. Milrinone volume was 2.07 [95% CI 1.87–2.27] times greater at birth than the population standard and decreased over the first days of life with a half-time of 0.977 days [95% CI 0.833–1.12]. Conclusion Milrinone is predominately renally eliminated and so renal function is an important covariate describing variability in clearance. Increasing clearance over time likely reflects increasing cardiac output and renal perfusion due to milrinone and return to baseline following CPB.

Aims: We propose using glomerular filtration rate (GFR) as the physiological basis for distinguishing components of renal clearance.Methods: Gentamicin, amikacin and vancomycin are thought to be predominantly excreted by the kidneys. A mixed‐effects joint model of the pharmacokinetics of these drugs was developed, with a wide dispersion of weight, age and serum creatinine. A dataset created from 18 sources resulted in 27,338 drug concentrations from 9,901 patients. Body size and composition, maturation and renal function were used to describe differences in drug clearance and volume of distribution.Results: This study demonstrates that GFR is a predictor of two distinct components of renal elimination clearance: (1) GFR clearance associated with normal GFR and (2) non‐GFR clearance not associated with normal GFR. All three drugs had GFR clearance estimated as a drug‐specific percentage of normal GFR (gentamicin 39%, amikacin 90% and vancomycin 57%). The total clearance (sum of GFR and non‐GFR clearance), standardized to 70 kg total body mass, 176 cm, male, renal function 1, was 5.58 L/h (95% confidence interval [CI] 5.50‐5.69) (gentamicin), 7.77 L/h (95% CI 7.26‐8.19) (amikacin) and 4.70 L/h (95% CI 4.61‐4.80) (vancomycin).Conclusions: GFR provides a physiological basis for renal drug elimination. It has been used to distinguish two elimination components. This physiological approach has been applied to describe clearance and volume of distribution from premature neonates to elderly adults with a wide dispersion of size, body composition and renal function. Dose individualization has been implemented using target concentration intervention.

Quantifying the effect of kidney disease on glomerular filtration rate (GFR) is important when describing variability in the clearance of drugs eliminated by the kidney. We aimed to develop a continuous model for renal function (RF) from prematurity to adulthood based on consistent models for fat‐free mass (FFM), creatinine production rate (CPR), and GFR. A model for fractional FFM in premature neonates to adults was developed using pooled data from 4462 subjects and 2847 FFM observations. It was found that girls have an FFM higher than that predicted from adult women based on height, total body mass, and sex, and boys have an FFM lower than adult men until around the onset of puberty, when it approaches adult male values. Data from 108 subjects with measurements of serum creatinine (Scr) and GFR were used to construct a model for CPR. Creatinine clearance was predicted using a model for CPR (based on FFM, postmenstrual age, and sex) and Scr that avoids discontinuous predictions between neonates, children, and adults. Individual CPR may then be used with individual Scr to predict the estimated GFR (eGFR; eGFR = CPR/Scr). A previously published model for human GFR based on 1153 GFR observations in 923 subjects without known kidney disease was updated using the model for fractional FFM to predict individual size and age‐consistent values for the expected normal GFR (nGFR). Individual renal function was then calculated using RF = eGFR/nGFR.

Aim: Renal function is an important covariate to describe variability in clearance of renally eliminated drugs. Estimating glomerular filtration rate (GFR) should account for differences in size, age and body composition consistently when using GFR to determine drug dosage. We aimed to develop a continuous model for renal function from prematurity to adulthood based on models for fat free mass (FFM), creatinine production rate (CPR) and GFR. Methods: A model for fractional FFM in premature neonates to adults was developed using pooled data from 4462 subjects and 2847 FFM observations. Data from 108 subjects with measurements of serum creatinine and GFR were used to construct a model for CPR by assuming that CLcr is equal to GFR. A previously published model for human GFR was updated using the model for fractional FFM and accounting for the effects of maturation and birth. Together these models were used to predict renal function. Results: Girls have a FFM larger than that predicted from adult women based on height, weight and sex. Boys have a FFM lower than adult men predicted FFM until around the onset of puberty, when it approaches adult values. CPR can be predicted using FFM, post menstrual age and sex and avoids discontinuous predictions between neonates, children and adults. The updated model for GFR maturation was used to describe expected normal GFR. Renal function calculated from the ratio of individual CLcr to normal GFR. Conclusion: Continuous models for FFM, CPR and GFR predict renal function independent of age and size.

Development of a vaccine to limit the impact of antibiotic resistant Neisseria gonorrhoeae is now a global priority. Serum bactericidal antibody (SBA) is a possible indicator of protective immunity to N. gonorrhoeae, but conventional assays measure colony forming units (CFU), which is time-consuming. A luminescent assay that quantifies ATP as a surrogate measure of bacterial viability was tested on N. gonorrhoeae strains FA1090, MS11 and P9-17 and compared to CFU-based readouts. There was a linear relationship between CFU and ATP levels for all three strains (r > 0.9). Normal human serum (NHS) is a common source of complement for SBA assays, but needs to be screened for non-specific bactericidal activity. NHS from 10 individuals were used for serum sensitivity assays—sensitivity values were significantly reduced with the ATP method for FA1090 (5/10, p < 0.05) and MS11 (10/10, p < 0.05), whereas P9-17 data were comparable for all donors. Our results suggest that measuring ATP underestimates serum sensitivity of N. gonorrhoeae and that the CFU method is a better approach. However, mouse anti-P9-17 outer membrane vesicles (OMV) SBA titres to P9-17 were comparable with both methods (r = 0.97), suggesting this assay can be used to rapidly screen sera for bactericidal antibodies to gonococci.