DAY 3 - Evaluating the Costs and Benefits of Implementing Guideline Recommendations for Common GI Disorders

Ravy K. Vajravelu, MD, from the University of Pittsburgh School of Medicine in Pennsylvania, presented epidemiologic data on iron deficiency anemia in US adults. The 2020 AGA guidelines liberalized the WHO diagnostic threshold for iron deficiency, from a serum ferritin of 15 ng/ml to 45 ng/ml. In addition, bidirectional endoscopy is recommended for post-menopausal women and men with iron deficiency, but the AGA 2020 guidelines expanded this to include pre-menopausal women. Dr. Vajravelu examined the potential implications of these changes on patient care. He used NHANES 2017-2020 cycle data on 14 million people to examine the distribution of ferritin levels, the prevalence of iron deficient anemia, indicators of gastrointestinal pathology, and iron intake. He found that liberalizing the anemia threshold identified far higher numbers of people who would qualify as iron deficient: 6 million by the WHO threshold vs. 9 million with the AGA threshold. Approximately 29% of those newly classified as iron deficient by AGA have risk factors for GI bleeding, but 32% of them are premenopausal with no symptoms, and likely have insufficient iron intake rather than gastrointestinal pathology. Dr. Vajravelu concluded that the AGA guidelines change of the ferritin threshold identifies too many people that would not have otherwise qualified for endoscopy, which is likely not cost-effective or beneficial to patients.

Jiaxin Wang, from the Capital Medical University Affiliated Beijing Friendship Hospital in Beijing, China, described a UK Biobank study to identify lifestyle and genetic factors associated with the risk of diverticular disease, with implications for patient recommendations to reduce morbidity and healthcare costs. The prevalence of diverticular disease has increased dramatically in those aged 18-40 years in the US and Europe, with approximately 50% of cases attributed to genetic factors. Using UK Biobank data on 472,000 patients (median age 58 years, 50% male, 94% white), Dr. Wang analyzed the incidence of diverticular disease, colonic diverticular disease, and perforation and abscesses; 17 potential lifestyle factors based on the published literature; and the polygenic risk score for diverticulitis (developed by De Roo AC et al. Ann Surg. 2023). They found that coffee, sleep, and healthy diet reduced the risk of diverticulitis, whereas smoking, alcohol, tea, sedentary time, and sleeplessness increased the risk. Smoking and insomnia were risk factors for perforation and abscess. They also found that the risk increased with the number of lifestyle factors present.

Jani Mäkinen, From Tampere University in Tampere, Finland, presented a cost-effectiveness analysis of population screening for celiac disease in childhood, using a simulation model to evaluate various approaches to genetic and serologic testing. Globally, 35-42% of people are genetically at risk of celiac disease and 1-3% develop symptoms. Trying to identify these individuals based on risk and symptoms alone can cause delays in diagnosis that lead to long-term complications, increased healthcare utilization, and reduced quality of life. It is not clear whether screening is cost-effective for improving outcomes. Dr. Mäkinen used a Markov model and a representative population to compare current practice to 272 different scenarios with screening performed at different ages and based on various risk factors. The representative population was based on data from patients in Sweden and Northern Europe, and assumed a 1.8% prevalence of celiac disease and a mean age of diagnosis of 32 years. For each simulation, they looked at the cost, benefits, and harms of screening. In the main analysis and sensitivity analysis, they found that the optimal single screen was at age 11 years and the optimal repeat screen was at age 4-7 years and then 10-14 years. The screens were cost-effective if the willingness to pay was €50,000 per QALY, the prevalence is greater than 5%, and diagnosis led to improvement in quality of life and follow-up care. Dr. Mäkinen noted the lack of data regarding the harms of not screening, but the main harm would be a delay in diagnosis. 

Brian Timothy Li, MD, from Beth Israel Deaconess Medical Center in Boston, Massachusetts, used an AI-powered pancreatic cyst safety net to assess differences in pancreatic cyst management guidelines. Dr. Li described the common scenario in which a patient has a CT scan showing a pancreatic cyst and needs to be monitored. However, there are several guidelines—from AGA, ACR, ACG, Fukoka, Kyoto—that provide monitoring recommendations, and some give conflicting recommendations. He noted there are two components to decision-making in this scenario, the patient/doctor relationship and data review. There is opportunity to use AI to offload the data review task so that more time can be spent discussing options with the patient. AI could also be used to identify patients at higher risk or who have been lost to follow-up to bring them back to care. They fed an LLM-powered natural language processing algorithm with 77,787 CT and MRI reports from his institution in 2022, and asked it to search for key terms, extract pertinent cyst features and risk-related findings, and create a structured dataset. Of 2,593 candidate reports with a high likelihood of pancreatic cysts, 2,097 patients had confirmed pancreatic cysts, and the relevant features were extracted (e.g., size, number, location, solid component, wall thickening, calcification, enhancing mural nodule, main duct diameter and communication, local lymphadenopathy). The AI tool also evaluated changes over time to calculate a stable period and extracted patient clinical features (family history, new-onset diabetes, elevated CA 19-9, jaundice). Finally, the tool compared the AGA, Kyoto, and ACR guideline recommendations to determine what each recommended for a particular patient and the cost of each option. The safety net code can be updated to match the most current guidelines, saving physicians significant time. 

Oliver White, MSc, from Erasmus MC in Den Haag, The Netherlands, presented findings from a microsimulation study of harms and benefits of differing pancreatic cyst surveillance guidelines. As described by Dr. Li above, the differences between current guidelines pose a problem for patient care. Dr. White used a MISCAN-PANCREAS microsimulation model to compare three of the guidelines for pancreatic cyst surveillance (AGA, Kyoto, Fukoka), with demography, natural history and screening assumptions. He simulated random screening of 100 million individuals follow with the surveillance recommendations in each of the three guidelines compared to no surveillance. The AGA guidelines had the most effective surveillance recommendations based on number of appointments, potential surgeries, mortality, and number needed to test, followed by the Kyoto and Fukoka guidelines (which had the most intense surveillance recommendations). Ultimately, there is a low mortality benefit of 7% with surveillance, so it is important to use the most efficient strategy. Stopping surveillance of non-growing cysts or based on age would further increase surveillance efficiency.

Ravi Teja Pasam, MBBS, MPH, from Wentworth Douglass Hospital in Dover, New Hampshire, shared results from a component network meta-analysis of the best strategy to prevent pancreatitis post-endoscopic retrograde cholangiopancreatography (ERCP). The incidence of post-ERCP pancreatitis (PEP) is 10-14% and adds $8k to the cost of care. A great deal of research has been done to identify options for PEP prophylaxis (NSAIDs, IV fluids, pancreatic duct stent, and nitrates are the currently recommended approaches), but the best single or combination of prevention methods is not known. Dr. Pasam conducted a meta-analysis studies evaluating various options: per-rectal, IV, IM, and oral NSAIDs; pancreatic duct stent; aggressive hydration, standard hydration; nitrates; IV somatostatin; and subcutaneous or IV octreotide. The incidence of PEP was compared across single agents and combinations of agents. The analysis included 104 studies, in 29,404 patients, and 44 unique interventions made up of 29 individual interventions. There were 11 multi-arm RCTs and three sub-networks of authors. Pancreatic duct stent, aggressive hydration, indomethacin, and diclofenac were the most studied prophylactic approaches. Overall, adding interventions provided incremental benefits, and somatostatin 4 mcg/kg IV ranked highest among individual interventions for reduction of PEP incidence. Diclofenac 100 mg per rectal and aggressive hydration were superior to indomethacin 100 mg per rectal. Diclofenac 75 mg IV was efficacious. Somatostatin IV infusion and octreotide were beneficial, but 12-24 hours of treatment would be required to reach effective doses. Somatostatin IV bolus and isosorbide dinitrate sublingual could be considered in patients with contraindications to NSAIDs and aggressive hydration, or when administering fluids over 8 hours post-procedure is not feasible.