ABSTRACT
BACKGROUND: Machine learning (ML) approaches have become increasingly popular in predicting surgical outcomes. However, it is unknown whether they are superior to traditional statistical methods such as logistic regression (LR). This study aimed to perform a systematic review and meta-analysis to compare the performance of ML vs LR models in predicting postoperative outcomes for patients undergoing gastrointestinal (GI) surgery. METHODS: A systematic search of Embase, MEDLINE, Cochrane, Web of Science, and Google Scholar was performed through December 2022. The primary outcome was the discriminatory performance of ML vs LR models as measured by the area under the receiver operating characteristic curve (AUC). A meta-analysis was then performed using a random effects model. RESULTS: A total of 62 LR models and 143 ML models were included across 38 studies. On average, the best-performing ML models had a significantly higher AUC than the LR models (ΔAUC, 0.07; 95% CI, 0.04-0.09; P < .001). Similarly, on average, the best-performing ML models had a significantly higher logit (AUC) than the LR models (Δlogit [AUC], 0.41; 95% CI, 0.23-0.58; P < .001). Approximately half of studies (44%) were found to have a low risk of bias. Upon a subset analysis of only low-risk studies, the difference in logit (AUC) remained significant (ML vs LR, Δlogit [AUC], 0.40; 95% CI, 0.14-0.66; P = .009). CONCLUSION: We found a significant improvement in discriminatory ability when using ML over LR algorithms in predicting postoperative outcomes for patients undergoing GI surgery. Subsequent efforts should establish standardized protocols for both developing and reporting studies using ML models and explore the practical implementation of these models.
Subject(s)
Digestive System Surgical Procedures , Machine Learning , Postoperative Complications , Humans , Digestive System Surgical Procedures/adverse effects , Postoperative Complications/epidemiology , Logistic Models , ROC Curve , Area Under CurveABSTRACT
Background: Entrustable Professional Activities (EPAs) allow for the assessment of specific, observable, essential tasks in medical education. Since being developed in non-surgical fields, EPA assessments have been implemented in surgery to explore intraoperative entrustment. However, assessment burden is a significant problem for faculty, and it is unknown whether EPA assessments enable formative technical feedback. EPAs' formative utility could inform how surgical programs facilitate technical feedback for trainees. We aimed to assess the extent to which narrative comments provided through the Fellowship Council (FC) EPA assessments contained technical feedback. Methods: The FC previously collected EPA assessments for subspecialty surgical fellows from September 2020 to October 2022. Two raters reviewed assessments' narrative comments for inclusion of each skill area that makes up part of the Objective Structured Assessment of Technical Skills (OSATS). A third rater reconciled discrepant ratings. Results: During the study period, there were 3302 completed EPA assessments, including 1191 fellow self-assessments, 1124 faculty assessments, and 987 assessments without an identified assessor role. We found that assessments' narrative comments related to a median of two of the seven OSATS areas (IQR:1-2). There were no comments relevant to any of the seven OSATS areas in 16.0 % of all assessments. Conclusions: In this review of narrative comments for EPA assessments from the FC, we found that limited technical feedback of the kind included in the OSATS was provided in many assessments. These results suggest benefit to adjusting the EPA form, enhancing faculty development, or continuing additional types of targeted technical assessment intraoperatively. Key message: This analysis of narrative comments from fellowship EPA assessments showed that many assessments included limited technical feedback. To allow for continued technical feedback for fellows, these results highlight the need for further refinements of the EPA assessment form, additional faculty development, or ongoing use of other types of technical assessment.
ABSTRACT
BACKGROUND: The impact of concurrent chemotherapy and immunotherapy has been well characterized in patients with recurrent and metastatic head and neck squamous cell carcinoma (RM-SCCHN). Here, we report outcomes in patients treated sequentially with immune checkpoint inhibition (ICI) followed by carboplatin and paclitaxel. METHODS: Patients with RM-SCCHN treated with ICI followed by carboplatin/paclitaxel at a single institution were identified retrospectively. ICI therapy history, p16, and PD-L1 status were collected. The best overall response was assessed by RECIST v1.1. RESULTS: Twelve patients met inclusion criteria. Eight patients received pembrolizumab, three durvalumab, and one nivolumab. The median duration of ICI was 3.44 months, median PFS was 5.8 months, and median OS was 15.2 months. 66.7% of patients had an objective response on carboplatin/paclitaxel. CONCLUSIONS: Carboplatin/paclitaxel can induce objective responses in patients with prior treatment with ICI and clinical outcomes in this small series compare favorably to those seen in ICI naïve patients.
Subject(s)
Carcinoma, Squamous Cell , Head and Neck Neoplasms , Humans , Squamous Cell Carcinoma of Head and Neck/drug therapy , Squamous Cell Carcinoma of Head and Neck/etiology , Carboplatin , Paclitaxel , B7-H1 Antigen , Retrospective Studies , Neoplasm Recurrence, Local/pathology , Carcinoma, Squamous Cell/drug therapy , Carcinoma, Squamous Cell/etiology , Head and Neck Neoplasms/drug therapy , Head and Neck Neoplasms/etiology , Antineoplastic Combined Chemotherapy Protocols/therapeutic useABSTRACT
Despite robust literature associating IL-31 with pruritic inflammatory skin diseases, its influence on cutaneous inflammation and the interplay between inflammatory and neurosensory pathways remain unmapped. Here, we examined the consequences of disrupting Il31 and its receptor Il31ra in a mouse model of house dust mite (HDM)-induced allergic dermatitis. Il31-deficient mice displayed a deficit in HDM dermatitis-associated scratching, consistent with its well-established role as a pruritogen. In contrast, Il31 deficiency increased the number and proportion of cutaneous type 2 cytokine-producing CD4+ T cells and serum IgE in response to HDM. Furthermore, Il4ra+ monocytes and macrophages capable of fueling a feedforward type 2 inflammatory loop were selectively enriched in Il31ra-deficient HDM dermatitis skin. Thus, IL-31 is not strictly a proinflammatory cytokine but rather an immunoregulatory factor that limits the magnitude of type 2 inflammatory responses in skin. Our data support a model wherein IL-31 activation of IL31RA+ pruritoceptors triggers release of calcitonin gene-related protein (CGRP), which can mediate neurogenic inflammation, inhibit CD4+ T cell proliferation, and reduce T cell production of the type 2 cytokine IL-13. Together, these results illustrate a previously unrecognized neuroimmune pathway that constrains type 2 tissue inflammation in the setting of chronic cutaneous allergen exposure and may explain paradoxical dermatitis flares in atopic patients treated with anti-IL31RA therapy.
Subject(s)
Dermatitis, Atopic , Neurogenic Inflammation , Animals , Mice , Cytokines , Immunity , Pyroglyphidae , Skin/immunology , Interleukins/immunology , Interleukins/metabolismABSTRACT
The mitochondrial antiviral signaling protein (MAVS) orchestrates host antiviral innate immune response to RNA virus infection. However, how MAVS signaling is controlled to eradicate virus while preventing self-destructive inflammation remains obscure. Here, we show that protein geranylgeranylation, a posttranslational lipid modification of proteins, limits MAVS-mediated immune signaling by targeting Rho family small guanosine triphosphatase Rac1 into the mitochondria-associated endoplasmic reticulum (ER) membranes (MAMs) at the mitochondria-ER junction. Protein geranylgeranylation and subsequent palmitoylation promote Rac1 translocation into MAMs upon viral infection. MAM-localized Rac1 limits MAVS' interaction with E3 ligase Trim31 and hence inhibits MAVS ubiquitination, aggregation, and activation. Rac1 also facilitates the recruitment of caspase-8 and cFLIPL to the MAVS signalosome and the subsequent cleavage of Ripk1 that terminates MAVS signaling. Consistently, mice with myeloid deficiency of protein geranylgeranylation showed improved survival upon influenza A virus infection. Our work revealed a critical role of protein geranylgeranylation in regulating antiviral innate immune response.