Search

Background Due to potentially fatal consequences of missed bacteremia, blood cultures are often overused. While there are several prediction models that can be used to identify patients who truly need blood cultures, physicians often rely on their gestalt. We evaluated the diagnostic performance of physician gestalt for bacteremia in comparison with 2 existing prediction models: Takeshima and Shapiro. Methods The study enrolled consecutive adult patients with suspected infection who were in the process of being admitted to the general medicine department at 2 hospitals between April 2017 and January 2019. Attending physicians provided gestalt regarding risk of bacteremia (0%–100%). Patients with a <10% risk estimated via each strategy (ie, physician gestalt or 2 existing models) were categorized as bacteremia excluded (ie, blood cultures were considered unnecessary). Strategies were compared in terms of safety (proportion of patients with bacteremia among those classified as bacteremia excluded) and efficiency (proportion of patients classified as bacteremia excluded among the total cohort). Results Among 2014 patients, 292 (14.5%) were diagnosed with bacteremia. The safety of physician gestalt and the Takeshima and Shapiro models was 3.7% (95% confidence interval [CI], 2.2% to 5.7%), 6.5% (95% CI, 5.0% to 7.9%), and 10.8% (95% CI, 9.4% to 12.3%), whereas the efficiency of each strategy was 22.4% (95% CI, 22.5% to 26.3%), 52.7% (95% CI, 50.5% to 54.9%), and 87.8% (95% CI, 86.3% to 89.2%), respectively. Conclusions Physician gestalt was safer but less efficient than existing models. Clinical prediction models could help reduce the overuse of blood cultures. [ABSTRACT FROM AUTHOR]

BackgroundDuring a pandemic, it is important for clinicians to stratify patients and decide who receives limited medical resources. Machine learning models have been proposed to accurately predict COVID-19 disease severity. Previous studies have typically tested only one machine learning algorithm and limited performance evaluation to area under the curve analysis. To obtain the best results possible, it may be important to test different machine learning algorithms to find the best prediction model. ObjectiveIn this study, we aimed to use automated machine learning (autoML) to train various machine learning algorithms. We selected the model that best predicted patients’ chances of surviving a SARS-CoV-2 infection. In addition, we identified which variables (ie, vital signs, biomarkers, comorbidities, etc) were the most influential in generating an accurate model. MethodsData were retrospectively collected from all patients who tested positive for COVID-19 at our institution between March 1 and July 3, 2020. We collected 48 variables from each patient within 36 hours before or after the index time (ie, real-time polymerase chain reaction positivity). Patients were followed for 30 days or until death. Patients’ data were used to build 20 machine learning models with various algorithms via autoML. The performance of machine learning models was measured by analyzing the area under the precision-recall curve (AUPCR). Subsequently, we established model interpretability via Shapley additive explanation and partial dependence plots to identify and rank variables that drove model predictions. Afterward, we conducted dimensionality reduction to extract the 10 most influential variables. AutoML models were retrained by only using these 10 variables, and the output models were evaluated against the model that used 48 variables. ResultsData from 4313 patients were used to develop the models. The best model that was generated by using autoML and 48 variables was the stacked ensemble model (AUPRC=0.807). The two best independent models were the gradient boost machine and extreme gradient boost models, which had an AUPRC of 0.803 and 0.793, respectively. The deep learning model (AUPRC=0.73) was substantially inferior to the other models. The 10 most influential variables for generating high-performing models were systolic and diastolic blood pressure, age, pulse oximetry level, blood urea nitrogen level, lactate dehydrogenase level, D-dimer level, troponin level, respiratory rate, and Charlson comorbidity score. After the autoML models were retrained with these 10 variables, the stacked ensemble model still had the best performance (AUPRC=0.791). ConclusionsWe used autoML to develop high-performing models that predicted the survival of patients with COVID-19. In addition, we identified important variables that correlated with mortality. This is proof of concept that autoML is an efficient, effective, and informative method for generating machine learning–based clinical decision support tools.

Patients with refractory or relapsed (R/R) large B-cell lymphoma (LBCL) refractory to first-line chemotherapy or with early relapse have poor outcomes. While chimeric antigen receptor (CAR) T-cell therapy has impressive efficacy after two or more lines of chemotherapy, it's still uncertain if these outcomes remain consistent in the context of third-line CAR T-cell therapy. We conducted a retrospective study of 107 R/R LBCL patients. Patients with relapse 12 months or more after their first-line chemoimmunotherapy (late failure: n = 25) had significantly longer overall survival (OS) than patients with refractory disease or relapse within 12 months (early failure: n = 82) (median OS: not achieved vs. 18.4 months; P < 0.001). Among patients who proceeded to autologous hematopoietic stem-cell transplantation (auto-HSCT), those with late failure had significantly longer event-free survival (EFS) than those with early failure (median EFS: 26.9 vs. 3.1 months; P = 0.012). However, no significant difference in EFS was detected among patients who underwent CAR T-cell therapy (median EFS: not reached vs. 11.8; P = 0.091). Cox regression with restricted cubic spline demonstrated that timing of relapse had significant impact on EFS in patients with auto-HSCT but not in patients with CAR T-cell therapy. Of patients who were scheduled for CAR T-cell therapy, those with late failure were significantly more likely to receive CAR T-cell therapy than those with early failure (90% vs. 57%; P = 0.008). In conclusion, patients with early failure still experienced poor outcomes after the approval of third-line CAR T-cell therapy.

Pola-BR (polatuzumab vedotin, bendamustine, and rituximab) therapy received approval for relapsed/refractory diffuse large B-cell lymphoma (R/R DLBCL) in Japan in March 2021. There have been few reports on the efficacy and safety of Pola-BR therapy in Japanese clinical practice. A retrospective analysis was performed on twenty-nine patients with R/R DLBCL who received Pola-BR therapy at our institution (intent to cellular immunotherapy cohort: 20 patients, stand-alone treatment cohort: nine patients). The overall response rate was 69.0% (complete response 27.6%). The median progression-free survival was 5.1 months, with a 9.5-month median overall survival. In the intent to cellular immunotherapy cohort, 11 of 19 patients received chimeric antigen receptor T-cell (CAR-T) infusions, and one patient received allogeneic stem cell transplantation. Four patients received Pola-BR therapy, including bendamustine before leukapheresis, and all produced CAR-T products successfully. 3 of the 28 patients experienced grade3 or higher adverse events, and two required treatment discontinuation. Our single institution, a real-world cohort of R/R DLBCL patients showed high efficacy outcomes and a tolerable toxicity profile for Pola-BR therapy, which is comparable to previous studies. More cases are needed to determine its impact on CAR-T therapy and stem cell transplantation.

Chimeric antigen receptor (CAR) T-cell therapy has revolutionized the approach to patients with relapsed or refractory diffuse large B-cell lymphoma (r/r DLBCL). This study retrospectively analyzed patients treated with commercially available tisagenlecleucel at our hospital and evaluated its safety and effectiveness. Of the 21 patients evaluated, any grade and grade ≥3 cytokine release syndrome (CRS) occurred in 85.7% and 9.5% of the patients, respectively. A total of 66.7% received tocilizumab and 28.6% received glucocorticoids for the treatment of CRS. The complete response (CR) rate at 3 months was 61.9% (95% confidence interval [CI] 38.4-81.9). After a median follow-up of 6.3 months following CAR-T infusion, the progression-free survival (PFS) and overall survival rates at 6 months were 53.1% (95%CI 28.3-72.7) and 69.2% (95%CI 43.7-84.9), respectively. Severe cytopenia and hypogammaglobulinemia occurred frequently following CAR-T infusion. Eight patients (38.1%) had comorbidities that would have made them ineligible for leukapheresis in the JULIET trial. However, the presence of comorbidities at the time of leukapheresis had no significant effect on the rates of CR, PFS, and adverse events. Tisagenlecleucel for r/r DLBCL in the real-world setting showed high efficacy and manageable safety profile comparable with the pivotal trial.