Pang, Songhan

AbstractBackground and AimEffective clinical event classification is essential for clinical research and quality improvement. The validation of artificial intelligence (AI) models like Generative Pre‐trained Transformer 4 (GPT‐4) for this task and comparison with conventional methods remains unexplored.MethodsWe evaluated the performance of the GPT‐4 model for classifying gastrointestinal (GI) bleeding episodes from 200 medical discharge summaries and compared the results with human review and an International Classification of Diseases (ICD) code‐based system. The analysis included accuracy, sensitivity, and specificity evaluation, using ground truth determined by physician reviewers.ResultsGPT‐4 exhibited an accuracy of 94.4% in identifying GI bleeding occurrences, outperforming ICD codes (accuracy 63.5%, P < 0.001). GPT‐4's accuracy was either slightly lower or statistically similar to individual human reviewers (Reviewer 1: 98.5%, P < 0.001; Reviewer 2: 90.8%, P = 0.170). For location classification, GPT‐4 achieved accuracies of 81.7% and 83.5% for confirmed and probable GI bleeding locations, respectively, with figures that were either slightly lower or comparable with those of human reviewers. GPT‐4 was highly efficient, analyzing the dataset in 12.7 min at a cost of 21.2 USD, whereas human reviewers required 8–9 h each.ConclusionOur study indicates GPT‐4 offers a reliable, cost‐efficient, and faster alternative to current clinical event classification methods, outperforming the conventional ICD coding system and performing comparably to individual expert human reviewers. Its implementation could facilitate more accurate and granular clinical research and quality audits. Future research should explore scalability, prompt and model tuning, and ethical implications of high‐performance AI models in clinical data processing.

Abstract Background: Previous studies demonstrated that obesity is associated with poor outcomes and significantly worse pathological complete response (pCR) in triple-negative breast cancer (TNBC) treated with conventional chemotherapy. Emerging studies described a phenomenon termed the obesity paradox, which referred to improved outcomes observed in obese patients treated with immune checkpoint inhibitors (ICIs). Our study aimed to determine distinct immune features in TNBC across body mass index (BMI). Methods: Characteristics and outcomes of TNBC patients treated with neoadjuvant pembrolizumab-based chemotherapy at Mayo Clinic were collected. Normal BMI was defined as 18-25 kg/m2, overweight 25-30, and obese BMI ≥ 30. NanoString CosMx, a spatial multi-omics single-cell imaging (SMI), was performed in 75 samples from the Mayo Clinic TNBC cohort (Leon-Ferre BCRT 2018). Differential expression as log 2-fold change (FC) was estimated from the linear mixed model with significance defined as two-sided p&lt; 0.05. Results: Between 2021-2023, there were 108 TNBC patients treated with neoadjuvant pembrolizumab. 25.9% were obese, 32.4% were overweight, and 41.7% had normal BMI. In contrast to previous studies with conventional chemotherapy, obese patients numerically had a higher pCR rate of 71.4% with pembrolizumab-based chemotherapy compared to 55.6% in normal and 62.9% in overweight patients. However, this difference did not reach statistical significance (obese vs. normal weight, OR 2.26, 95%CI 0.89-6.06, p 0.094). We further evaluated spatially-resolved immune features associated with obesity using CosMx SMI in 75 samples from the MC TNBC cohort. Obese patients had significantly higher intratumoral B cells (log2FC 1.88, p 0.017), stromal B cells (log2FC 2.79, p 0.001), stromal plasmablasts (log2FC 1.94, p 0.047), and stromal macrophages (log2FC 1.80, p 0.017) compared to patients with normal BMI. Obese patients also had significantly lower intratumoral myeloid dendritic cells (mDC) (log2FC -1.27, p 0.037). Similar trends were observed after standardizing cell counts per 100 tumor or stromal cells, including higher intratumoral B cells (log2FC 3.00, p 0.001), intratumoral plasmablasts (log2FC 2.00, p 0.006), intratumoral natural killer (NK) cells (log2FC 1.66, p 0.044), and stromal B cells (log2FC 1.50, p 0.008). Differential gene analyses for intratumoral B cells revealed that genes (CD74, HLA-C, HLA-DRB1, HLA-DRA, HLA-DRB) associated with the positive regulation of type 2 immune or chemokine response were upregulated. Conclusions: Using in-depth analysis with spatially defined context, our study showed that obese TNBC patients had significantly higher B cell infiltration, which is associated with improved outcomes with ICIs. Future research on how obesity alters the immune microenvironment is needed to identify potential pathways to improve outcomes with ICIs in TNBC. Citation Format: Songhan Pang, Yi Liu, Yaohua Ma, Zhuo Li, Roberto A. Leon-Ferre, David Zahrieh, David W. Hillman, Judy C. Boughey, James N. Ingle, Krishna R. Kalari, Fergus J. Couch, Matthew P. Goetz, Keith L. Knutson, Saranya Chumsri, E. Aubrey Thompson. Tumor-immune microenvironment associated with obesity in triple-negative breast cancer in the Mayo Clinic cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 1162.