Search

The unprecedented global crisis brought about by the COVID-19 pandemic has sparked numerous efforts to create predictive models for the detection and prognostication of SARS-CoV-2 infections with the goal of helping health systems allocate resources. Machine learning models, in particular, hold promise for their ability to leverage patient clinical information and medical images for prediction. However, most of the published COVID-19 prediction models thus far have little clinical utility due to methodological flaws and lack of appropriate validation. In this paper, we describe our methodology to develop and validate multi-modal models for COVID-19 mortality prediction using multi-center patient data. The models for COVID-19 mortality prediction were developed using retrospective data from Madrid, Spain (N = 2547) and were externally validated in patient cohorts from a community hospital in New Jersey, USA (N = 242) and an academic center in Seoul, Republic of Korea (N = 336). The models we developed performed differently across various clinical settings, underscoring the need for a guided strategy when employing machine learning for clinical decision-making. We demonstrated that using features from both the structured electronic health records and chest X-ray imaging data resulted in better 30-day mortality prediction performance across all three datasets (areas under the receiver operating characteristic curves: 0.85 (95% confidence interval: 0.83–0.87), 0.76 (0.70–0.82), and 0.95 (0.92–0.98)). We discuss the rationale for the decisions made at every step in developing the models and have made our code available to the research community. We employed the best machine learning practices for clinical model development. Our goal is to create a toolkit that would assist investigators and organizations in building multi-modal models for prediction, classification, and/or optimization.

[Background] Locally advanced soft tissue sarcoma (STS) management may include neoadjuvant or adjuvant treatment by radiotherapy (RT), chemotherapy (CT) or chemoradiotherapy (CRT) followed by wide surgical excision. While pathological complete response (pCR) to preoperative treatment is prognostic for survival in osteosarcomas, its significance for STS is unclear. We aimed to evaluate the prognostic significance of pCR to pre-operative treatment on 3-year disease-free survival (3y-DFS) in STS patients., [Methods] This is an observational, retrospective, international, study of adult patients with primary non-metastatic STS of the extremities and trunk wall, any grade, diagnosed between 2008 and 2012, treated with at least neoadjuvant treatment and surgical resection and observed for a minimum of 3 years after diagnosis. The primary objective was to evaluate the effect of pCR. (≤5% viable tumor cells or ≥95% necrosis/fibrosis) on 3y-DFS. Effect on local recurrence-free survival (LRFS), distant recurrence-free survival (MFS) overall survival (OS) at 3 years was also analyzed. Statistical univariate analysis utilized chi-square independence test and odds ratio confidence interval (CI) estimate, multivariate analysis was performed using LASSO., [Results] A total of 330 patients (median age 56 years old, range:19–95) treated by preoperative RT (67%), CT (15%) or CRT (18%) followed by surgery were included. pCR was achieved in 74/330 (22%) of patients, of which 56/74 (76%) had received RT. 3-yr DFS was observed in 76% of patients with pCR vs 61% without pCR (p < 0.001). Multivariate analysis showed that pCR is statistically associated with better MFS (95% CI, 1.054–3.417; p = 0.033), LRFS (95% CI, 1.226–5.916; p = 0.014), DFS (95% CI, 1.165–4.040; p = 0.015) and OS at 3 years (95% CI, 1.072–5.210; p = 0.033)., [Conclusions] In a wide, heterogeneous STS population we showed that pCR to preoperative treatment is prognostic for survival.

The unprecedented global crisis brought about by the COVID-19 pandemic has sparked numerous efforts to create predictive models for the detection and prognostication of SARS-CoV-2 infections with the goal of helping health systems allocate resources. Machine learning models, in particular, hold promise for their ability to leverage patient clinical information and medical images for prediction. However, most of the published COVID-19 prediction models thus far have little clinical utility due to methodological flaws and lack of appropriate validation. In this paper, we describe our methodology to develop and validate multi-modal models for COVID-19 mortality prediction using multi-center patient data. The models for COVID-19 mortality prediction were developed using retrospective data from Madrid, Spain (N=2547) and were externally validated in patient cohorts from a community hospital in New Jersey, USA (N=242) and an academic center in Seoul, Republic of Korea (N=336). The models we developed performed differently across various clinical settings, underscoring the need for a guided strategy when employing machine learning for clinical decision-making. We demonstrated that using features from both the structured electronic health records and chest X-ray imaging data resulted in better 30-day-mortality prediction performance across all three datasets (areas under the receiver operating characteristic curves: 0.85 (95% confidence interval: 0.83-0.87), 0.76 (0.70-0.82), and 0.95 (0.92-0.98)). We discuss the rationale for the decisions made at every step in developing the models and have made our code available to the research community. We employed the best machine learning practices for clinical model development. Our goal is to create a toolkit that would assist investigators and organizations in building multi-modal models for prediction, classification and/or optimization.

BACKGROUND: In a phase III study, sunitinib led to a significant increase in progression-free survival (PFS) versus placebo in patients with pancreatic neuroendocrine tumours (panNETs). This study was a post-marketing commitment to support the phase III data. METHODS: In this ongoing, open-label, phase IV trial (NCT01525550), patients with progressive, advanced unresectable/metastatic, well-differentiated panNETs received continuous sunitinib 37.5 mg once daily. Eligibility criteria were similar to those of the phase III study. The primary endpoint was investigator-assessed PFS per Response Evaluation Criteria in Solid Tumours v1.0 (RECIST). Other endpoints included PFS per Choi criteria, overall survival (OS), objective response rate (ORR), and adverse events (AEs). RESULTS: Sixty-one treatment-naive and 45 previously treated patients received sunitinib. By March 19, 2016, 82 (77%) patients had discontinued treatment, mainly due to disease progression. Median treatment duration was 11.7 months. Investigator-assessed median PFS per RECIST (95% confidence interval [CI]) was 13.2 months (10.9-16.7): 13.2 (7.4-16.8) and 13.0 (9.2-20.4) in treatment-naive and previously treated patients, respectively. ORR (95% CI) per RECIST was 24.5% (16.7-33.8) in the total population: 21.3% (11.9-33.7) in treatment-naive and 28.9% (16.4-44.3) in previously treated patients. Median OS, although not yet mature, was 37.8 months (95% CI, 33.0-not estimable). The most common treatment-related AEs were neutropenia (53.8%), diarrhoea (46.2%), and leukopenia (43.4%). CONCLUSIONS: This phase IV trial confirms sunitinib as an efficacious and safe treatment option in patients with advanced/metastatic, well-differentiated, unresectable panNETs, and supports the phase III study outcomes. AEs were consistent with the known safety profile of sunitinib.

The influence of culture density on growth, stress system and metabolism has been assessed in A. regius juveniles submitted to different initial stocking densities (3, 7, 10 and 13 g L− 1) for 40 days. On days 0, 12, 27 and 40 of the experiment, biometric parameters were measured. In addition, plasma, liver and muscle samples were taken for biochemical analysis at the end of experimental time. Our results showed better growth rates and feed utilization in those specimens held at high stocking densities when compared to those under low stocking densities. In addition, higher plasma cortisol, glucose and triglycerides levels found in the lowest stocking density tested suggest an activation of the stress system in fish held at this condition. Moreover, no changes in hepatic metabolites were observed, while values of muscle triglycerides, amino acids and lactate enhanced when stocking density increased. In conclusion, the culture of juvenile A. regius at higher initial densities (at least 13 g L− 1) ensures better growth and the optimization of the existing facilities, improving welfare of this species in aquaculture systems.