Surgical Oncology

AI at the Forefront: Navigating Oncologic Care for Six Gastrointestinal Cancers According to the NCCN Guidelines Utilizing Gemini-1.0 Ultra and ChatGPT-4.0 Tamir Bresler, MD, Tyler Wilson, MD, Tadevos Makaryan, MD, Shivam Pandya, MD, MS, Kevin Palmer, MD, MPH, Ryan Meyer, MD, Zin M Htway, PhD, Manabu Fujita, MD, PhD Los Robles Regional Medical Center, Thousand Oaks, CA; General Surgical Associates, Thousand Oaks, CA Introduction: We explored the ability of public-domain large language models (LLMs) ChatGPT-4.0 (ChatGPT; OpenAI, San Francisco, CA) and Gemini 1.0 Ultra (Gemini; Google, Mountain View, CA) in guiding clinical decision-making for six gastrointestinal cancers using the National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines as a framework. Methods: We reviewed the NCCN Guidelines for Anal Squamous Cell Carcinoma, Small Bowel Adenocarcinoma, Ampullary Adenocarcinoma, Biliary Tract Cancers, Pancreatic Adenocarcinoma, and Gastric Cancer. Clinical questions were designed for each decision-making page and categorized by type. LLMs were queried with up to 3 follow-up questions. Responses were rated on a Likert scale: 5) Correct; 4) Correct following clarification; 3) Correct but incomplete; 2) Partially incorrect; 1) Absolutely incorrect. Subgroup analysis was conducted on Correctness (scores 1-2 vs. 3-5) and Accuracy (scores 1-3 vs. 4-5). Descriptive statistics were used, and significance was tested with Mann-Whitney U-test. Results: A total of 270 questions were generated (range-per-cancer 32-68). ChatGPT vs. Gemini score frequency distribution was: 5) 45.56%-vs-41.85%; 4) 19.26%-vs-15.56%; 3) 12.96%-vs-18.52%; 2) 13.70%-vs-13.33%; and 1) 8.52%-vs-10.74%. The score difference was not statistically-significant (Mean Rank [MR] 278.30-vs-262.70, p = .222). Correctness was seen in 210-vs-205 responses (77.78%-vs-75.93%), and Accuracy in 175-vs-155 (64.81%-vs-57.41%). There was no statistically-significant difference in Correctness or Accuracy between LLMs in terms of question or cancer type (Figure).Conclusion: Both LLMs demonstrated a limited capacity to assist with complex clinical decision-making in six gastrointestinal cancers. Their current Accuracy level is below the acceptable threshold for clinical use. Future studies exploring LLMs in the healthcare domain are warranted. Analyzing the Biologic Behavior and Malignant Potential of Mixed-Type vs. Branch-Duct Intraductal Papillary Mucinous Neoplasms (IPMNs) Using High-Risk Mutation Patterns in Cyst Fluid Next Generation Sequencing (NGS): A Large Multicenter Study Amudhan Kannan, MD, Tarek Sawas, MD, Nisa Kubiliun, MD, John C Mansour, MD, FACS, Amer H Zureikat, MBBCh, FACS, Herbert J Zeh III, MD, FACS, Aatur D Singhi, MD, Patricio M Polanco, MD, FACS University of Texas Southwestern Medical Center, Dallas, TX; University of Pittsburgh Medical Center, Sewickley, PA Introduction: Mixed-type IPMNs (IPMNs with MPD dilation ≥5 mm) carry a greater malignant potential compared to BD-IPMNs. To study the underlying reason for this increased propensity, we examined the mutation patterns in mixed-type and BD-IPMNs using cyst fluid NGS. Methods: Prospective NGS analysis on cyst fluid samples from 31 centers was conducted between January 2018 and February 2020. Patients with IPMNs were diagnosed based on KRAS/GNAS mutations and were categorized into two groups (based on EUS): mixed-type IPMNs (MPD ≥5 mm) with BD-IPMNs (MPD <5 mm). High-risk mutations included mutations of TP53, SMAD4, MTOR, and CTNNB1 genes. Advanced neoplasia was defined as IPMNs with HGD or PDAC. The primary outcome was the rate of high-risk mutations in the two groups. Results: Out of 674 patients (mixed-type IPMNs = 202, BD-IPMNs = 472), 106 (15.7%) had high-risk mutations. Mixed-type IPMNs had a higher rate of high-risk mutations (31% vs 9%; p < 0.0001) and co-occurrence of mutations (12% vs 3%; p < 0.0001). Within the cohort of resected patients (n = 158), the rate of advanced neoplasms was higher in mixed-type IPMNs also compared to BD-IPMNs (62% vs 42%). On multivariate logistic regression analysis, mixed-type IPMNs showed an independent association with high-risk mutations (OR = 3.42; 95% CI = 1.72-6.82; p < 0.0001) (Table). Conclusion: Mixed-type IPMNs are more likely to harbor high-risk mutations associated with advanced neoplasms. The rate of advanced neoplasms was also higher in mixed-type IPMNs. Therefore, NGS should be recommended for patients with mixed-type IPMNs regardless of the presence of additional worrisome features. Table 1. - Mutation Patterns and Rate of Advanced Neoplasia in Mixed-type IPMNs and BD-IPMNs Variables Mixed-type IPMN (n = 202) BD-IPMN (n = 472) p-value TP53 mutation (%) 47 (23.3) 26 (5.5) <0.0001* SMAD4 mutation (%) 32 (16) 23 (5) <0.0001* CTNNB1 mutation (%) 1 (0.5) 3 (0.64) 0.652 mTOR mutation (%) 14 (7) 11 (2.33) 0.004* Presence of TP53/SMAD4/CTNNB1/mTOR mutations (%) 62 (31) 44 (9.32) <0.0001* Co-occurrence of two or more high-risk mutations (%) 25 (12.4) 14 (3) <0.0001* Advanced neoplasms (%) 61 (62.2) 37 (38) 0.02* Are We There Yet? The Changing Role of Lymph Node Assessment as a Quality Indicator in Resection of Colorectal Cancer David P Cyr, MD, PhD, Omar Vergara-Fernandez, Noemie Elfassy, MD, Amanpreet Brar, James R Conner, MD, PhD, Richard Kirsch, Mantaj Brar, Erin D Kennedy, MD, Anand Govindarajan, MD, Carol J Swallow, MD, FACS University of Toronto, Toronto, ON, Canada Introduction: Adequate lymph node assessment (LNA) has been regarded as a key quality indicator in patients who undergo resection of colorectal cancer (CRC), and depends on both surgical and pathologic retrieval. Several guidelines stipulate that a minimum of 12 regional LNs should be assessed to achieve accurate nodal staging. We examined temporal trends and prognostic associations of LNA at our institution. Methods: Consecutive patients who underwent resection of a primary CRC at a high-volume center between 1997 and 2016 were identified from a comprehensive clinicopathologic database. Patients were grouped into 5-year time periods according to date of surgery. Results: The entire study cohort comprised 2137 patients (929F:1208M; median age 64 y, range 16-98). There were no observed differences across time periods for patient (sex, age) or tumor (TNM stage, resection margin) features. Adequate LNA markedly increased over time, regardless of primary tumor site or neoadjuvant therapy status (p < 0.0001, Figure 1A,B). Inadequate LNA during the 2007-2016 time-period was strongly associated with inferior DSS, but this trend was not observed during the 1996-2006 time-period (Figure 1C,D). In the recent era, inadequate LNA was rare, and was strongly associated with other adverse prognostic features, including markers of diminished systemic Adequate LNA increased over time, the of clinical guidelines in the quality of for patients with cancer centers in the era, LNA is to surgical quality and an Assessment of Oncologic a AI and a Oncologic A MD, J MD, MD, MD, MD, FACS Introduction: by has in decision-making in study to in complex compared to a from Methods: We conducted a analysis of 32 at a National of in from to February A of surgical reviewed and each We key as or the for and surgery. Comprehensive each we compared with the using the and conducted a to Results: The of between the AI ChatGPT and the was to for more studies = p = more = p = and for the of = p = The for the of = p = and the of = p = Conclusion: ChatGPT with the on and but assessment for other potential in to in clinical in to a MD, MS, PhD, MD, David PhD, of Introduction: to and has been to to and as to the increased mutations and cancers as and cancers. The of mutations in has not been Methods: We conducted in for mutations. mutations were as frequency was for Results: samples from a total of were age who were by of mutations in tumor and of had of with cancer in the by included and (Figure studies are to including in to cancer that in to should also be Future studies should the of compared with other and A Prospective of and MD, J Kirsch, MD, MD, PhD, PhD, D MD, FACS, MD, FACS University of Introduction: The of on is We that for the of cancer increased but of compared to Methods: We conducted a study of patients an cancer in the to for 3 from were and were Results: the patients were during the study and were in and 3 There were 3 for patients underwent were the for (n = had higher and at and the independent of surgical and status (Figure had higher compared to at p = and at p = status was not associated with of (OR CI p = Conclusion: is associated with increased and in patients for National Practice Patterns in the of Pancreatic MD, MD, MD, MS, MD, MD, David J MD, FACS University of Introduction: and National Comprehensive Cancer Network guidelines surgical resection for but with up should be for to with surgical resection should be the in guidelines for resection this study to the and of surgical resection for for Methods: A cohort study was using patients diagnosed with clinical from the in the National Cancer The primary outcome is the of surgical logistic regression was to of surgery. Results: the patients that underwent surgical of patients were not surgical In a logistic age and of tumor in the or of the were the of Patients and were likely to undergo compared to patients (OR 95% CI in the or of the were also more likely to be resected compared to other (OR 95% CI Conclusion: The of surgical resection compared to is to of in the of John MD, PhD, C MD, FACS, A MD, FACS, MD, PhD, FACS, MD, PhD, MD, FACS CA Introduction: The tumor in and Within the and both and tumor on study and To the of the and between we the comprehensive of the to date from over from and samples across cancer from Methods: We of for into our To and the identified we and Results: We from over from and samples across Using this we defined the of and the and by the that are in tumor samples and associated with patient we identified for for the of the we demonstrated of the between and Conclusion: We demonstrated and of and in to across cancer in and tumor in to with Cancer MD, MPH, MD, MPH, MD, C MD, David C PhD, MPH, MD, FACS General Introduction: in time to between and with is study time to cancer Methods: We the National Cancer from to and included and or diagnosed with cancer patients with and diagnosed and at The primary outcome was of as been to was the primary independent were by Results: and was by of with a higher patients compared to patients analysis no difference in the of between patients and patients 95% a of of compared to (OR 95% CI Conclusion: study time to surgical for the to patterns to and in cancer and the Patients with Cancer in MD, MD, MD, PhD, MPH, R H MD, MD, MS, M MD, C MD, MS, FACS Medical Center, Introduction: and a cancer with in patients with cancer has been and limited studies a cancer Methods: with a cancer were identified the Cancer on and were a in score and were over time and compared with a of from that was as a Results: with cancer and were analysis that compared to patients with cancer in total p = p and p as as in score p compared with a in score in the of colorectal and cancers a in an for (Figure Conclusion: to healthcare and of and score cancer and comprehensive are to patients and the associated with a cancer of Cancer Patients MD, MD, FACS, MD, FACS, D MD, FACS, C MD, MPH, FACS University of Introduction: in surgical and We to and to cancer in in Methods: were in between and groups and were to between and and and were Results: A total of were with age of was were and and were the of and to The to was for to = = and = were of was also of = and < = and = were of was not associated with or Conclusion: for and cancer were higher this patient and and for studies are to of in surgical the of on in Pancreatic MD, MD, MD, FACS, MD, FACS, J MD, FACS, MD, FACS Introduction: Pancreatic a with by The has been to and study the of on surgical the in Methods: We conducted a analysis of the cancer on patients were categorized into and (Figure resection patients with median score was The distribution across was and with an increased of (p = and (p = differences were also observed in including (p = and the for (p = differences were in and and (p but analysis across (p = (Figure Conclusion: study a across patients with an association of status with of patient healthcare as the distribution of to the in and for with High-Risk in Cyst Fluid Next Generation Sequencing in Patients with Small Intraductal Papillary Mucinous Neoplasms A Multicenter Study Amudhan Kannan, MD, Tarek Sawas, MD, Nisa Kubiliun, MD, John C Mansour, MD, FACS, Amer H Zureikat, MBBCh, FACS, Herbert J Zeh III, MD, FACS, Aatur D Singhi, MD, Patricio M Polanco, MD, FACS University of Texas Southwestern Medical Center, Dallas, TX; University of Pittsburgh Medical Center, PA Introduction: The presence of high-risk mutations in cyst fluid of IPMNs by NGS with the presence of advanced neoplasia are no guidelines for cyst fluid NGS in patients with IPMNs of We to study the associated with in IPMNs of Methods: Prospective NGS analysis on cyst fluid samples from 31 centers was conducted between January 2018 and February 2020. Patients with IPMNs were diagnosed based on the presence of KRAS/GNAS mutations. included mutations of TP53, SMAD4, MTOR, and CTNNB1 genes. Small IPMNs were defined as IPMNs of and were categorized into two with and The primary was to the associated with in IPMNs of Results: A total of patients with IPMNs = = were included in this The presence of MPD and were higher in IPMNs with resected with had a higher rate of advanced neoplasia vs. On multivariate logistic regression analysis, showed an independent association with IPMNs with (OR = 95% CI = p < 0.0001) (Table). Conclusion: Presence of is associated with high-risk mutations in IPMNs of Therefore, NGS should be in patients with IPMNs of with of with High-Risk in Small - Variables p p - - - - Presence of Presence of <0.0001* MPD dilation mm) Presence of <0.0001* <0.0001* cyst - - in Cancer and with MS, MD, J MS, PhD, David MD, FACS, MD, is a key of the colorectal cancer and has also been to as To also be associated with cancer we the of from and Methods: from (n = 11 of (n = (n = 14 was for using in and for and and were assessed using Subgroup were compared using the association between and was assessed using was in including of was in with or p = Within was higher in vs. vs p = associations were observed between and (p = or (p = is in and is at increased in and by in and showed a association with and studies are to the significance of in and with Clinical in Patients with A MD, MD, M MD, PhD, FACS, General Introduction: was the to be for of the is We assessed are that are associated with tumor Methods: A analysis of patients with was clinical and tumor of were Results: patients were for of had (Figure patients had a clinical (Figure Six had a with regression of but of had a with regression of but of had no of were more likely to (Figure and a and primary tumor (Figure was in patients (Figure were in at and but patient had at mutation The were and mutation was associated with compared to no 3) vs p = the using in at a institution. patients who from of a Pancreatic MD, PhD, MD, Manabu PhD, MD, FACS University of Introduction: The by with the comprised of and (Figure has been to and we that a more to we on was to and in was with and and tumor were Results: demonstrated were p < p < compared to (Figure Conclusion: the showed a that the of differences in that tumor and should be strongly in studies of that not are Future on the type of in and in Resection for Colorectal Cancer in a Large Regional R MD, MD, FACS, MD, FACS, PhD, MPH, J MD, FACS, PhD, MD, FACS Introduction: resection for patients with colorectal cancer with of surgical and an of healthcare patient We an healthcare cancer to colorectal cancer Methods: cohort study patients from to other the and lymph were from the of and and were using analysis, and Results: the patients had were (n = by (n = (n = and (n = p < The rate of increased patients with up (p < was associated with a in 5-year compared to no vs p < Conclusion: We in of in patients with across the underlying systemic of healthcare across We to as for in to in with in Small and Large MD, PhD, MD, PhD, MD, PhD, MD, PhD, FACS University of Introduction: is into and large and we a in and surgical in each a that the of including be between tumor with to not been We of in the status of based on of Methods: resected were were into and large and and were to the status of in each Results: The large showed higher on no difference was observed in of other The of with on was higher in large vs In of and in were higher in the the Conclusion: We differences of between of There was a of strongly in the and in to from by of therapy for each to of of into the with for with MD, C PhD, MD, PhD, MS, MS, M PhD, C MD, FACS in Introduction: to and presence of in for is a of of using with in a with an Methods: is a 1 of in with as with or is over with 3 In were conducted to of Results: in the of the and in with an increased and In of from with of a with a in Conclusion: of with systemic in patients in clinical The and systemic observed were with tumor of in was to the as observed by the the and a in of to for in Gastric Cancer PhD, PhD, PhD, J MD, FACS, MD, FACS, MD, FACS, MD, MS, FACS Medical of Introduction: been to following for the in cancer and across this study to across Methods: on patients for cancer were identified using the National Cancer of models for were in and included with significance using Results: included were were were and were across with and (p < was the on = and was on the = to in the = the of was higher = and = and = p < Conclusion: study the for models across groups to clinical was included as a during and in was that in this likely the in across groups. of Patients the of Care of in an Network MD, MD, D MD, FACS, MD, FACS CA; University of San - CA; CA Introduction: in and and associated with The of this study was to the of the of of in an on the of Methods: We examined the of patients between and and following the of in 2016 the healthcare and regression Patients and patients diagnosed with or were Results: A total of patients with were identified between and 2020. There was no difference and 2016 tumor or surgery. increased from to with a in therapy from to (p < was compared to in patients with (p < Figure 1. on multivariate regression analysis, the and age groups were independent in terms of (p < The of of in in 2016 in increased for of and with to in Cancer MD, MS, MD, M MD, John M PhD, MD, PhD, PhD, MD, FACS Comprehensive Cancer Center, University of and University of Introduction: are as a for the to neoadjuvant in cancer on the of been limited to we the of type that is associated with Methods: We of from independent with was to the type and of the were and using to the association between and Results: = 95% = = = = and = were associated with (p < In = and = were associated with (p < on that = = and = were associated with in the = were associated with in (p < was not associated with or Conclusion: of and were associated with in and in Patients with Cancer PhD, MD, FACS University of Introduction: Cancer by a in patients with study to and associated with Methods: were from patients with during was for and included and analysis and Results: was identified in including and more in increased and associated with status (p = analysis showed in (p = with = p = identified and with to and with analysis associated with and to and Conclusion: In our study on the cancer in and the at and patient to A MD, MPH, PhD, C MD, FACS, Kevin C MD, MD, FACS, R MD, FACS, MD, P MD, FACS, Cancer Center, Introduction: is the of and has demonstrated greater Methods: from were compared cohort from for of rate and were were compared between and and were to and median with Results: patients underwent were for or and were vs (p < vs (p = vs (p < and vs (p = patients had inadequate 23 patients with of in six over demonstrated and were not between (n = and (n = for vs p median vs p = vs p = vs p = and vs p = (Table). Conclusion: is the was associated with higher with of and and (%) - Variables = = p of = = p of 12 3 of in Advanced in the MD, MD, MD, FACS, M MD, FACS, MD, FACS Cancer Center, Introduction: is a of regional this had rate but the of the of study a follow-up of patients with in the Methods: cohort of patients diagnosed with and with in a between and clinical was and statistics were Results: patients of patients were at time of with median age of follow-up time was of patients of and of was 64 rate to was patients and was with median of 62 patients of on In this and was and median was Conclusion: study the of in the era, with or for patients who had were with and patients of Advanced Patients to of MD, FACS, MD, FACS, M MD, FACS Cancer Center, Introduction: We demonstrated that patients with advanced who are for surgical resection of patients resected to no of Methods: of an identified patients with who underwent resection of from were examined by and were by the Results: follow-up was age was at patients had Patients were as or at time of surgery. patients the with a median follow-up of of no therapy systemic patients (n = had a median of and of were and 5-year in who was in patients was associated with (p = and (p = Conclusion: to in the of our patients who The of the of for advanced in with of to or Cancer MD, MPH, FACS, MS, MS, PhD, MD, MD, PhD, MD, PhD, Cancer PA Introduction: is an for the of or but clinical of of are in and we that the be associated with of Methods: Using and analysis we and in a of from patients according to of HGD or on during = of HGD or cancer during at of = Subgroup were using the and logistic regression for the outcome of to HGD or Results: A total of of were and were in vs. (p = and p = as was the vs p < 0.0001). On logistic (p = and (p = were associated with was not Conclusion: and as as are associated with of of to HGD or are to the as a to and of in and of National Comprehensive Cancer Network Care for Patients with M MPH, C MD, M MD, FACS, M MD, FACS, M MD, MPH, PhD, FACS, D MD, FACS The University of Introduction: National Comprehensive Cancer Network (NCCN) guidelines lymph node for patients with high-risk and and systemic therapy for patients with We to the association between status as defined by the and to NCCN guidelines and patients with Methods: Patients diagnosed with from were identified in the and regression analysis examined the association between and clinical at of in and and of systemic therapy in included and time models were to Results: (n = in a In was associated with advanced clinical at (OR CI p < of (OR CI p < and p = for stage, and was not associated with of systemic therapy on Conclusion: Patients from had a clinical at were likely to undergo and had to and in the and of patients with are warranted. with and as for Colorectal Cancer MD, MD, MD, MD, PhD, MD, FACS University Medical of Introduction: in colorectal cancer of tumor and and and of We to therapy C as a in colorectal Methods: The study and cancer were at 37 for were with A or were with or a a Cell and Results: and in and the and compared to in and compared in p < and and in p < In increased and and and and and compared to not on p < Conclusion: The the potential of and A as for colorectal clinical of and for MD, MD, PhD, J PhD, J PhD, MD, FACS of Medical Center, CA; University of San San CA; National of Introduction: using the of cancer surgery. A is that to the from in the In this we of to an using to the of in a Methods: was to or or colorectal cancer of and were at were with the Small and of were over tumor and was using Results: Using the for of the had the at time by and Using the for of the had the Using the for of

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.