Faculty Bibliography

Early detection of pancreatic cancer (PC) remains challenging largely due to the low population incidence and few known risk factors. However, screening in at-risk populations and detection of early cancer has the potential to significantly alter survival. In this study, we aim to develop a predictive model to identify patients at risk for developing new-onset PC at two and a half to three year time frame. We used the Electronic Health Records (EHR) of a large medical system from 2000 to 2021 (N = 537,410). The EHR data analyzed in this work consists of patients' demographic information, diagnosis records, and lab values, which are used to identify patients who were diagnosed with pancreatic cancer and the risk factors used in the machine learning algorithm for prediction. We identified 73 risk factors of pancreatic cancer with the Phenome-wide Association Study (PheWAS) on a matched case-control cohort. Based on them, we built a large-scale machine learning algorithm based on EHR. A temporally stratified validation based on patients not included in any stage of the training of the model was performed. This model showed an AUROC at 0.742 [0.727, 0.757] which was similar in both the general population and in a subset of the population who has had prior cross-sectional imaging. The rate of diagnosis of pancreatic cancer in those in the top 1 percentile of the risk score was 6 folds higher than the general population. Our model leverages data extracted from a 6-month window of time in the electronic health record to identify patients at nearly sixfold higher than baseline risk of developing pancreatic cancer 2.5-3 years from evaluation. This approach offers an opportunity to define an enriched population entirely based on static data, where current screening may be recommended.

BACKGROUND:Measurement of colorectal polyps is typically performed via visual estimation, which is prone to bias. Studies have evaluated the accuracy of visual estimation and utility of assistive tools, but results have been mixed. This study aims to clarify the accuracy of visual estimation as a measurement tool, and the benefits of artificial intelligence. METHODS:MEDLINE and Embase were searched through October 2024. Extraction and quality assessment were performed independently by two authors. The primary outcome was the pooled absolute mean difference in size between visual estimation and control. Secondary outcomes included subgroup analysis of expert vs trainee status, accuracy of artificial intelligence, study origin (East vs. West), comparator type, definition of accuracy, polyp size, direction of estimation, and image type. RESULTS:35 studies with 42,964 polyp measurements were included in our analysis. All studies were of high quality and there was no evidence of publication bias. The pooled absolute mean difference from comparator was 1.68mm (CI 1.21-2.15) with high variability explained by differences in the comparator, the direction of estimation, image type, and size of the polyp. Overall accuracy was 60% with high variability as well, with increased accuracy with video displayed over photos. Artificial intelligence improved accuracy with an odds ratio of 7.46. CONCLUSION/CONCLUSIONS:Visual estimation is an inaccurate and imprecise way to measure colorectal polyps. Further research is needed to determine the impact on clinical outcomes related to colorectal cancer. Investment in new technology to aid in polyp measurement is an important next step.

Introduction: Colorectal cancer (CRC) is the second leading cause of cancer-related mortality in the world. While effective at preventing CRC, standard colonoscopy can miss precancerous polyps placing patients at risk for interval CRC. Endoscopic mechanical attachments and artificial intelligence (AI) are technologies that have independently shown improvement in adenoma detection rate (ADR). We sought to compare the performance of Endocuff-assisted colonoscopy (EAC) to combined AI and EAC (AEAC) in relation to ADR.
Method(s): This was a single-center study involving patients who underwent either AEAC or EAC between December 2021 and May 2022. Demographic (age, sex) and clinical (indication, Boston Bowel preparation scale (BBPS), withdrawal time, polyp location, histology and size) data on patients was obtained from the electronic health record. The primary outcome was ADR. Secondary outcomes were polyp detection rate (PDR), adenomas per colonoscopy (APC), polyps per colonoscopy (PPC), sessile serrated lesion rate (SSR) and sessile serrated lesions per colonoscopy (SSPC). Categorical variables were analyzed using a two-sided chi square test. Continuous variables were assessed using the student's t-test or Mann-Whitney U-test. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using logistic regression.
Result(s): 148 patients (50.7% men, mean age 60.9 years; 74 AEAC vs 74 EAC) were included. The AEAC group did not differ by age, sex, indication or BBPS from the EAC group (Table). ADR in the AEAC group was higher (71.6% vs 60.8%; OR 1.63; 95% CI 0.82-3.24; P = 0.17). SSR was 14.9% in the EAC group versus 24.3% in the AEAC group (P < 0.05) (Table). For adenomas .5-10mm in size, the AEAC group had a significantly higher ADR (28.4% vs 14.9%; OR 2.27; 95% CI 1.00-5.13; P = 0.05). Withdrawal time was longer in the AEAC group (8.0min vs 7.3min; P = 0.03). Subgroup analysis by indication revealed that ADR trended towards significance for patients in the AEAC group undergoing colonoscopy for CRC screening (70.3% vs 52.3%; OR 2.17; 95% CI 0.94-4.98; P = 0.068).
Conclusion(s): Combining AI with Endocuff-assisted colonoscopy increased ADR, PDR, APC, PPC, SSR and SSPC when compared to EAC. ADR trended towards significance for patients in the AEAC group undergoing CRC screening. This study highlights the potential benefits of maximizing surface area exposure (mechanical enhancement) combined with enhanced mucosal inspection (AI). Future larger studies will be needed to further validate this combination

INTRODUCTION/BACKGROUND:The impact of lowering the colon cancer screening age from 50 to 45 years on endoscopist adenoma detection rate (ADR) is not well studied. METHODS:We used average risk screening colonoscopies submitted to the GI Quality Improvement Consortium, Ltd. registry from 2014-2020 among individuals age 45 to 75 years. We used one way ANOVA test to determine differences between ADRs among 45-49, 50-54 and 50-75 year-olds. RESULTS:A total of 2,806,539 screening colonoscopies were performed by 814 endoscopists. The mean ADR in the 45-49 group was 28.6% compared to 31.8% for 50-54 group (p<0.001) and 36.3% for 50-75 (p<0.001). DISCUSSION/CONCLUSIONS:Endoscopists might see a small drop in their ADR once a higher proportion of 45-49 year old patients start undergoing screening colonoscopy.

Introduction: Early detection of pancreatic cancer (PC) remains challenging largely due to the low population incidence and few known risk factors. However, screening in at-risk populations and detection of early cancer has the potential to significantly alter survival. We used an Electronic Health Records (EHR) based large-scale machine learning algorithm to identify disease codes that are associated with the development of PC at least 3 years before diagnosis and developed a predictive model to identify patients at risk for PC 27-33 months later.
Method(s): EHR data was analyzed between 2000 and 2021 and individuals with at least 3 years of continuous presence in the database were included. A 1:4 case-control matching based on age, sex, length of medical history to all diagnosed with PC was performed. In one model, all patients meeting database presence were included, whereas in a second model only those without known prior pancreatic disease were evaluated. Among demographic and 19,304 disease variables 27-33 months prior to PC diagnosis, we used the P-value of associations to select significant variables (cut-off P-value < 0.01), and trained a logistic regression model. Final predictive performance was tested on a held-out validation cohort.
Result(s): 544,000 patients were analyzed. 2091 patients with PC were matched to 8364 cancer-free patients. We identified 73 variables with significant association of development of PC, including pancreatic cysts, diabetes, family or personal history of breast cancer, and chronic pancreatitis (ranked results and statistical analysis are shown in Table 1). These variables were selected for the regression model, which we trained in over 541,602 patients. In our second model, in patients without prior pancreatic diseases, 541,377 patients were included. The area under the receiver operating characteristic curve (AUROC) were 0.790 [0.772, 0.809] and 0.779 [0.759, 0.789] in the two models respectively.
Conclusion(s): In a robust EHR-based analysis, we identified a list of diagnostic variables associated with pancreatic cancer development in a 3-year time frame and developed a model to identify patients at risk. Although the inclusion of additional variables such as laboratory results and radiomics will likely improve the accuracy of the model, the current algorithm will allow us to develop an EHR-based identification of patients at risk for PC. (Table Presented)

Introduction: Duodenoscopes are used in more than 500,000 procedures annually in the US as a minimally invasive diagnostic and therapeutic modality for hepatobiliary and pancreatic diseases. Traditionally, these devices have been intended for re-use after undergoing strict cleaning and disinfection protocols to reduce the risk of infection between patients. However, since 2008, several major outbreaks of infections linked to duodenoscopes have resulted in devices recalled from the market. Understanding what occurred in instances of device failure leading to recalls is important to improve the safety and efficacy of these devices.
Method(s): This institution review board-exempt study reviewed the FDA Center for Devices and Radiologic Health database for all duodenoscope-related recall events from November 1, 2002 through December 31, 2020. Market entry data, recall characteristics, and adverse reports were collected for each device.
Result(s): Seventeen class II duodenoscope-related recall events were identified, affecting at least 24,611 units in distribution. 12 out of the 17 (70%) recall events were for duodenoscopes, 3 out of 17 (18%) recalls were for operation manuals, and 2 out of 17 (12%) recalls were for reprocessors. 15 out of 17 recalled devices (88%) had at least 1 documented occurrence of an adverse event at the time of recall. All recall events were approved via the 510k pathway, however postmarket-related issues accounted for 88% of recalls.
Conclusion(s): Given the wide utilization of duodenoscopes in treating pancreaticobiliary diseases, an understanding of their recall events and associated public health impact are important for endoscopists to have a greater awareness of potential safety concerns. Recalls by three duodenoscope manufacturers and one scope reprocessor manufacturer highlight the need for innovation in design and improved post-marketing surveillance mechanisms

INTRODUCTION/BACKGROUND:Adenoma detection rate (ADR) is highly variable across practices, and national or population-based estimates are not available. Our aim was to study the ADR, variability of rates over time, and factors associated with detection rates of ADR in a national sample of patients undergoing colonoscopy. METHODS:We used colonoscopies submitted to the GI Quality Improvement Consortium, Ltd. registry from 2014 to 2018 on adults aged 50-89 years. We used hierarchical logistic models to study factors associated with ADR. RESULTS:A total of 2,646,833 colonoscopies were performed by 1,169 endoscopists during the study period. The average ADR for screening colonoscopies per endoscopist was 36.80% (SD 10.21), 44.08 (SD 10.98) in men and 31.20 (SD 9.65) in women. Adjusted to the US population, the ADR was 39.08%. There was a significant increase in ADR from screening colonoscopies over the study period from 33.93% in 2014 to 38.12% in 2018. DISCUSSION/CONCLUSIONS:The average ADR from a large national US sample standardized to the US population is 39.05% and has increased over time.

BACKGROUND:The novel Coronavirus (SARS-CoV-2) has caused almost 2 million deaths worldwide. Both Food and Drug Administration and European Medicines Agency have recently approved the first COVID-19 vaccines, and a few more are going to be approved soon. METHODS:Several different approaches have been used to stimulate the immune system in mounting a humoral response. As more traditional approaches are under investigation (inactivated virus vaccines, protein subunit vaccines, recombinant virus vaccines), more recent and innovative strategies have been tried (non-replicating viral vector vaccines, RNA based vaccines, DNA based vaccines). RESULTS:Since vaccinations campaigns started in December 2020 in both the US and Europe, gastroenterologists will be one of the main sources of information regarding SARS-CoV 2 vaccination for patients in their practice, including vulnerable patients such as those with Inflammatory Bowel Disease (IBD), patients with chronic liver disease, and GI cancer patients. CONCLUSIONS:Thus, we must ourselves be well educated and updated in order to provide unambiguous counseling to these categories of vulnerable patients. In this commentary, we aim to provide a comprehensive review of both approved COVID-19 vaccines and the ones still under development, and explore potential risks, benefits and prioritization of vaccination.

BACKGROUND:Coronavirus SARS-CoV-2 affects multiple organs. Studies have reported mild elevations of lipase levels of unclear significance. Our study aims to determine the outcomes in patients with COVID-19 and hyperlipasemia, and whether correlation with D-dimer levels explains the effect on outcomes. METHODS:Case-control study from two large tertiary care health systems, of patients with COVID-19 disease admitted between March 1 and May 1, 2020 who had lipase levels recorded. Data analyzed to study primary outcomes of mortality, length of stay (LOS) and intensive care utilization in hyperlipasemia patients, and correlation with D-dimer and outcomes. RESULTS:992 out of 5597 COVID-19 patients had lipase levels, of which 429 (43%) had hyperlipasemia. 152 (15%) patients had a lipase > 3x ULN, with clinical pancreatitis in 2 patients. Hyperlipasemia had a higher mortality than normal lipase patients (32% vs. 23%, OR = 1.6,95%CI = 1.2-2.1, P = 0.002). In subgroup analysis, hyperlipasemia patients had significantly worse LOS (11vs.15 days, P = 0.01), ICU admission rates (44% vs. 66%,OR = 2.5,95%CI = 1.3-5.0,P = 0.008), ICU LOS (12vs.19 days,P = 0.01), mechanical ventilation rates (34% vs. 55%,OR = 2.4,95%CI = 1.3-4.8,P = 0.01), and durations of mechanical ventilation (14 vs. 21 days, P = 0.008). Hyperlipasemia patients were more likely to have a D-dimer value in the highest two quartiles, and had increased mortality (59% vs. 15%,OR = 7.2,95%CI = 4.5-11,P < 0.001) and LOS (10vs.7 days,P < 0.001) compared to those with normal lipase and lower D-dimer levels. CONCLUSION/CONCLUSIONS:There is high prevalence of hyperlipasemia without clinical pancreatitis in COVID-19 disease. Hyperlipasemia was associated with higher mortality and ICU utilization, possibly explained by elevated D-dimer.