PurposeArtificial intelligence (AI) is a rapidly growing phenomenon poised to instigate large-scale changes in medicine. However, medical education has not kept pace with the rapid advancements of AI. Despite several calls to action, the adoption of teaching on AI in undergraduate medical education (UME) has been limited. This scoping review aims to identify gaps and key themes in the peer-reviewed literature on AI training in UME. MethodThe scoping review was informed by Arksey and O'Malley's methodology. Seven electronic databases including MEDLINE and EMBASE were searched for articles discussing the inclusion of AI in UME between January 2000 and July 2020. A total of 4,299 articles were independently screened by 3 co-investigators and 22 full-text articles were included. Data were extracted using a standardized checklist. Themes were identified using iterative thematic analysis.

e21127 Background: KRAS alterations constitute the most common driver mutations in metastatic non-small cell lung cancers (mNSCLC) and occur in approximately 30% of patients. KRAS mutational subtype as well as the presence of co-mutations has been associated with altered activation of downstream signaling pathways in preclinical models. We hypothesize that different KRAS G12C mutational subsets will be associated with variable clinical outcome and response to therapy. To this end, we have performed a retrospective analysis of survival and treatment outcomes by KRAS mutation subtype (G12C vs non-G12C). Methods: A review of KRAS-mutated mNSCLC patients treated with immunotherapy between 2013 and 2020 was conducted. Patient demographics, smoking status, KRAS mutational subtype, co-mutations and PD-L1 status were collected. Overall response rate (ORR) and progression-free survival (PFS) were analyzed in each subgroup. Results: 98 KRAS mutant mNSCLC patients were treated with immune checkpoint inhibitors (ICI): 37% with a KRAS G12C mutation, 62% with a non-G12C mutation. Patients with a G12C mutation were more likely to be of Caucasian ancestry (86% vs 56%; p = 0.01) whereas all other characteristics were similar between the groups including smoking history, PD-L1 expression ≥50% (61% vs 40%) and the presence of a TP53 co-mutation (48% vs. 54%); all p > 0.05. Treatment patterns were similar between the groups, with PD-1 inhibitor monotherapy given in 86% vs 79% of KRAS G12C and non-G12C patients. Overall response rate was 51% vs 27% in G12C vs non-G12C (p = 0.03). PFS was superior in G12C mutants (19.6 months vs 4.0 months), even after adjusting for smoking history, TP53 co-mutation status and PD-L1 expression (adjusted HR 0.51; p = 0.02). In subgroup analyses, the superiority in PFS was driven by the G12C mutants with high PD-L1 expression (n = 19): 26.8 months in G12C, PD-L1 high vs 4.7 months in G12C, PD-L1 low vs. 4.7 months in KRAS transversion mutations, PD-L1 high vs 4.0 months in transversion mutations, PD-L1 low vs. 3.0 months in transition mutations; p < 0.001. Conclusions: The presence of a KRAS G12C mutation is associated with improved ORR and PFS after treatment with ICI compared to non-G12C mutations in mNSCLC. The greatest benefit in PFS was observed in the subgroup with G12C mutation and high PD-L1 expression. Differential activation of downstream signaling associated with specific KRAS codon 12 mutation variants may modulate the composition of the tumor immune microenvironment thereby contributing to the variable response to immunotherapy. Further understanding on these molecular mechanisms may direct the development of new treatment strategies in KRAS mutant lung cancers.

An 85-year-old man with a known history of abdominal aortic aneurysm (AAA) presented to a vascular surgery clinic with a severely swollen, tender and erythematous left leg. An urgent CT angiogram demonstrated a left-sided, proximal deep vein thrombosis, and a permanent, Bird’s Nest inferior vena cava (IVC) filter (Cook, Inc., Bloomington, Ind.) penetrating his AAA. The patient was treated with a course of apixaban 5 mg two times per day and the decision was made to closely observe his IVC filter and AAA, given his numerous comorbidities and age. This case highlights the unique considerations associated with an approach to permanent IVC filter complications among patients with AAAs.

Letters to the Editor who were admitted with both pneumonia and asthma typically went straight to the intensive care unit, hence experienced fewer complications, but there was no way of including this contextual information in the AI's algorithm.Given the complexity of AI algorithms and their inability to deal with the context of patient care (at least for the foreseeable future), it is hard to see how AI tools can operate transparently. It is also difficult to imagine that such AI systems could replace human clinicians any more than a Google search could.