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.

Population pharmacokinetics and evaluation of the predictive performance of pharmacokinetic models in critically ill patients receiving continuous infusion meropenem: a comparison of eight pharmacokinetic models.

Background: Several population pharmacokinetic (PopPK) models for meropenem dosing in ICU patients are available. It is not known to what extent these models can predict meropenem concentrations in an independent validation dataset when meropenem is infused continuously. Patients and methods: A PopPK model was developed with concentration-time data collected from routine care of 21 ICU patients (38 samples) receiving continuous infusion meropenem. The predictability of this model and seven other published PopPK models was studied using an independent dataset that consisted of 47 ICU patients (161 samples) receiving continuous infusion meropenem. A statistical comparison of imprecision (mean square prediction error) and bias (mean prediction error) was conducted. Results: A one-compartment model with linear elimination and creatinine clearance as a covariate of clearance best described our data. The mean ± SD parameter estimate for CL was 9.89 ±âŸ3.71 L/h. The estimated volume of distribution was 48.1 L. The different PopPK models showed a bias in predicting serum concentrations from the validation dataset that ranged from -8.76 to 7.06 mg/L. Imprecision ranged from 9.90 to 42.1 mg/L. Conclusions: Published PopPK models for meropenem vary considerably in their predictive performance when validated in an external dataset of ICU patients receiving continuous infusion meropenem. It is necessary to validate PopPK models in a target population before implementing them in a therapeutic drug monitoring program aimed at optimizing meropenem dosing.

Oral Anticoagulation in Patients With Advanced Chronic Kidney Disease and Atrial Fibrillation: Beyond Anticoagulation.

The optimal approach to prevent stroke and systemic embolism in patients with advanced chronic kidney disease (CKD) and atrial fibrillation remains unresolved. We conducted a narrative review to explore areas of uncertainty and opportunities for future research. First, the relationship between atrial fibrillation and stroke is more complex in patients with advanced CKD than in the general population. The currently employed risk stratification tools do not adequately discriminate between patients deriving a net benefit and those suffering a net harm from oral anticoagulation. Anticoagulation initiation should probably be more restrictive than is currently advocated by official guidelines. Recent evidence reveals that the superior benefit-risk profile of non-vitamin K antagonist oral anticoagulants (NOACs) vs vitamin K antagonists (VKAs) observed in the general population and in moderate CKD can be extended to advanced CKD. The NOACs yield better protection against stroke, cause less major bleeding, are associated with less acute kidney injury and a slower decline of CKD, and are associated with a lower incidence of cardiovascular events than VKAs. The VKAs may be harmful in CKD patients, in particular in patients with a high bleeding risk and labile international normalized ratio. The better safety and efficacy of NOACs as opposed to VKAs may be particularly evident in advanced CKD as a result of better on-target anticoagulation with NOACs, harmful off-target vascular effects of VKAs, and beneficial off-target vascular effects of NOACs. The intrinsic vasculoprotective effects of NOACs are supported by animal experimental evidence as well as by findings of large clinical trials and may result in use of NOACs beyond their anticoagulant properties.

Why We May Need Higher Doses of Beta-Lactam Antibiotics: Introducing the 'Maximum Tolerable Dose'.

The surge in antimicrobial resistance and the limited availability of new antimicrobial drugs has fueled the interest in optimizing antibiotic dosing. An ideal dosing regimen leads to maximal bacterial cell kill, whilst minimizing the risk of toxicity or antimicrobial resistance. For beta-lactam antibiotics specifically, PK/PD-based considerations have led to the widespread adoption of prolonged infusion. The rationale behind prolonged infusion is increasing the percentage of time the beta-lactam antibiotic concentration remains above the minimal inhibitory concentration (%fT>MIC). The ultimate goal of prolonged infusion of beta-lactam antibiotics is to improve the outcome of infectious diseases. However, merely increasing target attainment (or the %fT>MIC) is unlikely to lead to improved clinical outcome for several reasons. First, the PK/PD index and target are dynamic entities. Changing the PK (as is the case if prolonged instead of intermittent infusion is used) will result in different PK/PD targets and even PK/PD indices necessary to obtain the same level of bacterial cell kill. Second, the minimal inhibitory concentration is not a good denominator to describe either the emergence of resistance or toxicity. Therefore, we believe a different approach to antibiotic dosing is necessary. In this perspective, we introduce the concept of the maximum tolerable dose (MTD). This MTD is the highest dose of an antimicrobial drug deemed safe for the patient. The goal of the MTD is to maximize bacterial cell kill and minimize the risk of antimicrobial resistance and toxicity. Unfortunately, data about what beta-lactam antibiotic levels are associated with toxicity and how beta-lactam antibiotic toxicity should be measured are limited. This perspective is, therefore, a plea to invest in research aimed at deciphering the dose−response relationship between beta-lactam antibiotic drug concentrations and toxicity. In this regard, we provide a theoretical approach of how increasing uremic toxin concentrations could be used as a quantifiable marker of beta-lactam antibiotic toxicity.

Early target attainment of continuous infusion piperacillin/tazobactam and meropenem in critically ill patients: A prospective observational study.

PURPOSE: To evaluate target attainment of empirically dosed continuous infusion piperacillin/tazobactam (TZP) and meropenem (MER) in critically ill patients. PATIENTS AND METHODS: Patients were sampled on a daily basis. TZP or MER concentrations were evaluated during the first two days antibiotic therapy. The lower limit of the target range was defined as unbound concentrations equaling 4 times the epidemiological cutoff value of P. aeruginosa. The upper limit of the target range was based on the risk of toxicity, i.e. unbound concentrations >160 mg/L for TZP and > 45 mg/L for MER. Multivariable logistic regression was used to evaluate factors associated with target attainment. RESULTS: Data from 253 patients were analyzed. Overall, 76/205 (37.1%) and 36/48 (75%) of the patients receiving TZP or MER respectively, attained target concentrations. In multivariable analysis, estimated creatinine clearance was identified as a risk factor for target non-attainment (OR 0.988, 95%CI [0.982;0.994]). Patients receiving MER were more likely to attain target concentrations compared with patients receiving TZP (OR 6.02, 95%CI [2.12;18.4]). CONCLUSION: Target attainment of empiric antibiotic therapy in critically ill patients was low (37%) for TZP and moderate (75%) for MER, despite the use of a loading dose and despite optimization of the mode of infusion.

Population pharmacokinetics of continuous infusion of piperacillin in critically ill patients.

Dosing recommendations for continuous infusion of piperacillin, a broad-spectrum beta-lactam antibiotic, are mainly guided by outputs from population pharmacokinetic models constructed with intermittent infusion data. However, the probability of target attainment in patients receiving piperacillin by continuous infusion may be overestimated when drug clearance estimates from population pharmacokinetic models based on intermittent infusion data are used, especially when higher doses (e.g. 16 g/24 h or more) are simulated. Therefore, the purpose of this study was to describe the population pharmacokinetics of piperacillin when infused continuously in critically ill patients. For this analysis, 270 plasma samples from 110 critically ill patients receiving piperacillin were available for population pharmacokinetic model building. A one-compartment model with linear clearance best described the concentration-time data. The mean ± standard deviation parameter estimates were 8.38 ± 9.91 L/h for drug clearance and 25.54 ± 3.65 L for volume of distribution. Creatinine clearance improved the model fit and was supported for inclusion as a covariate. In critically ill patients with renal clearance higher than 90 mL/min/1.73 m2, a high-dose continuous infusion of 24 g/24 h is insufficient to achieve adequate exposure (pharmacokinetic/pharmacodynamic target of 100% fT>4 x MIC) against susceptible Pseudomonas aerginosa isolates (MIC ≤16 mg/L). These findings suggest that merely increasing the dose of piperacillin, even with continuous infusion, may not always result in adequate piperacillin exposure. This should be confirmed by evaluating piperacillin target attainment rates in critically ill patients exhibiting high renal clearance.

Emergence of antimicrobial resistance to piperacillin/tazobactam or meropenem in the ICU: Intermittent versus continuous infusion. A retrospective cohort study.

BACKGROUND: Prolonged infusion of beta-lactam antibiotics is broadly recognized as a strategy to optimize antibiotic therapy by achieving a higher percentage of time that concentrations remain above the minimal inhibitory concentration (% fT>MIC), i.e. the pharmacokinetic/pharmacodynamic (PK/PD) index. However, %fT>MIC may not be the PK/PD index of choice for inhibition of resistance emergence and it is therefore unsure what impact prolonged infusion of beta-lactam antibiotics may have on the emergence of resistance. METHODS: A retrospective cohort study including 205 patients receiving either intermittent (101 patients) or continuous (104 patients) infusion of piperacillin/tazobactam or meropenem was conducted in the ICU of the Ghent University Hospital. Logistic regression analysis was used to develop a prediction model and to determine whether the mode of infusion was a predictor of emergence of antimicrobial resistance. RESULTS: Resistant strains emerged in 24 out of the 205 patients (11.7%). The mode of infusion was no predictor of emergence of antimicrobial resistance. Infection with Pseudomonas aeruginosa was associated with a significantly higher risk for emergence of resistance. CONCLUSIONS: In this retrospective cohort study, the emergence of antimicrobial resistance to piperacillin/tazobactam or meropenem was not related to the mode of infusion.