Exploring the role of Artificial Intelligence in Bariatric and Metabolic Surgery

Exploring the role of Artificial Intelligence in Bariatric and Metabolic Surgery | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Exploring the role of Artificial Intelligence in Bariatric and Metabolic Surgery Carlos J Pérez-Padrón, Hugo Sánchez-Aguilar, Rafael H Pérez-Soto, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8427770/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Introduction Artificial intelligence (AI) has gained increasing relevance in bariatric and metabolic surgery (BMS), yet its concordance with expert clinical judgement remains uncertain. This study evaluates the agreement between two AI models and bariatric and metabolic surgeons across representative clinical scenarios, assessing the impact of directed exposure to bibliography. Material and methods Ten evidence-based clinical scenarios were constructed using high-quality bibliography. Two AI models, ChatGPT-4 and DeepSeek, along with board-certified Mexican BMS surgeons, evaluated surgical candidacy and determined whether each patient was candidate for surgery, as well as the most appropriate procedure, under two conditions: without access to high-quality bibliography (Phase 1), and after reviewing it (Phase 2). Majority surgeon responses served as the reference standard. Agreement and concordance were analyzed descriptively and with Cohen’s kappa (p < 0.05). Results Thirty surgeons completed both phases. In Phase 1, surgeons agreement varied, ranging from 96.6% to 23.3% in selected cases; when comparing to AI models, ChatGPT-4 showed 60% agreement (K = 0.344), while DeepSeek 70% (k = 0.508). In Phase 2, surgeons demonstrated 90% agreement across phases (k = 0.787). ChatGPT-4 showed 60% agreement without significant concordance (k = 0.048), whereas DeepSeek maintained 70% agreement, with fair concordance (k = 0.318). Conclusions Surgeons maintained high consistency between phases, while AI models showed variable alignment with clinical decision-making. DeepSeek demonstrated higher adaptability to bibliographic evidence, whereas ChatGPT-4 didn’t. These findings highlight the need for continued refinement, contextual training and validation of AI tools to ensure safe and clinically applicable support in BMS decision-making. Bariatric surgery Artificial intelligence Large language models gastric sleeve Roux-en-Y gastric bypass Introduction Artificial intelligence (AI) represents a sustained human endeavor to enhance the quality of life and meet fundamental societal needs. First conceptualized by John McCarthy in 1956 [ 1 ], AI has evolved rapidly with the era of digital globalization, becoming increasingly integrated into everyday life. Beyond improving the efficiency and accuracy of diverse services through large-scale data processing, AI now enables the detection of complex patterns beyond human analytical capacity, particularly in clinical imaging, predictive modeling, and the development of decision support systems that assist physicians in diagnosis and clinical management. In the field of bariatric and metabolic surgery (BMS), the integration of AI is no exception. A notable early example is the study by Jazi et al. [ 2 ], which compared the recommendations generated by a large language model (ChatGPT-4) with those provided by bariatric and metabolic surgeons for selected clinical scenarios. The authors reported a concordance rate of approximately 30% between both groups, which was considered poor, underscoring the critical importance of human clinical judgment in the evaluation of candidates for weight-loss surgery and in the selection of the most appropriate surgical procedure. Another potential application of AI lies across different stages of the surgical process, including preoperative planning, intraoperative guidance, postoperative assessment and outcome prediction [ 3 ]. Yung Lee et al. [ 4 ] compared the performance of three AI models, ChatGPT-4, Bing, and Bard, in responding to clinical questions related to bariatric and metabolic surgery, finding that ChatGPT-4 provided the highest proportion of correct answers compared with the other models. In a subsequent study, Yung Lee et al. [ 5 ] evaluated the same three models using questions derived from the American Society for Metabolic and Bariatric Surgery (ASMBS) Textbook of Bariatric Surgery , reporting accuracies of 83.0% for ChatGPT-4, 76.0% for Bard, and 65.0% for Bing. These results demonstrate a relatively high level of accuracy for ChatGPT-4, suggesting its potential reliability as an educational and decision-support tool in BMS. The SOPHIA AI-based weight-loss prediction tool [ 6 ], was developed using supervised machine learning techniques, including LASSO regression and decision tree-based algorithms. Trained on data from more than 10,000 bariatric patients across Europe, the model analyzed hundreds of baseline variables and identified seven key predictors of postoperative weight trajectory. It was externally validated to estimate individual weight outcomes up to five years after surgery. This AI driven approach enables personalized, data informed predictions that support clinical decision-making in BMS. To date, no studies have evaluated whether different AI models generate distinct clinical recommendations based on there original training or whether these recommendations change after contextual bibliographic training. Given the widespread public accessibility of these tools, comparing AI outputs is essential to understand the recommendations that patients may receive and the extent to which such responses can be modulated by high-quality evidence. The performance and responses of AI models depend on the information to which they have been exposed and their capacity to learn. This training process is crucial for developing greater fluency and accuracy in answering complex queries. Comparing different versions of the same model can reveal marked variations in precision, reasoning, and alignment with scientific evidence. This phenomenon has been described, for instance, when comparing the performance of ChatGPT-3 and ChatGPT-4, where a noticeable improvement in coherence and accuracy was observed in clinical question responses [ 7 ]. Within this context, a new question arises: can artificial intelligence truly optimize the quality of surgical decision-making in BMS by enhancing medical and surgical reasoning? The principal objective of this study was to compare surgical recommendations made by AI models (ChatGPT-4 and Deepseek) with those of bariatric and metabolic surgeons in representative obesity clinical scenarios based on high-quality evidence. Materials and Methods Ten representative clinical scenarios of patients with obesity were designed, incorporating variables such as age, sex, body mass index (BMI), and the presence of obesity-associated comorbidities. The creation of these scenarios was based on the highest-quality evidence currently available [ 8 – 14 ], primarily derived from clinical trials in which bariatric and metabolic surgery demonstrated a significant impact on improvement or resolution of obesity-associated comorbidities. The ten simulated clinical scenarios used for surgeon and AI evaluation are provided in Supplementary Material 2. Two AI models, ChatGPT-4 and DeepSeek, were asked to determine whether each hypothetical patient was a candidate for BMS from the perspective of an expert surgeon. When surgery was indicated, the models were further requested to recommend the most appropriate surgical procedure among the following options: gastric sleeve (GS), Roux-en-Y gastric bypass (RYGB), one-anastomosis gastric bypass (OAGB), biliopancreatic diversion with duodenal switch (BPD-DS), single-anastomosis duodeno–ileal bypass with sleeve gastrectomy (SADI-S), intestinal transit bipartition (ITB), or other. Recommendations were obtained under two different conditions: without access to bibliographic references ( Phase 1 ) and with access to key clinical references ( Phase 2 ) [ 8 – 14 ], aiming to analyze the impact of bibliographic support on the surgical decision-making process of AI models. The same ten clinical scenarios were distributed to board-certified bariatric and metabolic surgeons, listed in Supplementary Material 1, through Google Forms, under the same conditions previously described. In Phase 1, participants were individually asked to determine whether each hypothetical patient was a candidate for bariatric surgery and, if so, to specify the most appropriate procedure. In Phase 2, 4 weeks after the prior phase, the same group of surgeons received the bibliographic references and, after reviewing and analyzing them, were asked to provide updated surgical recommendations for the same cases. In both phases, all responses were collected anonymously employing an anonymized matching strategy using non-identifiable demographic variables, with participants blinded to the answers of their peers and those generated by the AI models. A descriptive analysis was performed on the responses obtained from both the surgeons and the AI models to identify patterns and trends. The consensus answer from the surgical group was defined as the option selected by the majority and was considered the reference standard for comparison with the AI models. To assess the concordance between surgeons’ and AI models’ recommendations, Cohen’s kappa coefficient was estimated, with a p value < 0.05 considered statistically significant. All analyses were performed using SPSS software, version 30 (IBM, USA). Results For Phase 1 of the study, 43 bariatric and metabolic surgeons were invited to participate. Only the 30 surgeons who participated in both phases were included in the final analysis. The sample consisted of 74.2% men and 25.8% women. The predominant age group was between 40 and 50 years (38.7%). Regarding to geographical distribution, 42% of participants reported practicing in Mexico City, followed by Jalisco and Nuevo León. The responses to the clinical scenarios from Phase 1 are presented in Table 1 . Agreement among surgeons was variable, with the highest agreement observed in Case 5, where 96.6% (n = 29) selected RYGB as the most appropriate surgical option. Conversely, the lowest agreement was noted in Case 9, with only 23.3% of the surgeons selecting ITB. Table 1 Summary of responses provided by Bariatric and Metabolic Surgeons in Phase 1 and Phase 2 of the study. (GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno–ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition). Case / Surgical option No surgery (N) GS (N) RYGB (N) OAGB (N) DS (N) SADI-S (N) ITB (N) Other (N) Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 Phase 1 Phase 2 1 3.3% (1) 13.3% (4) 83.3% (25) 76.6% (23) 6.6% (2) 10% (3) 3.3% (1) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 3.3% (1) 0% (0) 2 0% (0) 0% (0) 3.3% (1) 0% (0) 76.6% (23) 76.6% (23) 13.3% (4) 13.3% (4) 0% (0) 0% (0) 0% (0) 0% (0) 6.6% (2) 10% (3) 0% (0) 0% (0) 3 0% (0) 0% (0) 40% (12) 26.6% (8) 46.6% (14) 56.6% (17) 6.6% (3) 10% (3) 0% (0) 0% (0) 0% (0) 0% (0) 6.6% (2) 6.6% (2) 0% (0) 0% (0) 4 0% (0) 0% (0) 26.6% (8) 16.6% (5) 50% (15) 60% (18) 6.6% (2) 6.6% (2) 0% (0) 0% (0) 3.3% (1) 3.3% (1) 13.3% (6) 13.3% (4) 0% (0) 0% (0) 5 0% (0) 0% (0) 0% (0) 0% (0) 96.6% (29) 96.6% (29) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 3.3% (1) 3.3% (1) 0% (0) 0% (0) 6 0% (0) 0% (0) 26.6% (8) 20% (6) 33.3% (10) 30% (9) 3.3% (1) 13.3% (4) 0% (0) 0% (0) 13.3% (4) 20% (6) 23.3% (7) 16.6% (5) 0% (0) 0% (0) 7 3.3% (1) 0% (0) 33.3% (10) 20% (6) 43.3% (13) 66.6% (20) 10% (3) 6.6% (2) 0% (0) 0% (0) 0% (0) 0% (0) 10.0% (3) 6.6% (2) 0% (0) 0% (0) 8 0% (0) 0% (0) 6.6% (2) 3.3% (1) 70% (21) 73.3% (22) 10% (3) 10% (3) 0% (0) 0% (0) 0% (0) 3.3% (1) 13.3% (4) 10% (3) 0% (0) 0% (0) 9 0% (0) 0% (0) 20% (6) 20% (6) 16% (5) 26.6% (8) 20% (6) 13.3% (4) 0% (0) 0% (0) 20% (6) 30% (9) 23.3% (7) 10% (3) 0% (0) 0% (0) 10 0% (0) 0% (0) 80% (24) 66.6% (20) 16.6% (5) 33.3% (10) 3.3% (1) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) 0% (0) In Phase 1 an overall agreement of 60% was observed between DeepSeek and ChatGPT-4 (Table 2 ), in the same clinical scenarios, with a moderate concordance ( k = 0.420, p = 0.012). When comparing these results with those from the surgeons, an agreement of 60% was found between ChatGPT-4 and the surgeons with a fair concordance ( k = 0.344, p = 0.041) , and 70% between DeepSeek and the surgeons with a moderate concordance (k = 0.508, p = 0.003) (Table 4 ). Table 2 Responses provided by AI models and the majority of surgeons in Phase 1 of the study. (GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno–ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition). Case ChatGPT-4 (Phase 1) DeepSeek (Phase 1) Bariatric and Metabolic Surgeons group (Phase 1) (%) 1 No surgery GS GS (83.30%), RYGB (6.67%) 2 RYGB RYGB RYGB (76.70%), OAGB (13.3%) 3 GS RYGB RYGB (46.7%), GS (40%) 4 RYGB DS RYGB (50%), GS (26.67%) 5 RYGB RYGB RYGB (96.70%), ITB (3.33%) 6 SADI-S DS RYGB (33.3%), GS (26.67%) 7 RYGB RYGB RYGB (43.3%), GS (33.3%) 8 RYGB RYGB RYGB (70%), ITB (13.3%) 9 SADI-S SADI-S ITB (23.3%), SADI-S (20%), OAGB (20%), GS (20%) 10 GS GS GS (80%), RYGB (16.67%) Table 4 Cohen’s kappa coefficient between surgeons and AI models in both phases of the study. Comparatives Cohen’s Kappa coefficient Interpretation p Bariatric and Metabolic Surgeons Phase 1 vs. Bariatric and Metabolic Surgeons Phase 2 0.787 Substantial agreement < 0.001 ChatGPT-4 Phase 1 vs. ChatGPT-4 Phase 2 0.474 Moderate agreement 0.004 DeepSeek Phase 1 vs. DeepSeek Phase 2 0.167 Slight agreement 0.114 Bariatric and Metabolic Surgeons Phase 1 vs. ChatGPT-4 Phase 1 0.344 Fair agreement 0.041 Bariatric and Metabolic Surgeons Phase 1 vs. DeepSeek Phase 1 0.508 Moderate agreement 0.003 ChatGPT-4 Phase 1 vs. DeepSeek Phase 1 0.420 Moderate agreement 0.012 Bariatric and Metabolic Surgeons Phase 2 vs. ChatGPT-4 Phase 2 0.048 No agreement 0.812 Bariatric and Metabolic Surgeons Phase 2 vs. DeepSeek Phase 2 0.318 Fair agreement 0.016 ChatGPT-4 Phase 2 vs DeepSeek Phase 2 0.412 Moderate agreement 0.084 Regarding Phase 2, (Table 1 ), once again, the highest level of agreement was observed in Case 5, with 96.6%. When comparing surgeons’ responses between Phases 1 and 2, an overall agreement of 90% was observed, with the only discrepancy occurring in Case 9. In Phase 1, the most frequent selection for Case 9 was ITB (23.3%), whereas in Phase 2, SADI-S became the predominant choice (30%). According to the statistical analysis (Table 4 ), the comparison between surgeons’ responses in both phases showed a substantial concordance ( k = 0.787, p < 0.001). For the responses of surgeons and AI models in Phase 2 (Table 3 ), ChatGPT-4 demonstrated a 60% agreement with surgeons, but with no significant concordance ( k = 0.048, p = 0.812). In contrast, DeepSeek exhibited a 70% agreement with the surgeons, corresponding to a fair concordance ( k = 0.318, p = 0.016). Among the AI models, an 80% agreement was observed, with a moderate concordance (k = 0. 412, p = 0.084) (Table 4 ). Table 3 Responses provided by AI models in Phase 2 of the study. (GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno–ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition). Case ChatGPT-4 (Phase 2) DeepSeek (Phase 2) Bariatric and Metabolic Surgeons group (Phase 2) (%) 1 No surgery No surgery GS (76.70%), No surgery (13.3%) 2 RYGB RYGB RYGB (76.70%), OAGB (13.3%) 3 GS RYGB RYGB (56.70%), GS (26.6%) 4 RYGB RYGB RYGB (60%), GS (16.67%) 5 RYGB RYGB RYGB (96.70%), ITB (3.33%) 6 RYGB RYGB RYGB (30%), GS (20%), SADIS-S (20%) 7 RYGB RYGB RYGB (66.70%), GS (20%) 8 RYGB RYGB RYGB (73.3%), OAGB (10%), ITB (10%) 9 RYGB RYGB SADI-S (30%), RYGB (26.67%) 10 RYGB No surgery GS (66.70%), RYGB (33.3%) Between ChatGPT-4 responses from Phase 1 and Phase 2 (Table 4 ), an agreement of 70% was presented, corresponding to a moderate agreement ( k = 0.474, p = 0.004 ). For DeepSeek, agreement was lower between both phases, with a 50% agreement, corresponding to a slight concordance ( k = 0.167, p = 0.114) (Table 4 ). As an exploratory secondary analysis, the cases with discrepant recommendations between ChatGPT-4 and DeepSeek, limited to scenarios involving GS and RYGB, Case 3 in Phase 1, and Case 3 in Phase 2, were evaluated using the SOPHIA AI-based weight-loss prediction tool to estimate predicted excess weight loss (%EWL) at 6, 12 and 24 months after the proposed surgery (Table 5 ). Table 5 SOPHIA-predicted excess weight loss (%EWL) following GS or RYGB in cases with discordant recommendations between AI models. Phase / Case %EWL at 6 months (range) %EWL at 12 months (range) %EWL at 24 months (range) ChatGPT-4: GS DeepSeek: RYGB ChatGPT-4: GS DeepSeek: RYGB ChatGPT-4: GS DeepSeek: RYGB Phase 1 / Case 3 64% (53–72%) 60% (48–70%) 80% (63–91%) 84% (66–94% 80% (55–93%) 105% (79–116%) Phase 2 / Case 3 64% (53–72%) 60% (48–70%) 80% (63–91%) 84% (66–94% 80% (55–93%) 105% (79–116%) Discussion In our analysis, we compared the recommendations generated by both AI models with the surgical decisions made by participating bariatric and metabolic surgeons. In Phase 1 (without bibliographic support), ChatGPT-4 matched the majority surgical choice in 60% of the cases, showing fair agreement, while DeepSeek achieved the same coincidence rate but with moderate agreement. This finding represents a methodological strength of the present study, as it considers freely accessible AI models to explore their potential applicability in a specialized surgical field. In Phase 2, after providing access to high-quality scientific references, we explored whether bibliographic support improved the accuracy and consistency of AI recommendations. ChatGPT-4 maintained the same coincidence rate (60%), with no statistically significant concordance. In contrast, DeepSeek improved its performance, achieving fair significant concordance. These findings suggest that both models show some alignment with clinical reasoning, but DeepSeek may be more effective at incorporating external evidence into its decision-making, while ChatGPT-4 did not demonstrate a comparable improvement after exposure to the same literature. Is it possible that, after benign exposed to the bibliography exposure, the AI model focused more on providing the ‘right’ answer, rather than the most appropriate choice considering the patient’s context? It’s important to acknowledge that the training process of LLM is not fully transparent, and thus it cannot be conclusively verified whether ChatGPT-4 or DeepSeek had prior exposure to the bibliographic sources provided in Phase 2. In the other hand, the study design intentionally focused on evaluating how explicit evidence provision influences model outputs, independent of potential latent knowledge, and by standardizing the evidence supplied, Phase 2 represents a controlled simulation of evidence-guided clinical reasoning. The growing accessibility of AI models such as ChatGPT-4 and DeepSeek has sparked increasing interest in exploring their potential applications in clinical practice. In this context, Leng Yu et al. [ 15 ] compared ChatGPT-4’s responses with the consensus of three expert bariatric surgeons, who assessed the quality of the information based on international guidelines and consensus statements. In their study, ChatGPT-4 performed remarkably well, with 50% of its responses receiving the highest score for clinical guideline alignment. Our findings were consistent with this trend; however, our study introduced an additional AI model, DeepSeek, to compare the performance of different publicly available tools accessible to both clinicians and patients. This approach is particularly relevant because these systems are not specifically designed for medical use, and their responses may draw from sources of variable reliability. Our results are consistent with those reported by López-Gonzalez et al. [ 16 ], who found that although ChatGPT-4 can process large amounts of clinical data, its agreement with surgical decisions made by institutional algorithms or expert panels remains limited. Similarly, Sánchez-Cordero et al. [ 17 ] observed that ChatGPT-4 modified its recommendations after exposure to scientific literature, yet it did not fully align with real-world surgical practice. By including two different AI models and testing their performance before and after bibliographic exposure, our study adds a new dimension to understanding how evidence can shape AI-driven reasoning in complex surgical decision-making. When comparing both AI models with each other, a moderate concordance was observed in both phases, suggesting that, despite their different origins, architecture, and contextual training, both models preserved similar reasoning patterns when confronted with equivalent scenarios. DeepSeek presented greater variability between phases, indicating that its reasoning changed substantially after being exposed to additional information. In contrast, ChatGPT-4 maintained a reasonably stable decision patter, with only a partial adjustment of its recommendations following the introduction of bibliographic support. This observation is partially consistent with the findings of Sánchez-Cordero et al. [ 17 ], who evaluated the performances of ChatGPT-4 in selecting surgical weight-loss options before and after exposure to scientific literature, finding that after “contextual” training a shift in its recommendations is present, although it still failed to achieve strong concordance with real-world clinical practice. Among the surgical group, greater consistency was observed between phases, even with the addition of bibliography, suggesting that surgeons maintain a substantially higher degree of diagnostic reproducibility compared with the AI models. It is important to highlight that the observed variability among surgeons regarding the selection of the most appropriate procedure, may reflect the diversity of clinical judgment or may be driven by factors not related to scientific evidence, such as personal skills, experience, and surgical training, or the characteristics of the facilities and community where they practice. In the exploratory analysis, SOPHIA favored RYGB, with differences becoming more pronounced at longer-term points, particularly at 24 months. Although this secondary analysis was not designed to determine procedure superiority or predict real-world outcomes, it illustrates how disagreement between AI models may translate into clinically meaningful differences in long-term predicted weight-loss trajectories. These findings underscore the importance of contextualizing AI-generated recommendations with expert clinical judgement while acknowledging that this analysis was restricted to GS and RYGB, as these are the procedures supported by the SOPHIA model. Among the potential limitations of the study, one is that it is possible that some participants had not reviewed the provided references in depth before submitting their responses, which may have affected the consistency in guideline-based decision-making during Phase 2. On the other hand, one of the key strengths of this study is the comparison of two different AI models, other is the participation of a substantial number of surgeons from diverse geographic regions, which provides heterogeneity and broader representations, and finally that high-quality studies were selected as reference for constructing the clinical scenarios and as the evidence base for solving them. Conclusions SG and RYGB, remained the most frequently recommended procedures among bariatric and metabolic surgeons, while both AI models favored uniform, evidence-based recommendations predominantly supporting these techniques; DeepSeek showed greater alignment with surgeons’ consensus and improved concordance after bibliographic exposure, whereas ChatGPT-4, demonstrated no additional benefit and reduced concordance in Phase 2. Overall, these findings highlight the importance of ongoing efforts in contextual training, clinical validation, and human oversight to develop AI tools that are more accurate, safe and clinically applicable, serving as complementary resources than enhance, rather than replace, the clinical judgment of bariatric and metabolic surgeons. Declarations Author Contribution C.P and M.H wrote the main manuscript text, contributed equally to the conception and design of the study, development of the simulated clinical scenarios, data analysis and interpretation, manuscript drafting, critical revision, and final approval of the manuscript.H.S manuscript drafting, critical revision, and final approval of the manuscript.R.P data analysis and interpretation, manuscript drafting, critical revision, and final approval of the manuscript.All authors reviewed the manuscript. Acknowledgement The authors thank the expert bariatric and metabolic surgeons who participated in the evaluation of the clinical scenarios and provided their expert recommendations: Andrei Coria, Antonio Giovanni Spaventa Ibarrola, Azucena Reyes Pérez, Carlos Alberto Gutiérrez Rojas, Carlos González de Cosío Corredor, Carlos Valenzuela Salazar, Carlos Zerrweck López, Diana Gabriela Maldonado Pintado, Eduardo Jaramillo, Eduardo Vidrio Duarte, Ernesto Paez, Gilberto Ungson, Guillemo Ponce de León Ballesteros, Iliana González Pezzat, Israel Augusto González González, Itzel Fernández, Jesús Morales Maza, Juan Carlos Ramírez Almaral, Juan Francisco Arellano, Karla Carolina Flores Maciel, Lizbeth Guilbert, Luis Zorrilla, Luis Zurita Macías Valadez, Manuel Aceves Ávalos, Martha Patricia Sánchez Muñoz, Melba Rivera Félix, Miguel Ángel Zapata, Noé Nuñez Jasso, Roberto Estrada, Sergio Verboonen. References Turing AM. Computing machinery and intelligence. Mind. 1950;LIX(236):433–60. Jazi AHD, Mahjoubi M, Shahabi S, Alqahtani AR, Haddad A, Pazouki A, et al. Bariatric evaluation through AI: a survey of expert opinions versus ChatGPT-4 (BETA-SEOV). Obes Surg. 2023;33(12):3971–80. Bellini V, Valente M, Turetti M, Del Rio P, Saturno F, Maffezzoni M, et al. Current applications of artificial intelligence in bariatric surgery. Obes Surg. 2022;32(8):2717–33. Lee Y, Shin T, Tessier L, Javidan A, Jung J, Hong D, et al. Harnessing artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in generating clinician-level bariatric surgery recommendations. Surg Obes Relat Dis. 2024;20(7):603–8. Lee Y, Tessier L, Brar K, Malone S, Jin D, McKechnie T, et al. Performance of artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in the American Society for Metabolic and Bariatric Surgery textbook of bariatric surgery questions. Surg Obes Relat Dis. 2024;20(7):609–13. Saux P, Bauvin P, Raverdy V, Teigny J, Verkindt H, Soumphonphakdy T, et al. Development and validation of an interpretable machine learning–based calculator for predicting 5-year weight trajectories after bariatric surgery: a multinational retrospective cohort SOPHIA study. Lancet Digit Health. 2023;5(10):e692–702. Samaan JS, Rajeev N, Ng WH, Srinivasan N, Busam JA, Yeo YH, et al. ChatGPT as a source of information for bariatric surgery patients: a comparative analysis of accuracy and comprehensiveness between GPT-4 and GPT-3.5. Obes Surg. 2024;34(5):1987–9. Schauer PR, Bhatt DL, Kirwan JP, Wolski K, Aminian A, Brethauer SA, et al. Bariatric surgery versus intensive medical therapy for diabetes—5-year outcomes. N Engl J Med. 2017;376(7):641–51. Schiavon CA, Bersch-Ferreira AC, Santucci EV, Oliveira JD, Torreglosa CR, Bueno PT, et al. Effects of bariatric surgery in obese patients with hypertension: the GATEWAY randomized trial (gastric bypass to treat obese patients with steady hypertension). Circulation. 2018;137(11):1132–42. Ikramuddin S, Korner J, Lee WJ, Connett JE, Inabnet WB, Billington CJ, et al. Roux-en-Y gastric bypass vs intensive medical management for the control of type 2 diabetes, hypertension, and hyperlipidemia: the Diabetes Surgery Study randomized clinical trial. JAMA. 2013;309(21):2240–9. Grönroos S, Helmiö M, Juuti A, Tiusanen R, Hurme S, Löyttyniemi E, et al. Effect of laparoscopic sleeve gastrectomy vs Roux-en-Y gastric bypass on weight loss and quality of life at 7 years in patients with morbid obesity: the SLEEVEPASS randomized clinical trial. JAMA Surg. 2021;156(2):137–46. Verrastro O, Panunzi S, Castagneto-Gissey L, De Gaetano A, Lembo E, Capristo E, et al. Bariatric-metabolic surgery versus lifestyle intervention plus best medical care in non-alcoholic steatohepatitis (BRAVES): a multicentre, open-label, randomised trial. Lancet. 2023;401(10390):1786–97. Małczak P, Mizera M, Lee Y, Pisarska-Adamczyk M, Wysocki M, Bała M, Witowski J, Rubinkiewicz M, Dudek A, Stefura T, Torbicz G, Tylec P, Gajewska N, Vongsurbchart T, Su M, Major P, Pędziwiatr M. Quality of Life After Bariatric Surgery—a Systematic Review with Bayesian Network Meta-analysis. Obes Surg. 2021;31(12):5213–23. 10.1007/s11695-021-05687-1 . Schauer PR, Kashyap SR, Wolski K, Brethauer SA, Kirwan JP, Pothier CE, et al. Bariatric surgery versus intensive medical therapy in obese patients with diabetes. N Engl J Med. 2012;366(17):1567–76. Leng Y, Yang Y, Liu J, Jiang J, Zhou C. Evaluating the feasibility of ChatGPT-4 as a knowledge resource in bariatric surgery: a preliminary assessment. Obes Surg. 2025;35:645–50. 10.1007/s11695-024-07666-8 . López-Gonzalez R, Sanchez-Cordero S, Pujol-Gebellí J, Castellvi J. Evaluation of the impact of ChatGPT on the selection of surgical technique in bariatric surgery. Obes Surg. 2025;35:19–24. 10.1007/s11695-024-07279-1 . Sánchez-Cordero S, López-Gonzalez R, Fernández H, Pujol-Gebellí J. Training ChatGPT for surgical decisions: bariatric surgery analysis using algorithms and evidence. Obes Res Clin Pract. 2025;19:352–5. 10.1016/j.orcp.2025.08.002 . Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial1and2.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8427770","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":577104021,"identity":"741555e7-2545-43cb-b228-430adfa4d776","order_by":0,"name":"Carlos J Pérez-Padrón","email":"","orcid":"","institution":"Centro Médico ABC","correspondingAuthor":false,"prefix":"","firstName":"Carlos","middleName":"J","lastName":"Pérez-Padrón","suffix":""},{"id":577104024,"identity":"662f06e2-b72f-437f-835b-009b33b63b24","order_by":1,"name":"Hugo Sánchez-Aguilar","email":"","orcid":"","institution":"Centro Médico ABC","correspondingAuthor":false,"prefix":"","firstName":"Hugo","middleName":"","lastName":"Sánchez-Aguilar","suffix":""},{"id":577104030,"identity":"f9099572-bb9f-48d4-b9d1-4138f6ea6d7a","order_by":2,"name":"Rafael H Pérez-Soto","email":"","orcid":"","institution":"Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán","correspondingAuthor":false,"prefix":"","firstName":"Rafael","middleName":"H","lastName":"Pérez-Soto","suffix":""},{"id":577104031,"identity":"d6b9477a-9c4b-47af-a190-8acadc523761","order_by":3,"name":"Miguel F Herrera","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/UlEQVRIiWNgGAWjYDACZhBRIJcA5doAMWPjAcJaDIxhWtJAWhrwa2FA1XIYTOLVotvOfk3ig4FBHr/02YePK2rO261tPwy0pcYmGpcWs8M8ZZIzDAyKJfvSjQ3PHLudvO1MIlDLsbTcBtxa0qR5DP4kbjjDxibZwHY72ewAUAtjw2H8Wv4YGCTuP8PG/rPh37lks/MPCWlhPybNANSygYeNjbGx7YCd2Q3CtjBb9gC1zDjDxizZ2JecYHYDaEsCPr+cP/7wxo8Kg8T+HjbGjw3f7OzNzqc/fPChxganFgYGHgMUbiJYZQJO5SDA/gCFa49X8SgYBaNgFIxIAACr3mECqd/8bgAAAABJRU5ErkJggg==","orcid":"","institution":"Centro Médico ABC","correspondingAuthor":true,"prefix":"","firstName":"Miguel","middleName":"F","lastName":"Herrera","suffix":""}],"badges":[],"createdAt":"2025-12-22 18:38:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8427770/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8427770/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100690633,"identity":"c900252c-56e6-437c-b92f-5d8bf08c1ca0","added_by":"auto","created_at":"2026-01-20 13:55:34","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38921,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptAIinBMS.docx","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/6dcc7e0147958dc19048cf83.docx"},{"id":100690632,"identity":"d4c26755-b680-4b96-9db8-66ef0387b835","added_by":"auto","created_at":"2026-01-20 13:55:32","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7208,"visible":true,"origin":"","legend":"","description":"","filename":"1287c9f5006f46cab65a65744e019f52.json","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/85192dd240a0a573d2851111.json"},{"id":100690953,"identity":"b4bd3002-c7b5-44ea-a1d8-7f0478f2f068","added_by":"auto","created_at":"2026-01-20 14:00:33","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17942,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial1and2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/d098f3e12e62480b77c38f7d.docx"},{"id":100690942,"identity":"9f07c2c4-ef1d-4b8d-b9c8-7ff786c33181","added_by":"auto","created_at":"2026-01-20 14:00:04","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":93096,"visible":true,"origin":"","legend":"","description":"","filename":"1287c9f5006f46cab65a65744e019f521enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/6679bd0c2ab8becc380d22aa.xml"},{"id":100690531,"identity":"67512217-5241-493b-8d51-6b5560717f4c","added_by":"auto","created_at":"2026-01-20 13:54:09","extension":"xml","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":92493,"visible":true,"origin":"","legend":"","description":"","filename":"1287c9f5006f46cab65a65744e019f521structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/44dfbd7d1c97a7819596d69c.xml"},{"id":100690542,"identity":"6103f0b6-be1a-4ddb-9147-ae1d8234f4e2","added_by":"auto","created_at":"2026-01-20 13:54:15","extension":"html","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100613,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/a55df8662e48cacb2764d767.html"},{"id":104145500,"identity":"a8ea169e-3462-4211-bf3c-5b3b745cf981","added_by":"auto","created_at":"2026-03-07 22:08:56","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":891289,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/da303428-146e-415e-9a1a-461b11f4760d.pdf"},{"id":100690729,"identity":"00e20ddb-bad6-4d45-b3f2-84a8d56fcb23","added_by":"auto","created_at":"2026-01-20 13:57:20","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":17942,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial1and2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8427770/v1/ea2ec58f8a1fbe621504e108.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Exploring the role of Artificial Intelligence in Bariatric and Metabolic Surgery","fulltext":[{"header":"Introduction","content":"\u003cp\u003eArtificial intelligence (AI) represents a sustained human endeavor to enhance the quality of life and meet fundamental societal needs. First conceptualized by John McCarthy in 1956 [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], AI has evolved rapidly with the era of digital globalization, becoming increasingly integrated into everyday life. Beyond improving the efficiency and accuracy of diverse services through large-scale data processing, AI now enables the detection of complex patterns beyond human analytical capacity, particularly in clinical imaging, predictive modeling, and the development of decision support systems that assist physicians in diagnosis and clinical management.\u003c/p\u003e \u003cp\u003eIn the field of bariatric and metabolic surgery (BMS), the integration of AI is no exception. A notable early example is the study by Jazi et al. [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], which compared the recommendations generated by a large language model (ChatGPT-4) with those provided by bariatric and metabolic surgeons for selected clinical scenarios. The authors reported a concordance rate of approximately 30% between both groups, which was considered poor, underscoring the critical importance of human clinical judgment in the evaluation of candidates for weight-loss surgery and in the selection of the most appropriate surgical procedure.\u003c/p\u003e \u003cp\u003eAnother potential application of AI lies across different stages of the surgical process, including preoperative planning, intraoperative guidance, postoperative assessment and outcome prediction [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Yung Lee et al. [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] compared the performance of three AI models, ChatGPT-4, Bing, and Bard, in responding to clinical questions related to bariatric and metabolic surgery, finding that ChatGPT-4 provided the highest proportion of correct answers compared with the other models. In a subsequent study, Yung Lee et al. [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] evaluated the same three models using questions derived from the \u003cem\u003eAmerican Society for Metabolic and Bariatric Surgery (ASMBS) Textbook of Bariatric Surgery\u003c/em\u003e, reporting accuracies of 83.0% for ChatGPT-4, 76.0% for Bard, and 65.0% for Bing. These results demonstrate a relatively high level of accuracy for ChatGPT-4, suggesting its potential reliability as an educational and decision-support tool in BMS.\u003c/p\u003e \u003cp\u003eThe SOPHIA AI-based weight-loss prediction tool [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], was developed using supervised machine learning techniques, including LASSO regression and decision tree-based algorithms. Trained on data from more than 10,000 bariatric patients across Europe, the model analyzed hundreds of baseline variables and identified seven key predictors of postoperative weight trajectory. It was externally validated to estimate individual weight outcomes up to five years after surgery. This AI driven approach enables personalized, data informed predictions that support clinical decision-making in BMS. To date, no studies have evaluated whether different AI models generate distinct clinical recommendations based on there original training or whether these recommendations change after contextual bibliographic training. Given the widespread public accessibility of these tools, comparing AI outputs is essential to understand the recommendations that patients may receive and the extent to which such responses can be modulated by high-quality evidence.\u003c/p\u003e \u003cp\u003eThe performance and responses of AI models depend on the information to which they have been exposed and their capacity to learn. This training process is crucial for developing greater fluency and accuracy in answering complex queries. Comparing different versions of the same model can reveal marked variations in precision, reasoning, and alignment with scientific evidence. This phenomenon has been described, for instance, when comparing the performance of ChatGPT-3 and ChatGPT-4, where a noticeable improvement in coherence and accuracy was observed in clinical question responses [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Within this context, a new question arises: can artificial intelligence truly optimize the quality of surgical decision-making in BMS by enhancing medical and surgical reasoning? The principal objective of this study was to compare surgical recommendations made by AI models (ChatGPT-4 and Deepseek) with those of bariatric and metabolic surgeons in representative obesity clinical scenarios based on high-quality evidence.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003eTen representative clinical scenarios of patients with obesity were designed, incorporating variables such as age, sex, body mass index (BMI), and the presence of obesity-associated comorbidities. The creation of these scenarios was based on the highest-quality evidence currently available [\u003cspan additionalcitationids=\"CR9 CR10 CR11 CR12 CR13\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], primarily derived from clinical trials in which bariatric and metabolic surgery demonstrated a significant impact on improvement or resolution of obesity-associated comorbidities. The ten simulated clinical scenarios used for surgeon and AI evaluation are provided in Supplementary Material 2.\u003c/p\u003e \u003cp\u003eTwo AI models, ChatGPT-4 and DeepSeek, were asked to determine whether each hypothetical patient was a candidate for BMS from the perspective of an expert surgeon. When surgery was indicated, the models were further requested to recommend the most appropriate surgical procedure among the following options: gastric sleeve (GS), Roux-en-Y gastric bypass (RYGB), one-anastomosis gastric bypass (OAGB), biliopancreatic diversion with duodenal switch (BPD-DS), single-anastomosis duodeno\u0026ndash;ileal bypass with sleeve gastrectomy (SADI-S), intestinal transit bipartition (ITB), or other.\u003c/p\u003e \u003cp\u003eRecommendations were obtained under two different conditions: without access to bibliographic references (\u003cem\u003ePhase 1\u003c/em\u003e) and with access to key clinical references (\u003cem\u003ePhase 2\u003c/em\u003e) [\u003cspan additionalcitationids=\"CR9 CR10 CR11 CR12 CR13\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], aiming to analyze the impact of bibliographic support on the surgical decision-making process of AI models.\u003c/p\u003e \u003cp\u003eThe same ten clinical scenarios were distributed to board-certified bariatric and metabolic surgeons, listed in Supplementary Material 1, through Google Forms, under the same conditions previously described. In Phase 1, participants were individually asked to determine whether each hypothetical patient was a candidate for bariatric surgery and, if so, to specify the most appropriate procedure. In Phase 2, 4 weeks after the prior phase, the same group of surgeons received the bibliographic references and, after reviewing and analyzing them, were asked to provide updated surgical recommendations for the same cases. In both phases, all responses were collected anonymously employing an anonymized matching strategy using non-identifiable demographic variables, with participants blinded to the answers of their peers and those generated by the AI models.\u003c/p\u003e \u003cp\u003eA descriptive analysis was performed on the responses obtained from both the surgeons and the AI models to identify patterns and trends. The consensus answer from the surgical group was defined as the option selected by the majority and was considered the reference standard for comparison with the AI models. To assess the concordance between surgeons\u0026rsquo; and AI models\u0026rsquo; recommendations, Cohen\u0026rsquo;s kappa coefficient was estimated, with a \u003cem\u003ep\u003c/em\u003e value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 considered statistically significant. All analyses were performed using SPSS software, version 30 (IBM, USA).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eFor Phase 1 of the study, 43 bariatric and metabolic surgeons were invited to participate. Only the 30 surgeons who participated in both phases were included in the final analysis. The sample consisted of 74.2% men and 25.8% women. The predominant age group was between 40 and 50 years (38.7%). Regarding to geographical distribution, 42% of participants reported practicing in Mexico City, followed by Jalisco and Nuevo Le\u0026oacute;n.\u003c/p\u003e \u003cp\u003eThe responses to the clinical scenarios from Phase 1 are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Agreement among surgeons was variable, with the highest agreement observed in Case 5, where 96.6% (n\u0026thinsp;=\u0026thinsp;29) selected RYGB as the most appropriate surgical option. Conversely, the lowest agreement was noted in Case 9, with only 23.3% of the surgeons selecting ITB.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of responses provided by Bariatric and Metabolic Surgeons in Phase 1 and Phase 2 of the study. \u003cem\u003e(GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno\u0026ndash;ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition).\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"17\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c13\" colnum=\"13\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c14\" colnum=\"14\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c15\" colnum=\"15\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c16\" colnum=\"16\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c17\" colnum=\"17\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase / Surgical option\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eNo surgery (N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eOAGB\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c11\" namest=\"c10\"\u003e \u003cp\u003eDS\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c13\" namest=\"c12\"\u003e \u003cp\u003eSADI-S\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c15\" namest=\"c14\"\u003e \u003cp\u003eITB\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c17\" namest=\"c16\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003cp\u003e(N)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c13\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c14\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c15\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c16\"\u003e \u003cp\u003ePhase 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c17\"\u003e \u003cp\u003ePhase 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e83.3% (25)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e76.6% (23)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e76.6% (23)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e76.6% (23)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40% (12)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e26.6% (8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e46.6% (14)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e56.6% (17)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6.6% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e26.6% (8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e16.6% (5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e50% (15)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e60% (18)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e13.3% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e96.6% (29)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e96.6% (29)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e6\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e26.6% (8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e33.3% (10)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e30% (9)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e23.3% (7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e16.6% (5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e7\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e33.3% (10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e43.3% (13)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e66.6% (20)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e10.0% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e8\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.6% (2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e70% (21)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e73.3% (22)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e9\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16% (5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e26.6% (8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e13.3% (4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e20% (6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e\u003cb\u003e30% (9)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e\u003cb\u003e23.3% (7)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e10% (3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e10\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e80% (24)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e66.6% (20)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16.6% (5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e33.3% (10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3.3% (1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c11\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c12\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c13\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c14\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c15\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c16\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c17\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003cp\u003e(0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn Phase 1 an overall agreement of 60% was observed between DeepSeek and ChatGPT-4 (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), in the same clinical scenarios, with a moderate concordance (\u003cem\u003ek\u0026thinsp;=\u0026thinsp;0.420, p\u0026thinsp;=\u0026thinsp;0.012).\u003c/em\u003e When comparing these results with those from the surgeons, an agreement of 60% was found between ChatGPT-4 and the surgeons with a fair concordance (\u003cem\u003ek\u0026thinsp;=\u0026thinsp;0.344, p\u0026thinsp;=\u0026thinsp;0.041)\u003c/em\u003e, and 70% between DeepSeek and the surgeons with a moderate concordance \u003cem\u003e(k\u0026thinsp;=\u0026thinsp;0.508, p\u0026thinsp;=\u0026thinsp;0.003)\u003c/em\u003e (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eResponses provided by AI models and the majority of surgeons in Phase 1 of the study. \u003cem\u003e(GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno\u0026ndash;ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition).\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-4 (Phase 1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDeepSeek\u003c/p\u003e \u003cp\u003e(Phase 1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons group\u003c/p\u003e \u003cp\u003e(Phase 1) (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo surgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGS (83.30%), RYGB (6.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (76.70%), OAGB (13.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (46.7%), GS (40%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (50%), GS (26.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (96.70%), ITB (3.33%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e6\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSADI-S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (33.3%), GS (26.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e7\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (43.3%), GS (33.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e8\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (70%), ITB (13.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e9\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSADI-S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSADI-S\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eITB (23.3%), SADI-S (20%), OAGB (20%), GS (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e10\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGS (80%), RYGB (16.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCohen\u0026rsquo;s kappa coefficient between surgeons and AI models in both phases of the study.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComparatives\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohen\u0026rsquo;s Kappa\u003c/p\u003e \u003cp\u003ecoefficient\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons Phase 1 vs. Bariatric and Metabolic Surgeons Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.787\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSubstantial agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChatGPT-4 Phase 1 vs. ChatGPT-4 Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.474\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModerate agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.004\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDeepSeek Phase 1 vs. DeepSeek Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.167\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSlight agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.114\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons Phase 1 vs. ChatGPT-4 Phase 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.344\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFair agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons Phase 1 vs. DeepSeek Phase 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.508\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModerate agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.003\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChatGPT-4 Phase 1 vs. DeepSeek Phase 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.420\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModerate agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons Phase 2 vs. ChatGPT-4 Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.048\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.812\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons Phase 2 vs. DeepSeek Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.318\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFair agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.016\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChatGPT-4 Phase 2 vs DeepSeek Phase 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.412\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModerate agreement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.084\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eRegarding Phase 2, (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), once again, the highest level of agreement was observed in Case 5, with 96.6%. When comparing surgeons\u0026rsquo; responses between Phases 1 and 2, an overall agreement of 90% was observed, with the only discrepancy occurring in Case 9. In Phase 1, the most frequent selection for Case 9 was ITB (23.3%), whereas in Phase 2, SADI-S became the predominant choice (30%). According to the statistical analysis (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e), the comparison between surgeons\u0026rsquo; responses in both phases showed a substantial concordance (\u003cem\u003ek\u0026thinsp;=\u0026thinsp;0.787, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFor the responses of surgeons and AI models in Phase 2 (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e), ChatGPT-4 demonstrated a 60% agreement with surgeons, but with no significant concordance (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.048, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.812). In contrast, DeepSeek exhibited a 70% agreement with the surgeons, corresponding to a fair concordance (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.318, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.016). Among the AI models, an 80% agreement was observed, with a moderate concordance \u003cem\u003e(k\u0026thinsp;=\u0026thinsp;0. 412, p\u0026thinsp;=\u0026thinsp;0.084)\u003c/em\u003e (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eResponses provided by AI models in Phase 2 of the study. \u003cem\u003e(GS, gastric sleeve; RYGB, Roux-en-Y gastric bypass; OAGB, one-anastomosis gastric bypass; DS, biliopancreatic diversion with duodenal switch; SADI-S, single-anastomosis duodeno\u0026ndash;ileal bypass with sleeve gastrectomy; ITB, intestinal transit bipartition).\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCase\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-4\u003c/p\u003e \u003cp\u003e(Phase 2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDeepSeek\u003c/p\u003e \u003cp\u003e(Phase 2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBariatric and Metabolic Surgeons group\u003c/p\u003e \u003cp\u003e(Phase 2) (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo surgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo surgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGS (76.70%), No surgery (13.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (76.70%), OAGB (13.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (56.70%), GS (26.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (60%), GS (16.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (96.70%), ITB (3.33%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e6\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (30%), GS (20%), SADIS-S (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e7\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (66.70%), GS (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e8\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRYGB (73.3%), OAGB (10%), ITB (10%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e9\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSADI-S (30%), RYGB (26.67%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e10\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo surgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGS (66.70%), RYGB (33.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eBetween ChatGPT-4 responses from Phase 1 and Phase 2 (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e), an agreement of 70% was presented, corresponding to a moderate agreement (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.474, \u003cem\u003ep\u0026thinsp;=\u0026thinsp;0.004\u003c/em\u003e). For DeepSeek, agreement was lower between both phases, with a 50% agreement, corresponding to a slight concordance (\u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.167, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.114) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAs an exploratory secondary analysis, the cases with discrepant recommendations between ChatGPT-4 and DeepSeek, limited to scenarios involving GS and RYGB, Case 3 in Phase 1, and Case 3 in Phase 2, were evaluated using the SOPHIA AI-based weight-loss prediction tool to estimate predicted excess weight loss (%EWL) at 6, 12 and 24 months after the proposed surgery (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSOPHIA-predicted excess weight loss (%EWL) following GS or RYGB in cases with discordant recommendations between AI models.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhase / Case\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e%EWL at 6 months (range)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e%EWL at 12 months (range)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e%EWL at 24 months (range)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-4: GS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDeepSeek:\u003c/p\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eChatGPT-4: GS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDeepSeek:\u003c/p\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eChatGPT-4: GS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eDeepSeek:\u003c/p\u003e \u003cp\u003eRYGB\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhase 1 / Case 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64% (53\u0026ndash;72%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60% (48\u0026ndash;70%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80% (63\u0026ndash;91%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e84% (66\u0026ndash;94%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e80% (55\u0026ndash;93%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e105% (79\u0026ndash;116%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhase 2 / Case 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64% (53\u0026ndash;72%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60% (48\u0026ndash;70%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e80% (63\u0026ndash;91%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e84% (66\u0026ndash;94%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e80% (55\u0026ndash;93%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e105% (79\u0026ndash;116%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn our analysis, we compared the recommendations generated by both AI models with the surgical decisions made by participating bariatric and metabolic surgeons. In Phase 1 (without bibliographic support), ChatGPT-4 matched the majority surgical choice in 60% of the cases, showing fair agreement, while DeepSeek achieved the same coincidence rate but with moderate agreement. This finding represents a methodological strength of the present study, as it considers freely accessible AI models to explore their potential applicability in a specialized surgical field.\u003c/p\u003e \u003cp\u003eIn Phase 2, after providing access to high-quality scientific references, we explored whether bibliographic support improved the accuracy and consistency of AI recommendations. ChatGPT-4 maintained the same coincidence rate (60%), with no statistically significant concordance. In contrast, DeepSeek improved its performance, achieving fair significant concordance. These findings suggest that both models show some alignment with clinical reasoning, but DeepSeek may be more effective at incorporating external evidence into its decision-making, while ChatGPT-4 did not demonstrate a comparable improvement after exposure to the same literature. Is it possible that, after benign exposed to the bibliography exposure, the AI model focused more on providing the \u0026lsquo;right\u0026rsquo; answer, rather than the most appropriate choice considering the patient\u0026rsquo;s context? It\u0026rsquo;s important to acknowledge that the training process of LLM is not fully transparent, and thus it cannot be conclusively verified whether ChatGPT-4 or DeepSeek had prior exposure to the bibliographic sources provided in Phase 2. In the other hand, the study design intentionally focused on evaluating how explicit evidence provision influences model outputs, independent of potential latent knowledge, and by standardizing the evidence supplied, Phase 2 represents a controlled simulation of evidence-guided clinical reasoning.\u003c/p\u003e \u003cp\u003eThe growing accessibility of AI models such as ChatGPT-4 and DeepSeek has sparked increasing interest in exploring their potential applications in clinical practice. In this context, Leng Yu et al. [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] compared ChatGPT-4\u0026rsquo;s responses with the consensus of three expert bariatric surgeons, who assessed the quality of the information based on international guidelines and consensus statements. In their study, ChatGPT-4 performed remarkably well, with 50% of its responses receiving the highest score for clinical guideline alignment. Our findings were consistent with this trend; however, our study introduced an additional AI model, DeepSeek, to compare the performance of different publicly available tools accessible to both clinicians and patients. This approach is particularly relevant because these systems are not specifically designed for medical use, and their responses may draw from sources of variable reliability.\u003c/p\u003e \u003cp\u003eOur results are consistent with those reported by L\u0026oacute;pez-Gonzalez et al. [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], who found that although ChatGPT-4 can process large amounts of clinical data, its agreement with surgical decisions made by institutional algorithms or expert panels remains limited. Similarly, S\u0026aacute;nchez-Cordero et al. [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] observed that ChatGPT-4 modified its recommendations after exposure to scientific literature, yet it did not fully align with real-world surgical practice. By including two different AI models and testing their performance before and after bibliographic exposure, our study adds a new dimension to understanding how evidence can shape AI-driven reasoning in complex surgical decision-making.\u003c/p\u003e \u003cp\u003eWhen comparing both AI models with each other, a moderate concordance was observed in both phases, suggesting that, despite their different origins, architecture, and contextual training, both models preserved similar reasoning patterns when confronted with equivalent scenarios. DeepSeek presented greater variability between phases, indicating that its reasoning changed substantially after being exposed to additional information. In contrast, ChatGPT-4 maintained a reasonably stable decision patter, with only a partial adjustment of its recommendations following the introduction of bibliographic support. This observation is partially consistent with the findings of S\u0026aacute;nchez-Cordero et al. [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], who evaluated the performances of ChatGPT-4 in selecting surgical weight-loss options before and after exposure to scientific literature, finding that after \u0026ldquo;contextual\u0026rdquo; training a shift in its recommendations is present, although it still failed to achieve strong concordance with real-world clinical practice.\u003c/p\u003e \u003cp\u003eAmong the surgical group, greater consistency was observed between phases, even with the addition of bibliography, suggesting that surgeons maintain a substantially higher degree of diagnostic reproducibility compared with the AI models.\u003c/p\u003e \u003cp\u003eIt is important to highlight that the observed variability among surgeons regarding the selection of the most appropriate procedure, may reflect the diversity of clinical judgment or may be driven by factors not related to scientific evidence, such as personal skills, experience, and surgical training, or the characteristics of the facilities and community where they practice.\u003c/p\u003e \u003cp\u003eIn the exploratory analysis, SOPHIA favored RYGB, with differences becoming more pronounced at longer-term points, particularly at 24 months. Although this secondary analysis was not designed to determine procedure superiority or predict real-world outcomes, it illustrates how disagreement between AI models may translate into clinically meaningful differences in long-term predicted weight-loss trajectories. These findings underscore the importance of contextualizing AI-generated recommendations with expert clinical judgement while acknowledging that this analysis was restricted to GS and RYGB, as these are the procedures supported by the SOPHIA model.\u003c/p\u003e \u003cp\u003e Among the potential limitations of the study, one is that it is possible that some participants had not reviewed the provided references in depth before submitting their responses, which may have affected the consistency in guideline-based decision-making during Phase 2. On the other hand, one of the key strengths of this study is the comparison of two different AI models, other is the participation of a substantial number of surgeons from diverse geographic regions, which provides heterogeneity and broader representations, and finally that high-quality studies were selected as reference for constructing the clinical scenarios and as the evidence base for solving them.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eSG and RYGB, remained the most frequently recommended procedures among bariatric and metabolic surgeons, while both AI models favored uniform, evidence-based recommendations predominantly supporting these techniques; DeepSeek showed greater alignment with surgeons\u0026rsquo; consensus and improved concordance after bibliographic exposure, whereas ChatGPT-4, demonstrated no additional benefit and reduced concordance in Phase 2. Overall, these findings highlight the importance of ongoing efforts in contextual training, clinical validation, and human oversight to develop AI tools that are more accurate, safe and clinically applicable, serving as complementary resources than enhance, rather than replace, the clinical judgment of bariatric and metabolic surgeons.\u003c/p\u003e "},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eC.P and M.H wrote the main manuscript text, contributed equally to the conception and design of the study, development of the simulated clinical scenarios, data analysis and interpretation, manuscript drafting, critical revision, and final approval of the manuscript.H.S manuscript drafting, critical revision, and final approval of the manuscript.R.P data analysis and interpretation, manuscript drafting, critical revision, and final approval of the manuscript.All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors thank the expert bariatric and metabolic surgeons who participated in the evaluation of the clinical scenarios and provided their expert recommendations: Andrei Coria, Antonio Giovanni Spaventa Ibarrola, Azucena Reyes P\u0026eacute;rez, Carlos Alberto Guti\u0026eacute;rrez Rojas, Carlos Gonz\u0026aacute;lez de Cos\u0026iacute;o Corredor, Carlos Valenzuela Salazar, Carlos Zerrweck L\u0026oacute;pez, Diana Gabriela Maldonado Pintado, Eduardo Jaramillo, Eduardo Vidrio Duarte, Ernesto Paez, Gilberto Ungson, Guillemo Ponce de Le\u0026oacute;n Ballesteros, Iliana Gonz\u0026aacute;lez Pezzat, Israel Augusto Gonz\u0026aacute;lez Gonz\u0026aacute;lez, Itzel Fern\u0026aacute;ndez, Jes\u0026uacute;s Morales Maza, Juan Carlos Ram\u0026iacute;rez Almaral, Juan Francisco Arellano, Karla Carolina Flores Maciel, Lizbeth Guilbert, Luis Zorrilla, Luis Zurita Mac\u0026iacute;as Valadez, Manuel Aceves \u0026Aacute;valos, Martha Patricia S\u0026aacute;nchez Mu\u0026ntilde;oz, Melba Rivera F\u0026eacute;lix, Miguel \u0026Aacute;ngel Zapata, No\u0026eacute; Nu\u0026ntilde;ez Jasso, Roberto Estrada, Sergio Verboonen.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eTuring AM. Computing machinery and intelligence. Mind. 1950;LIX(236):433\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJazi AHD, Mahjoubi M, Shahabi S, Alqahtani AR, Haddad A, Pazouki A, et al. Bariatric evaluation through AI: a survey of expert opinions versus ChatGPT-4 (BETA-SEOV). Obes Surg. 2023;33(12):3971\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBellini V, Valente M, Turetti M, Del Rio P, Saturno F, Maffezzoni M, et al. Current applications of artificial intelligence in bariatric surgery. Obes Surg. 2022;32(8):2717\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee Y, Shin T, Tessier L, Javidan A, Jung J, Hong D, et al. Harnessing artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in generating clinician-level bariatric surgery recommendations. Surg Obes Relat Dis. 2024;20(7):603\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee Y, Tessier L, Brar K, Malone S, Jin D, McKechnie T, et al. Performance of artificial intelligence in bariatric surgery: comparative analysis of ChatGPT-4, Bing, and Bard in the American Society for Metabolic and Bariatric Surgery textbook of bariatric surgery questions. Surg Obes Relat Dis. 2024;20(7):609\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaux P, Bauvin P, Raverdy V, Teigny J, Verkindt H, Soumphonphakdy T, et al. Development and validation of an interpretable machine learning\u0026ndash;based calculator for predicting 5-year weight trajectories after bariatric surgery: a multinational retrospective cohort SOPHIA study. Lancet Digit Health. 2023;5(10):e692\u0026ndash;702.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSamaan JS, Rajeev N, Ng WH, Srinivasan N, Busam JA, Yeo YH, et al. ChatGPT as a source of information for bariatric surgery patients: a comparative analysis of accuracy and comprehensiveness between GPT-4 and GPT-3.5. Obes Surg. 2024;34(5):1987\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchauer PR, Bhatt DL, Kirwan JP, Wolski K, Aminian A, Brethauer SA, et al. Bariatric surgery versus intensive medical therapy for diabetes\u0026mdash;5-year outcomes. N Engl J Med. 2017;376(7):641\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchiavon CA, Bersch-Ferreira AC, Santucci EV, Oliveira JD, Torreglosa CR, Bueno PT, et al. Effects of bariatric surgery in obese patients with hypertension: the GATEWAY randomized trial (gastric bypass to treat obese patients with steady hypertension). Circulation. 2018;137(11):1132\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIkramuddin S, Korner J, Lee WJ, Connett JE, Inabnet WB, Billington CJ, et al. Roux-en-Y gastric bypass vs intensive medical management for the control of type 2 diabetes, hypertension, and hyperlipidemia: the Diabetes Surgery Study randomized clinical trial. JAMA. 2013;309(21):2240\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGr\u0026ouml;nroos S, Helmi\u0026ouml; M, Juuti A, Tiusanen R, Hurme S, L\u0026ouml;yttyniemi E, et al. Effect of laparoscopic sleeve gastrectomy vs Roux-en-Y gastric bypass on weight loss and quality of life at 7 years in patients with morbid obesity: the SLEEVEPASS randomized clinical trial. JAMA Surg. 2021;156(2):137\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVerrastro O, Panunzi S, Castagneto-Gissey L, De Gaetano A, Lembo E, Capristo E, et al. Bariatric-metabolic surgery versus lifestyle intervention plus best medical care in non-alcoholic steatohepatitis (BRAVES): a multicentre, open-label, randomised trial. Lancet. 2023;401(10390):1786\u0026ndash;97.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMałczak P, Mizera M, Lee Y, Pisarska-Adamczyk M, Wysocki M, Bała M, Witowski J, Rubinkiewicz M, Dudek A, Stefura T, Torbicz G, Tylec P, Gajewska N, Vongsurbchart T, Su M, Major P, Pędziwiatr M. Quality of Life After Bariatric Surgery\u0026mdash;a Systematic Review with Bayesian Network Meta-analysis. Obes Surg. 2021;31(12):5213\u0026ndash;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11695-021-05687-1\u003c/span\u003e\u003cspan address=\"10.1007/s11695-021-05687-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchauer PR, Kashyap SR, Wolski K, Brethauer SA, Kirwan JP, Pothier CE, et al. Bariatric surgery versus intensive medical therapy in obese patients with diabetes. N Engl J Med. 2012;366(17):1567\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeng Y, Yang Y, Liu J, Jiang J, Zhou C. Evaluating the feasibility of ChatGPT-4 as a knowledge resource in bariatric surgery: a preliminary assessment. Obes Surg. 2025;35:645\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11695-024-07666-8\u003c/span\u003e\u003cspan address=\"10.1007/s11695-024-07666-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL\u0026oacute;pez-Gonzalez R, Sanchez-Cordero S, Pujol-Gebell\u0026iacute; J, Castellvi J. Evaluation of the impact of ChatGPT on the selection of surgical technique in bariatric surgery. Obes Surg. 2025;35:19\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11695-024-07279-1\u003c/span\u003e\u003cspan address=\"10.1007/s11695-024-07279-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eS\u0026aacute;nchez-Cordero S, L\u0026oacute;pez-Gonzalez R, Fern\u0026aacute;ndez H, Pujol-Gebell\u0026iacute; J. Training ChatGPT for surgical decisions: bariatric surgery analysis using algorithms and evidence. Obes Res Clin Pract. 2025;19:352\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.orcp.2025.08.002\u003c/span\u003e\u003cspan address=\"10.1016/j.orcp.2025.08.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Bariatric surgery, Artificial intelligence, Large language models, gastric sleeve, Roux-en-Y gastric bypass","lastPublishedDoi":"10.21203/rs.3.rs-8427770/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8427770/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eIntroduction\u003c/h2\u003e \u003cp\u003eArtificial intelligence (AI) has gained increasing relevance in bariatric and metabolic surgery (BMS), yet its concordance with expert clinical judgement remains uncertain. This study evaluates the agreement between two AI models and bariatric and metabolic surgeons across representative clinical scenarios, assessing the impact of directed exposure to bibliography.\u003c/p\u003e\u003ch2\u003eMaterial and methods\u003c/h2\u003e \u003cp\u003eTen evidence-based clinical scenarios were constructed using high-quality bibliography. Two AI models, ChatGPT-4 and DeepSeek, along with board-certified Mexican BMS surgeons, evaluated surgical candidacy and determined whether each patient was candidate for surgery, as well as the most appropriate procedure, under two conditions: without access to high-quality bibliography (Phase 1), and after reviewing it (Phase 2). Majority surgeon responses served as the reference standard. Agreement and concordance were analyzed descriptively and with Cohen\u0026rsquo;s kappa (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThirty surgeons completed both phases. In Phase 1, surgeons agreement varied, ranging from 96.6% to 23.3% in selected cases; when comparing to AI models, ChatGPT-4 showed 60% agreement (K\u0026thinsp;=\u0026thinsp;0.344), while DeepSeek 70% (k\u0026thinsp;=\u0026thinsp;0.508). In Phase 2, surgeons demonstrated 90% agreement across phases (k\u0026thinsp;=\u0026thinsp;0.787). ChatGPT-4 showed 60% agreement without significant concordance (k\u0026thinsp;=\u0026thinsp;0.048), whereas DeepSeek maintained 70% agreement, with fair concordance (k\u0026thinsp;=\u0026thinsp;0.318).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eSurgeons maintained high consistency between phases, while AI models showed variable alignment with clinical decision-making. DeepSeek demonstrated higher adaptability to bibliographic evidence, whereas ChatGPT-4 didn\u0026rsquo;t. These findings highlight the need for continued refinement, contextual training and validation of AI tools to ensure safe and clinically applicable support in BMS decision-making.\u003c/p\u003e","manuscriptTitle":"Exploring the role of Artificial Intelligence in Bariatric and Metabolic Surgery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-20 11:37:56","doi":"10.21203/rs.3.rs-8427770/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f16e57ef-c063-4935-b584-fb5870936dc5","owner":[],"postedDate":"January 20th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-07T22:08:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-20 11:37:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8427770","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8427770","identity":"rs-8427770","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}