Barrett esophagus and esophageal cancer: sometimes, less treatment is more

Affecting approximately 6% of persons older than 50 years, Barrett esophagus is a premalignant condition that increases the risk of developing esophageal cancer. However, the annual risk of progression to adenocarcinoma in the absence of high-grade dysplasia is low (0.12%-0.24%). Although expert consensus recommends endoscopic surveillance every 3 to 5 years in asymptomatic persons with Barrett esophagus, optimal intervals and the effectiveness of surveillance are not known.

A randomized controlled trial at 109 centers in the United Kingdom compared the outcomes of surveillance endoscopy every 2 years with “at need” endoscopy for symptoms only. In the trial, 3,453 participants with a recent diagnosis of Barrett esophagus with no or low-grade dysplasia were followed for a minimum of 10 years (mean 12.8 years). Symptoms that prompted endoscopy in the “at need” group included dysphagia, unexplained weight loss of more than 7 pounds, iron-deficiency anemia, recurrent vomiting, or worsening upper gastrointestinal symptoms. Within the participants, 93% of the surveillance group and 59% of the “at need” group received at least one endoscopy, with means of 3.5 and 1.4 endoscopies, respectively. Overall, 71 patients (2.1%) were diagnosed with esophageal cancer. There were no statistical differences in time to diagnosis of esophageal cancer, cancer stage at diagnosis, cancer-specific or overall survival.

Standard treatment for locally advanced esophageal cancer involves neoadjuvant chemoradiotherapy followed by esophagectomy. However, rates of serious postoperative complications (30%-50%), and in-hospital mortality (5%) are high. Active surveillance is a strategy to defer or avoid surgical complications in patients with a complete clinical response to chemoradiotherapy. A multicenter, cluster randomized, non-inferiority trial in 12 Dutch hospitals compared survival in 309 persons who received active surveillance vs esophagectomy within 2 weeks of chemoradiotherapy. Patients were eligible if they had no evidence of residual tumors on endoscopic biopsies, ultrasound, or PET-CT after chemoradiotherapy. After a median follow-up of 38 months, the intention-to-treat analysis found that a higher percentage of persons in the active surveillance group (75%) were alive than in the surgery group (70%). Those in the active surveillance group who underwent later surgery experienced similar postoperative complications as those who had standard surgery.

Although this study suggested that active surveillance may be a reasonable option for some with esophageal cancer, there are concerns about the durability of the findings beyond 2 years. A commentary on the study noted that the majority of patients do not have a complete response to chemoradiotherapy and would be ineligible for active surveillance. Surgeons outside of the trial also commented that the assessments for metastatic disease may not have been complete enough, resulting in many persons undergoing esophagectomy without benefit and reducing the apparent effectiveness of surgery.

**

This post first appeared on the AFP Community Blog.

Health policy that is neither big nor beautiful

During my second and third years of residency (2002-2004), I periodically volunteered to see patients at a bare-bones clinic at a homeless shelter in Lancaster. We had limited amounts of donated supplies and medications and mostly treated acute problems. One evening, a patient with diabetes came in. He had been taking oral medications but could no longer afford them and was instead giving himself shots of insulin whenever he could afford a few vials. His hemoglobin A1c level was 15, meaning that he was walking around with an average blood sugar level of around 400 and was one mild illness away from a health catastrophe. I asked him to come see me in my faculty supervised clinic at the hospital and referred him to a social worker. Miraculously (in those pre-Affordable Care Act days, Medicaid eligibility was much less generous for single "able bodied" working adults without children), it turned out that he qualified for Medicaid. By the time I graduated, his diabetes, blood pressure, and cholesterol were well-controlled and he had saved enough money to move into his own apartment.

Like every major medical and hospital association in the United States, I lobbied against the budget reconciliation bill that squeaked through the House and Senate and that President Trump signed on July 4th. I send multiple e-mails to my representatives in Congress and encouraged others to do the same. The cuts to Medicaid and the USDA's Supplemental Nutrition Assistance Program (SNAP), totaling $1 trillion over the next several years, don't even come close to filling the $4.3 trillion budget hole created by extending tax cuts that overwhelmingly benefit people like Elon Musk, Jeff Bezos, and Trump himself.

When some health policy researchers got wind of the options being considered to reduce federal Medicaid spending, they published a timely analysis in the Annals of Internal Medicine that projected the impacts of various cuts on Medicaid enrollment and the uninsured. The bill then being considered by the House was estimated to reduce the number of people with Medicaid by more than 10 million and increase the number of uninsured persons by nearly 8 million (because some people would be able to obtain another form of insurance due to increased income or becoming eligible for workplace coverage).

These estimates were horrifying enough, but the Senate version of the bill - the one that President Trump signed on America's 249th birthday - cut even deeper. 11.8 million people are expected to become uninsured, and (outside of the bill, due to Congressional inaction) an additional 5 million will lose private marketplace coverage due to no longer being able to afford to pay the premiums.

Medicaid is just the tip of the iceberg. As a fellow Pennsylvania physician observed, Medicaid cuts will hurt all American children - not just those publicly insured, since pediatric hospitals and health systems rely heavily on Medicaid rather than the relatively more generous Medicare payments that fund adult health care. States will either try to stretch reduced Medicaid funds to cover the same number of people, lower payments to doctors and hospitals, or both. Hospitals will be forced to close, leading to mass layoffs and more people with few insurance options other than Medicaid. Similarly, SNAP cuts will hurt American farmers and grocery stores in underserved areas where people will have less to spend on food. The Commonwealth Fund estimates that by 2029, the bill's impacts will include 1.2 million jobs lost nationally, depressing collective state gross domestic products by $154 billion and state and local tax revenue by $12.2 billion.

By then, a new administration will have been inaugurated that will need to clean up one big, ugly mess that the federal government created.

AI: augmenting the intelligence of family physicians

In a recent editorial, Dr. Joel Selanikio discussed how 24/7 access to generative artificial intelligence (AI) tools such as ChatGPT empowers patients to retrieve health information and self-manage low-acuity conditions that would have previously involved visiting a clinician. By embracing the capabilities of AI to reduce administrative burdens and improve clinical outcomes, Dr. Selanikio argued that practices can demonstrate “the unique and irreplaceable value doctors bring to health care.” Another opinion envisioned the rise of “AI-augmented generalists” who integrate the knowledge base of subspecialists and use large language models (LLMs) as “active cognitive collaborators.” New competencies required for the AI era include “AI system proficiency,” “collaborative problem-solving,” and “contextual adaptation.” Recently published and ongoing research provides several real-world examples.

A 2025 Graham Center Policy One-Pager synthesized information from online peer forums and vendor websites to compare costs and pros and cons of commercially available AI scribes. A study funded by the Agency for Healthcare Research and Quality is interviewing primary care clinicians and patients to identify barriers and facilitators to successful adoption of ambient digital scribe technology and to develop a prototype implementation guide for diverse primary care settings.

In addition to office notes, LLMs can be used to generate hospital discharge summaries. A study from the University of California, San Francisco, evaluated the accuracy and quality of LLM-generated discharge summaries for 100 randomly selected inpatient stays of 3 to 6 days’ duration. A team of blinded reviewers that included hospitalists, primary care physicians, and skilled nursing facility (SNF) physicians rated LLM and physician-authored summaries on comprehensiveness, concision, coherence, and errors (inaccuracies, omissions, and hallucinations). Overall, LLM narratives contained more errors but were rated as more concise and coherent than physician-generated narratives. Of note, primary care and SNF physicians—the end-users of discharge summaries—had more favorable views of LLM narratives than did hospitalists.

AI is being evaluated for its potential to assist clinical decision-making. In a single-center study of virtual urgent care visits for respiratory, urinary, vaginal, eye, or dental symptoms, AI-generated recommendations agreed with physician recommendations in 57% of cases and were more likely to be rated as optimal:

Our observations suggest that AI showed particular strength in adhering to clinical guidelines, recommending appropriate laboratory and imaging tests, and recommending necessary in-person referrals. It outperformed physicians in avoiding unjustified empirical treatments. … Conversely, physicians excelled in adapting to evolving or inconsistent patient narratives, … [and] also seemed to demonstrate better judgment in avoiding unnecessary ED referrals.

However, the AI in this study reported that it had insufficient confidence to provide a recommendation in 21% of cases.

Finally, a randomized trial examined the diagnostic accuracy of 50 US-licensed physicians who responded to clinical questions about a standardized chest pain video vignette featuring either a White male or Black female patient before and after receiving input from ChatGPT-4. This study showed that physicians were willing to modify their initial decisions based on suggestions from ChatGPT and that these changes led to improved accuracy without introducing or exacerbating demographic biases (eg, being less likely to diagnose the Black female patient with acute coronary syndrome).

**

This post first appeared on the AFP Community Blog.