Development and evaluation of uncertainty quantifying machine learning models to predict piperacillin plasma concentrations in critically ill patients.

BACKGROUND: Beta-lactam antimicrobial concentrations are frequently suboptimal in critically ill patients. Population pharmacokinetic (PopPK) modeling is the golden standard to predict drug concentrations. However, currently available PopPK models often lack predictive accuracy, making them less suited to guide dosing regimen adaptations. Furthermore, many currently developed models for clinical applications often lack uncertainty quantification. We, therefore, aimed to develop machine learning (ML) models for the prediction of piperacillin plasma concentrations while also providing uncertainty quantification with the aim of clinical practice. METHODS: Blood samples for piperacillin analysis were prospectively collected from critically ill patients receiving continuous infusion of piperacillin/tazobactam. Interpretable ML models for the prediction of piperacillin concentrations were designed using CatBoost and Gaussian processes. Distribution-based Uncertainty Quantification was added to the CatBoost model using a proposed Quantile Ensemble method, useable for any model optimizing a quantile function. These models are subsequently evaluated using the distribution coverage error, a proposed interpretable uncertainty quantification calibration metric. Development and internal evaluation of the ML models were performed on the Ghent University Hospital database (752 piperacillin concentrations from 282 patients). Ensuing, ML models were compared with a published PopPK model on a database from the University Medical Centre of Groningen where a different dosing regimen is used (46 piperacillin concentrations from 15 patients.). RESULTS: The best performing model was the Catboost model with an RMSE and [Formula: see text] of 31.94-0.64 and 33.53-0.60 for internal evaluation with and without previous concentration. Furthermore, the results prove the added value of the proposed Quantile Ensemble model in providing clinically useful individualized uncertainty predictions and show the limits of homoscedastic methods like Gaussian Processes in clinical applications. CONCLUSIONS: Our results show that ML models can consistently estimate piperacillin concentrations with acceptable and high predictive accuracy when identical dosing regimens as in the training data are used while providing highly relevant uncertainty predictions. However, generalization capabilities to other dosing schemes are limited. Notwithstanding, incorporating ML models in therapeutic drug monitoring programs seems definitely promising and the current work provides a basis for validating the model in clinical practice.

Continuous versus intermittent infusion of cefotaxime in critically ill patients: a randomized controlled trial comparing plasma concentrations.

BACKGROUND: In critical care patients, reaching optimal ß-lactam concentrations poses challenges, as infections are caused more often by microorganisms associated with higher MICs, and critically ill patients typically have an unpredictable pharmacokinetic/pharmacodynamic profile. Conventional intermittent dosing frequently yields inadequate drug concentrations, while continuous dosing might result in better target attainment. Few studies address cefotaxime concentrations in this population. OBJECTIVES: To assess total and unbound serum levels of cefotaxime and an active metabolite, desacetylcefotaxime, in critically ill patients treated with either continuously or intermittently dosed cefotaxime. METHODS: Adult critical care patients with indication for treatment with cefotaxime were randomized to treatment with either intermittent dosing (1 g every 6 h) or continuous dosing (4 g/24 h, after a loading dose of 1 g). We defined a preset target of reaching and maintaining a total cefotaxime concentration of 4 mg/L from 1 h after start of treatment. CCMO trial registration number NL50809.042.14, Clinicaltrials.gov NCT02560207. RESULTS: Twenty-nine and 30 patients, respectively, were included in the continuous dosing group and the intermittent dosing group. A total of 642 samples were available for analysis. In the continuous dosing arm, 89.3% met our preset target, compared with 50% in the intermittent dosing arm. Patients not reaching this target had a significantly higher creatinine clearance on the day of admission. CONCLUSIONS: These results support the application of a continuous dosing strategy of ß-lactams in critical care patients and the practice of therapeutic drug monitoring in a subset of patients with higher renal clearance and need for prolonged treatment for further optimization, where using total cefotaxime concentrations should suffice.

Marked Changes in Gut Microbiota in Cardio-Surgical Intensive Care Patients: A Longitudinal Cohort Study.

Background: Virtually no studies on the dynamics of the intestinal microbiota in patients admitted to the intensive care unit (ICU) are published, despite the increasingly recognized important role of microbiota on human physiology. Critical care patients undergo treatments that are known to influence the microbiota. However, dynamics and extent of such changes are not yet fully understood. To address this topic, we analyzed the microbiota before, during and after planned major cardio surgery that, for the first time, allowed us to follow the microbial dynamics of critical care patients. In this prospective, observational, longitudinal, single center study, we analyzed the fecal microbiota using 16S rRNA gene sequencing. Results: Samples of 97 patients admitted between April 2015 and November 2016 were included. In 32 patients, data of all three time points (before, during and after admission) were available for analysis. We found a large intra-individual variation in composition of gut microbiota. During admission, a significant change in microbial composition occurred in most patients, with a significant increase in pathobionts combined with a decrease in strictly anaerobic gut bacteria, typically beneficial for health. A lower bacterial diversity during admission was associated with longer hospitalization. In most patients analyzed at all three time points, the change in microbiota during hospital stay reverted to the original composition post-discharge. Conclusions: Our study shows that, even with a short ICU stay, patients present a significant change in microbial composition shortly after admission. The unique longitudinal setup of this study displayed a restoration of the microbiota in most patients to baseline composition post-discharge, which demonstrated its great restorative capacity. A relative decrease in benign or even beneficial bacteria and increase of pathobionts shifts the microbial balance in the gut, which could have clinical relevance. In future studies, the microbiota of ICU patients should be considered a good target for optimisation.

Target attainment with continuous dosing of piperacillin/tazobactam in critical illness: a prospective observational study.

Optimal dosing of ß-lactam antibiotics in critically ill patients is a challenge given the unpredictable pharmacokinetic profile of this patient population. Several studies have shown intermittent dosing to often yield inadequate drug concentrations. Continuous dosing is an attractive alternative from a pharmacodynamic point of view. This study evaluated whether, during continuous dosing, piperacillin concentrations reached and maintained a pre-defined target in critically ill patients. Adult patients treated with piperacillin by continuous dosing in the intensive care unit of a university medical centre in The Netherlands were prospectively studied. Total and unbound piperacillin concentrations drawn at fixed time points throughout the entire treatment course were determined by liquid chromatography-tandem mass spectrometry. A pharmacokinetic combined target of a piperacillin concentration ≥80 mg/L, reached within 1 h of starting study treatment and maintained throughout the treatment course, was set. Eighteen patients were analysed. The median duration of monitored piperacillin treatment was 60 h (interquartile range, 33-96 h). Of the 18 patients, 5 (27.8%) reached the combined target; 15 (83.3%) reached and maintained a less strict target of >16 mg/L. In this patient cohort, this dosing schedule was insufficient to reach the pre-defined target. Depending on which target is to be met, a larger initial cumulative dose is desirable, combined with therapeutic drug monitoring.

[Fatal pneumonitis due to oseltamivir-resistant new influenza A(H1N1) in the case of an intensive care patient]. / Fatale pneumonitis door oseltamivir-resistente nieuwe influenza A(H1N1) bij een intensive-carepatiënt.

A 58-year-old man was submitted to our intensive care ward with respiratory failure due to pneumonitis. He had previously been treated for non-Hodgkin lymphoma by autologous stem cell transplantation, as a result of which bone marrow function was reduced. Further analysis showed infection with new influenza A(H1N1); typing revealed an oseltamivir-resistant subpopulation (H275Y). The patient was treated with oseltamivir and intravenously with zanamivir, but died of respiratory disease progression. This is the first published case of oseltamivir-resistant new influenza A(H1N1) infection in the Netherlands.

[Diagnostic image. A woman with a swollen tongue]. / Een vrouw met een dikke tong.

A 46-year-old woman had a swollen tongue due to amyloid light chain (AL) amyloidosis caused by multiple myeloma.