A Machine Learning Approach to Predict Interdose Vancomycin Exposure.

INTRODUCTION: Estimation of vancomycin area under the curve (AUC) is challenging in the case of discontinuous administration. Machine learning approaches are increasingly used and can be an alternative to population pharmacokinetic (POPPK) approaches for AUC estimation. The objectives were to train XGBoost algorithms based on simulations performed in a previous POPPK study to predict vancomycin AUC from early concentrations and a few features (i.e. patient information) and to evaluate them in a real-life external dataset in comparison to POPPK. PATIENTS AND METHODS: Six thousand simulations performed from 6 different POPPK models were split into training and test sets. XGBoost algorithms were trained to predict trapezoidal rule AUC a priori or based on 2, 4 or 6 samples and were evaluated by resampling in the training set and validated in the test set. Finally, the 2-sample algorithm was externally evaluated on 28 real patients and compared to a state-of-the-art POPPK model-based averaging approach. RESULTS: The trained algorithms showed excellent performances in the test set with relative mean prediction error (MPE)/ imprecision (RMSE) of the reference AUC = 3.3/18.9, 2.8/17.4, 1.3/13.7% for the 2, 4 and 6 samples algorithms respectively. Validation in real patient showed flexibility in sampling time post-treatment initiation and excellent performances MPE/RMSE<1.5/12% for the 2 samples algorithm in comparison to different POPPK approaches. CONCLUSIONS: The Xgboost algorithm trained from simulation and evaluated in real patients allow accurate and precise prediction of vancomycin AUC. It can be used in combination with POPPK models to increase the confidence in AUC estimation.

Multimodal Imaging in Rhinoorbitocerebral Mucormycosis Associated with Type 2 Diabetes After COVID-19.

PURPOSE: This case series analyzed the appropriateness of computed tomography (CT) and magnetic resonance imaging (MRI) for visualization of rhinoorbitocerebral mucormycosis (ROCM) patterns associated with type 2 diabetes (T2D) post-recovery from coronavirus disease 2019 (COVID-19). METHODS: The study included 24 patients with invasive ROCM after having recovered from COVID-19. All patients underwent CT examinations and microbiological and histological verification; 5 patients underwent MRI. RESULTS: The CT and MRI patterns noted in our patients revealed involvement of skull orbits, paranasal sinuses, large arteries, and optic nerves, with intracranial spread and involvement of the cranial base bones. Using brain scan protocol for CT provided better soft-tissue resolution. We found that extending the MRI protocol by T2-sequence with fat suppression or STIR was better for periantral fat and muscle evaluations. CONCLUSION: Computed tomography of the paranasal sinuses is the method of choice for suspected fungal infections, particularly mucormycosis. However, MRI is recommended if there is suspicion of orbital, vascular, or intracranial complications, including cavernous sinus extension. The combination of both CT and MRI enables determination of soft tissue invasion and bony destruction, thereby facilitating the choice of an optimal ROCM treatment strategy. Invasive fungal infections are extremely rare in Europe; most of the related data are provided from India and Middle Eastern or African nations. Hence, this study is notable in its use of only diagnosed ROCM cases in Russia.

First clinical description of letermovir resistance mutation in cytomegalovirus UL51 gene and potential impact on the terminase complex structure.

BACKGROUND: Letermovir (LMV) is a human cytomegalovirus (HCMV) terminase inhibitor indicated as prophylaxis for HCMV-positive stem-cell recipients. Its mechanism of action involves at least the viral terminase proteins pUL56, pUL89 and pUL51. Despite its efficiency, resistance mutations were characterized in vitro and in vivo, largely focused on pUL56. To date, mutations in pUL51 in clinical resistance remain to be demonstrated. METHODS: The pUL51 natural polymorphism was described by sequencing 54 LMV-naive strains and was compared to UL51 HCMV genes from 16 patients non-responding to LMV therapy (prophylaxis or curative). Recombinant viruses were built by «en-passant¼ mutagenesis to measure the impact of the new mutations on antiviral activity and viral growth. Structure prediction was performed by homology modeling. The pUL51 final-model was analyzed and aligned with the atomic coordinates of the monomeric HSV-1 terminase complex (PDB:6M5R). RESULTS: Among the 16 strains from treated-patients with LMV, 4 never described substitutions in pUL51 (D12E, 17del, A95V, V113L) were highlighted. These substitutions had no impact on viral fitness. Only UL51-A95V conferred 13.8-fold increased LMV resistance level by itself (IC50 = 29.246 ± 0.788). CONCLUSION: As an isolated mutation in pUL51 in a clinical isolate can lead to LMV resistance, genotyping for resistance should involve sequencing of the pUL51, pUL56 and pUL89 genes. With terminase modelling, we make the hypothesis that LMV could bind to domains were UL56-L257I and UL51-A95V mutations were localized.