Conference Coverage

WASHINGTON, DC — A low-emulsifier-containing diet led to a threefold increased likelihood of improvement in symptoms of Crohn’s disease compared with an emulsifier-containing diet in a randomized double-blind dietary trial involving 154 patients with mildly active disease living across the United Kingdom.

The findings were reported at Gut Microbiota for Health (GMFH) World Summit 2025 by Benoit Chassaing, PhD, of the Institut Pasteur, Paris, France, whose research leading up to the trial has demonstrated that food additive emulsifiers —ubiquitous in processed foods — alter microbiota composition and lead to microbiota encroachment into the mucus layer of the gut and subsequent chronic gut inflammation.

Patients in the ADDapt trial, which was also reported in an abstract earlier this year at the European Crohn’s and Colitis Organization (ECCO) 2025 Congress, had a Crohn’s disease activity index (CDAI) of 150-250 and evidence of inflammation (faecal calprotectin (FCP) ≥ 150 µg/g or endoscopy/radiology). All “had been exposed in their regular diets to emulsifiers,” said Chassaing, a co-investigator, during a GMFH session on “Dietary Drivers of Health and Disease.”

They were randomized to either a low-emulsifier diet or to a low-emulsifier diet followed by emulsifier “resupplementation” — a design meant to “account for the very strong placebo effect that is always observed with dietary studies,” he said.

All patients received dietary counseling, a smart phone app and barcode scan to support shopping, and weekly support. They also received supermarket foods for 25% of their needs that were either free of emulsifiers or contained emulsifiers, and they were provided three snacks per day that were emulsifier-free or contained carrageenan, carboxymethycellulse (CMC), and polysorbate-80 (P80) — dietary emulsifiers that are commonly added to processed foods to enhance texture and extend shelf-life.

In the intention-to-treat (ITT) analysis, 49% of patients in the intervention group reached the primary endpoint of a 70-point reduction or more in CDAI response after 8 weeks compared with 31% of those in the control group (P = .019), with an adjusted relative risk of response of 3.1 (P = .003), Chassaing shared at the GMFH meeting, convened by the American Gastroenterological Association and the European Society of Neurogastroenterology and Motility.

In the per-protocol analysis (n = 119), 61% and 47% of patients in the intervention and control groups, respectively, reached the primary outcome of CDAI response, with an adjusted relative risk of response of 3.0 (P = .018), he said.

Secondary endpoints included CDAI remission at 24 weeks, and according to the abstract for the ECCO Congress, in the ITT analysis, patients in the intervention group were more than twice as likely to experience remission.

Chassaing noted at the GMFH meeting that as part of the study, he and coinvestigators have been investigating the participants’ gut microbiota with metagenomic analyses. The study was led by Kevin Whelan, PhD, head of the Department of Nutritional Sciences at King’s College London, London, England.

 

Can Emulsifier-Sensitive Individuals Be Identified?

In murine model research 10 years ago, Chassaing showed that the administration of CMC and P80 results in microbiota encroachment into the mucus layer of the gut, alterations in microbiota composition — including an increase in bacteria that produce pro-inflammatory flagellin — and development of chronic inflammation.

Wild-type mice treated with these compounds developed metabolic disease, and mice that were modified to be predisposed to colitis had a higher incidence of robust colitis. Moreover, fecal transplantation from emulsifier-treated mice to germ-free mice reproduced these changes, “clearly suggesting that the microbiome itself is sufficient to drive chronic inflammation,” he said.

In recent years, in humans, analyses from the large French NutriNet-Sante prospective cohort study have shown associations between exposure to food additive emulsifiers and the risk for cardiovascular disease, the risk for cancer (overall, breast, and prostate), and the risk for type 2 diabetes.

But to explore causality and better understand the mechanisms of emulsifier-driven changes on the microbiota, Chassaing and his colleagues also launched the FRESH study (Functional Research on Emulsifier in Humans), a double-blind randomized controlled-feeding study of the emulsifier CMC. For 11 days, nine healthy patients consumed an emulsifier-free diet and 11 consumed an identical diet enriched with 15 g/d of CMC.

Patients on the CMC-containing diet had reduced microbiota diversity and depletions of an array of microbiota-related metabolites, but only a small subset had profound alterations in microbiota composition and increased microbiota encroachment into the mucus layer. “Some seemed to be resistant to CMC-induced microbiota encroachment, while some were highly susceptible,” Chassaing said.

The pilot study raised the question, he said, of whether there is an “infectivity component” — some kind of “sensitive” gut microbiota composition — that may be associated with dietary emulsifier-driven inflammation and disease.

In other murine research, Chassaing and his team found that germ-free mice colonized with Crohn’s disease-associated adherent-invasive E coli (AIEC) and subsequently given CMC or P80 developed chronic inflammation and metabolic dysregulation, “clearly demonstrating that you can convert resistant mice to sensitive mice just by adding one bacteria to the ecosystem,” he said. “The presence of AIEC alone was sufficient to drive the detrimental effects of dietary emulsifiers.”

(In vitro research with transcriptomic analysis then showed that the emulsifiers directly elicit AIEC virulence gene expression, Chassaing and his coauthors wrote in their 2020 paper, facilitating AIEC’s “penetration of the mucus layer and adherence to epithelial cells and resulting in activation of host pro-inflammatory signaling.”)

“We don’t think it’s solely the AIEC bacteria that will drive emulsifier sensitivity, though…we think it’s more complex,” Chassaing said at the meeting. Overall, the findings raise the question of whether emulsifier-sensitive individuals can be identified.

This, he said, is one of his most recent research questions. His lab has led the development of an in vitro microbiota model built to predict an individual’s sensitivity to emulsifiers. In a study published in April, the model recapitulated the differential CMC sensitivity observed in the earlier FRESH study, suggesting that an individual’s sensitivity to emulsifiers can indeed be predicted by examining their baseline microbiota.

 

Interpreting the Epidemiology

Chassaing’s research arch illustrates the synergy between epidemiological research, basic/translational research, and clinical interventional research that’s needed to understand the diet-microbiome intersection in inflammatory bowel disease, said Ashwin Ananthakrishnan, MBBS, MPH, AGAF, associate professor of medicine at Massachusetts General Hospital, Boston, in an interview at the meeting.

“It’s a good example of how to really span the spectrum, starting from the big picture and going deeper to understand mechanisms, and starting from mechanisms and expanding it out,” Ananthakrishnan said.

In his own talk about research on IBD, Ananthakrishnan said that epidemiological data have shown over the past 10-15 years that total dietary fiber is inversely associated with the risk for Crohn’s disease (with the strongest associations with fiber from fruits and vegetables). Studies have also shown that a higher intake of polyunsaturated fatty acids is associated with a lower risk for ulcerative colitis, whereas “an n-6-fatty acid-rich diet is associated with a higher risk of ulcerative colitis,” he said.

Dietary cohort studies, meanwhile, have shed light on the influence of dietary patterns — such as the Mediterranean diet and diets with high inflammatory potential—on IBD. A diet rich in ultra-processed foods has also been shown in a prospective cohort study to be associated with a higher risk for Crohn’s disease, with certain categories of ultra-processed foods (eg, breads and breakfast foods) having the strongest associations.

Such studies are limited in part, however, by inadequate assessment of potentially relevant variables such as emulsifiers, preservatives, and how the food is processed, he said.

And in interpreting the epidemiological research on fiber and IBD, for instance, one must appreciate that “there are a number of mechanisms by which fiber is impactful…there’s a big picture to look at,” Ananthakrishnan said. Fiber “can affect the microbiome, clearly, it can affect the gut barrier, and it can affect bile acids, and there are detailed translational studies in support of each of these.”

But there are other constituents of fruits and vegetables “that could potentially influence disease risk, such as AhR ligands and polyphenols,” he said. “And importantly, people not eating a lot of fiber may be eating a lot of ultra-processed foods.”

Most interventional studies of fiber have not shown a benefit of a high-fiber diet, Ananthakrishnan said, but there are multiple possible reasons and factors at play, including potential population differences (eg, in inflammatory status or baseline microbiota), shortcomings of the interventions, and potentially inaccurate outcomes.

Abigail Johnson, PhD, RDN, associate director of the Nutrition Coordinating Center, University of Minnesota Twin Cities, which supports dietary analysis, said during the session that the focus of dietary research is “moving toward understanding overall dietary patterns” as opposed to focusing more narrowly on vitamins, minerals, and macronutrients such as proteins, fats, and carbohydrates.

This is an improvement, though “we still don’t have good approaches for understanding [the contributions of] things like additives and emulsifiers, food preparation and cooking, and food processing,” said Johnson, assistant professor in the Division of Epidemiology and Community Health at University of Minnesota Twin Cities. “Perhaps by looking at things at the food level we can overcome some of these limitations.”

Ananthakrishnan reported being a consultant for Geneoscopy and receiving a research grant from Takeda. Chassaing did not report any financial disclosures. Johnson reported that she had no financial disclosures.

A version of this article appeared on Medscape.com.

WASHINGTON, DC — A low-emulsifier-containing diet led to a threefold increased likelihood of improvement in symptoms of Crohn’s disease compared with an emulsifier-containing diet in a randomized double-blind dietary trial involving 154 patients with mildly active disease living across the United Kingdom.

The findings were reported at Gut Microbiota for Health (GMFH) World Summit 2025 by Benoit Chassaing, PhD, of the Institut Pasteur, Paris, France, whose research leading up to the trial has demonstrated that food additive emulsifiers —ubiquitous in processed foods — alter microbiota composition and lead to microbiota encroachment into the mucus layer of the gut and subsequent chronic gut inflammation.

Patients in the ADDapt trial, which was also reported in an abstract earlier this year at the European Crohn’s and Colitis Organization (ECCO) 2025 Congress, had a Crohn’s disease activity index (CDAI) of 150-250 and evidence of inflammation (faecal calprotectin (FCP) ≥ 150 µg/g or endoscopy/radiology). All “had been exposed in their regular diets to emulsifiers,” said Chassaing, a co-investigator, during a GMFH session on “Dietary Drivers of Health and Disease.”

They were randomized to either a low-emulsifier diet or to a low-emulsifier diet followed by emulsifier “resupplementation” — a design meant to “account for the very strong placebo effect that is always observed with dietary studies,” he said.

All patients received dietary counseling, a smart phone app and barcode scan to support shopping, and weekly support. They also received supermarket foods for 25% of their needs that were either free of emulsifiers or contained emulsifiers, and they were provided three snacks per day that were emulsifier-free or contained carrageenan, carboxymethycellulse (CMC), and polysorbate-80 (P80) — dietary emulsifiers that are commonly added to processed foods to enhance texture and extend shelf-life.

In the intention-to-treat (ITT) analysis, 49% of patients in the intervention group reached the primary endpoint of a 70-point reduction or more in CDAI response after 8 weeks compared with 31% of those in the control group (P = .019), with an adjusted relative risk of response of 3.1 (P = .003), Chassaing shared at the GMFH meeting, convened by the American Gastroenterological Association and the European Society of Neurogastroenterology and Motility.

In the per-protocol analysis (n = 119), 61% and 47% of patients in the intervention and control groups, respectively, reached the primary outcome of CDAI response, with an adjusted relative risk of response of 3.0 (P = .018), he said.

Secondary endpoints included CDAI remission at 24 weeks, and according to the abstract for the ECCO Congress, in the ITT analysis, patients in the intervention group were more than twice as likely to experience remission.

Chassaing noted at the GMFH meeting that as part of the study, he and coinvestigators have been investigating the participants’ gut microbiota with metagenomic analyses. The study was led by Kevin Whelan, PhD, head of the Department of Nutritional Sciences at King’s College London, London, England.

 

Can Emulsifier-Sensitive Individuals Be Identified?

In murine model research 10 years ago, Chassaing showed that the administration of CMC and P80 results in microbiota encroachment into the mucus layer of the gut, alterations in microbiota composition — including an increase in bacteria that produce pro-inflammatory flagellin — and development of chronic inflammation.

Wild-type mice treated with these compounds developed metabolic disease, and mice that were modified to be predisposed to colitis had a higher incidence of robust colitis. Moreover, fecal transplantation from emulsifier-treated mice to germ-free mice reproduced these changes, “clearly suggesting that the microbiome itself is sufficient to drive chronic inflammation,” he said.

In recent years, in humans, analyses from the large French NutriNet-Sante prospective cohort study have shown associations between exposure to food additive emulsifiers and the risk for cardiovascular disease, the risk for cancer (overall, breast, and prostate), and the risk for type 2 diabetes.

But to explore causality and better understand the mechanisms of emulsifier-driven changes on the microbiota, Chassaing and his colleagues also launched the FRESH study (Functional Research on Emulsifier in Humans), a double-blind randomized controlled-feeding study of the emulsifier CMC. For 11 days, nine healthy patients consumed an emulsifier-free diet and 11 consumed an identical diet enriched with 15 g/d of CMC.

Patients on the CMC-containing diet had reduced microbiota diversity and depletions of an array of microbiota-related metabolites, but only a small subset had profound alterations in microbiota composition and increased microbiota encroachment into the mucus layer. “Some seemed to be resistant to CMC-induced microbiota encroachment, while some were highly susceptible,” Chassaing said.

The pilot study raised the question, he said, of whether there is an “infectivity component” — some kind of “sensitive” gut microbiota composition — that may be associated with dietary emulsifier-driven inflammation and disease.

In other murine research, Chassaing and his team found that germ-free mice colonized with Crohn’s disease-associated adherent-invasive E coli (AIEC) and subsequently given CMC or P80 developed chronic inflammation and metabolic dysregulation, “clearly demonstrating that you can convert resistant mice to sensitive mice just by adding one bacteria to the ecosystem,” he said. “The presence of AIEC alone was sufficient to drive the detrimental effects of dietary emulsifiers.”

(In vitro research with transcriptomic analysis then showed that the emulsifiers directly elicit AIEC virulence gene expression, Chassaing and his coauthors wrote in their 2020 paper, facilitating AIEC’s “penetration of the mucus layer and adherence to epithelial cells and resulting in activation of host pro-inflammatory signaling.”)

“We don’t think it’s solely the AIEC bacteria that will drive emulsifier sensitivity, though…we think it’s more complex,” Chassaing said at the meeting. Overall, the findings raise the question of whether emulsifier-sensitive individuals can be identified.

This, he said, is one of his most recent research questions. His lab has led the development of an in vitro microbiota model built to predict an individual’s sensitivity to emulsifiers. In a study published in April, the model recapitulated the differential CMC sensitivity observed in the earlier FRESH study, suggesting that an individual’s sensitivity to emulsifiers can indeed be predicted by examining their baseline microbiota.

 

Interpreting the Epidemiology

Chassaing’s research arch illustrates the synergy between epidemiological research, basic/translational research, and clinical interventional research that’s needed to understand the diet-microbiome intersection in inflammatory bowel disease, said Ashwin Ananthakrishnan, MBBS, MPH, AGAF, associate professor of medicine at Massachusetts General Hospital, Boston, in an interview at the meeting.

“It’s a good example of how to really span the spectrum, starting from the big picture and going deeper to understand mechanisms, and starting from mechanisms and expanding it out,” Ananthakrishnan said.

In his own talk about research on IBD, Ananthakrishnan said that epidemiological data have shown over the past 10-15 years that total dietary fiber is inversely associated with the risk for Crohn’s disease (with the strongest associations with fiber from fruits and vegetables). Studies have also shown that a higher intake of polyunsaturated fatty acids is associated with a lower risk for ulcerative colitis, whereas “an n-6-fatty acid-rich diet is associated with a higher risk of ulcerative colitis,” he said.

Dietary cohort studies, meanwhile, have shed light on the influence of dietary patterns — such as the Mediterranean diet and diets with high inflammatory potential—on IBD. A diet rich in ultra-processed foods has also been shown in a prospective cohort study to be associated with a higher risk for Crohn’s disease, with certain categories of ultra-processed foods (eg, breads and breakfast foods) having the strongest associations.

Such studies are limited in part, however, by inadequate assessment of potentially relevant variables such as emulsifiers, preservatives, and how the food is processed, he said.

And in interpreting the epidemiological research on fiber and IBD, for instance, one must appreciate that “there are a number of mechanisms by which fiber is impactful…there’s a big picture to look at,” Ananthakrishnan said. Fiber “can affect the microbiome, clearly, it can affect the gut barrier, and it can affect bile acids, and there are detailed translational studies in support of each of these.”

But there are other constituents of fruits and vegetables “that could potentially influence disease risk, such as AhR ligands and polyphenols,” he said. “And importantly, people not eating a lot of fiber may be eating a lot of ultra-processed foods.”

Most interventional studies of fiber have not shown a benefit of a high-fiber diet, Ananthakrishnan said, but there are multiple possible reasons and factors at play, including potential population differences (eg, in inflammatory status or baseline microbiota), shortcomings of the interventions, and potentially inaccurate outcomes.

Abigail Johnson, PhD, RDN, associate director of the Nutrition Coordinating Center, University of Minnesota Twin Cities, which supports dietary analysis, said during the session that the focus of dietary research is “moving toward understanding overall dietary patterns” as opposed to focusing more narrowly on vitamins, minerals, and macronutrients such as proteins, fats, and carbohydrates.

This is an improvement, though “we still don’t have good approaches for understanding [the contributions of] things like additives and emulsifiers, food preparation and cooking, and food processing,” said Johnson, assistant professor in the Division of Epidemiology and Community Health at University of Minnesota Twin Cities. “Perhaps by looking at things at the food level we can overcome some of these limitations.”

Ananthakrishnan reported being a consultant for Geneoscopy and receiving a research grant from Takeda. Chassaing did not report any financial disclosures. Johnson reported that she had no financial disclosures.

A version of this article appeared on Medscape.com.

WASHINGTON, DC — A low-emulsifier-containing diet led to a threefold increased likelihood of improvement in symptoms of Crohn’s disease compared with an emulsifier-containing diet in a randomized double-blind dietary trial involving 154 patients with mildly active disease living across the United Kingdom.

The findings were reported at Gut Microbiota for Health (GMFH) World Summit 2025 by Benoit Chassaing, PhD, of the Institut Pasteur, Paris, France, whose research leading up to the trial has demonstrated that food additive emulsifiers —ubiquitous in processed foods — alter microbiota composition and lead to microbiota encroachment into the mucus layer of the gut and subsequent chronic gut inflammation.

Patients in the ADDapt trial, which was also reported in an abstract earlier this year at the European Crohn’s and Colitis Organization (ECCO) 2025 Congress, had a Crohn’s disease activity index (CDAI) of 150-250 and evidence of inflammation (faecal calprotectin (FCP) ≥ 150 µg/g or endoscopy/radiology). All “had been exposed in their regular diets to emulsifiers,” said Chassaing, a co-investigator, during a GMFH session on “Dietary Drivers of Health and Disease.”

They were randomized to either a low-emulsifier diet or to a low-emulsifier diet followed by emulsifier “resupplementation” — a design meant to “account for the very strong placebo effect that is always observed with dietary studies,” he said.

All patients received dietary counseling, a smart phone app and barcode scan to support shopping, and weekly support. They also received supermarket foods for 25% of their needs that were either free of emulsifiers or contained emulsifiers, and they were provided three snacks per day that were emulsifier-free or contained carrageenan, carboxymethycellulse (CMC), and polysorbate-80 (P80) — dietary emulsifiers that are commonly added to processed foods to enhance texture and extend shelf-life.

In the intention-to-treat (ITT) analysis, 49% of patients in the intervention group reached the primary endpoint of a 70-point reduction or more in CDAI response after 8 weeks compared with 31% of those in the control group (P = .019), with an adjusted relative risk of response of 3.1 (P = .003), Chassaing shared at the GMFH meeting, convened by the American Gastroenterological Association and the European Society of Neurogastroenterology and Motility.

In the per-protocol analysis (n = 119), 61% and 47% of patients in the intervention and control groups, respectively, reached the primary outcome of CDAI response, with an adjusted relative risk of response of 3.0 (P = .018), he said.

Secondary endpoints included CDAI remission at 24 weeks, and according to the abstract for the ECCO Congress, in the ITT analysis, patients in the intervention group were more than twice as likely to experience remission.

Chassaing noted at the GMFH meeting that as part of the study, he and coinvestigators have been investigating the participants’ gut microbiota with metagenomic analyses. The study was led by Kevin Whelan, PhD, head of the Department of Nutritional Sciences at King’s College London, London, England.

 

Can Emulsifier-Sensitive Individuals Be Identified?

In murine model research 10 years ago, Chassaing showed that the administration of CMC and P80 results in microbiota encroachment into the mucus layer of the gut, alterations in microbiota composition — including an increase in bacteria that produce pro-inflammatory flagellin — and development of chronic inflammation.

Wild-type mice treated with these compounds developed metabolic disease, and mice that were modified to be predisposed to colitis had a higher incidence of robust colitis. Moreover, fecal transplantation from emulsifier-treated mice to germ-free mice reproduced these changes, “clearly suggesting that the microbiome itself is sufficient to drive chronic inflammation,” he said.

In recent years, in humans, analyses from the large French NutriNet-Sante prospective cohort study have shown associations between exposure to food additive emulsifiers and the risk for cardiovascular disease, the risk for cancer (overall, breast, and prostate), and the risk for type 2 diabetes.

But to explore causality and better understand the mechanisms of emulsifier-driven changes on the microbiota, Chassaing and his colleagues also launched the FRESH study (Functional Research on Emulsifier in Humans), a double-blind randomized controlled-feeding study of the emulsifier CMC. For 11 days, nine healthy patients consumed an emulsifier-free diet and 11 consumed an identical diet enriched with 15 g/d of CMC.

Patients on the CMC-containing diet had reduced microbiota diversity and depletions of an array of microbiota-related metabolites, but only a small subset had profound alterations in microbiota composition and increased microbiota encroachment into the mucus layer. “Some seemed to be resistant to CMC-induced microbiota encroachment, while some were highly susceptible,” Chassaing said.

The pilot study raised the question, he said, of whether there is an “infectivity component” — some kind of “sensitive” gut microbiota composition — that may be associated with dietary emulsifier-driven inflammation and disease.

In other murine research, Chassaing and his team found that germ-free mice colonized with Crohn’s disease-associated adherent-invasive E coli (AIEC) and subsequently given CMC or P80 developed chronic inflammation and metabolic dysregulation, “clearly demonstrating that you can convert resistant mice to sensitive mice just by adding one bacteria to the ecosystem,” he said. “The presence of AIEC alone was sufficient to drive the detrimental effects of dietary emulsifiers.”

(In vitro research with transcriptomic analysis then showed that the emulsifiers directly elicit AIEC virulence gene expression, Chassaing and his coauthors wrote in their 2020 paper, facilitating AIEC’s “penetration of the mucus layer and adherence to epithelial cells and resulting in activation of host pro-inflammatory signaling.”)

“We don’t think it’s solely the AIEC bacteria that will drive emulsifier sensitivity, though…we think it’s more complex,” Chassaing said at the meeting. Overall, the findings raise the question of whether emulsifier-sensitive individuals can be identified.

This, he said, is one of his most recent research questions. His lab has led the development of an in vitro microbiota model built to predict an individual’s sensitivity to emulsifiers. In a study published in April, the model recapitulated the differential CMC sensitivity observed in the earlier FRESH study, suggesting that an individual’s sensitivity to emulsifiers can indeed be predicted by examining their baseline microbiota.

 

Interpreting the Epidemiology

Chassaing’s research arch illustrates the synergy between epidemiological research, basic/translational research, and clinical interventional research that’s needed to understand the diet-microbiome intersection in inflammatory bowel disease, said Ashwin Ananthakrishnan, MBBS, MPH, AGAF, associate professor of medicine at Massachusetts General Hospital, Boston, in an interview at the meeting.

“It’s a good example of how to really span the spectrum, starting from the big picture and going deeper to understand mechanisms, and starting from mechanisms and expanding it out,” Ananthakrishnan said.

In his own talk about research on IBD, Ananthakrishnan said that epidemiological data have shown over the past 10-15 years that total dietary fiber is inversely associated with the risk for Crohn’s disease (with the strongest associations with fiber from fruits and vegetables). Studies have also shown that a higher intake of polyunsaturated fatty acids is associated with a lower risk for ulcerative colitis, whereas “an n-6-fatty acid-rich diet is associated with a higher risk of ulcerative colitis,” he said.

Dietary cohort studies, meanwhile, have shed light on the influence of dietary patterns — such as the Mediterranean diet and diets with high inflammatory potential—on IBD. A diet rich in ultra-processed foods has also been shown in a prospective cohort study to be associated with a higher risk for Crohn’s disease, with certain categories of ultra-processed foods (eg, breads and breakfast foods) having the strongest associations.

Such studies are limited in part, however, by inadequate assessment of potentially relevant variables such as emulsifiers, preservatives, and how the food is processed, he said.

And in interpreting the epidemiological research on fiber and IBD, for instance, one must appreciate that “there are a number of mechanisms by which fiber is impactful…there’s a big picture to look at,” Ananthakrishnan said. Fiber “can affect the microbiome, clearly, it can affect the gut barrier, and it can affect bile acids, and there are detailed translational studies in support of each of these.”

But there are other constituents of fruits and vegetables “that could potentially influence disease risk, such as AhR ligands and polyphenols,” he said. “And importantly, people not eating a lot of fiber may be eating a lot of ultra-processed foods.”

Most interventional studies of fiber have not shown a benefit of a high-fiber diet, Ananthakrishnan said, but there are multiple possible reasons and factors at play, including potential population differences (eg, in inflammatory status or baseline microbiota), shortcomings of the interventions, and potentially inaccurate outcomes.

Abigail Johnson, PhD, RDN, associate director of the Nutrition Coordinating Center, University of Minnesota Twin Cities, which supports dietary analysis, said during the session that the focus of dietary research is “moving toward understanding overall dietary patterns” as opposed to focusing more narrowly on vitamins, minerals, and macronutrients such as proteins, fats, and carbohydrates.

This is an improvement, though “we still don’t have good approaches for understanding [the contributions of] things like additives and emulsifiers, food preparation and cooking, and food processing,” said Johnson, assistant professor in the Division of Epidemiology and Community Health at University of Minnesota Twin Cities. “Perhaps by looking at things at the food level we can overcome some of these limitations.”

Ananthakrishnan reported being a consultant for Geneoscopy and receiving a research grant from Takeda. Chassaing did not report any financial disclosures. Johnson reported that she had no financial disclosures.

A version of this article appeared on Medscape.com.

The role of hereditary syndromes in renal cell carcinoma (RCC) might be easily missed, a kidney cancer specialist said during a recent Association of VA Hematology and Oncology (AVAHO) seminar in Long Beach, California, though careful clinical evaluation can uncover genetic traits that may affect treatment and familial risk.

“The importance of finding or identifying hereditary forms of kidney cancer really should not be underestimated,” said urologist Brian Shuch, MD, director of the UCLA Kidney Cancer Program, on treating veterans with kidney cancer. 

According to Shuch, recent data suggest that about 4.5% of patients with RCC have a hereditary syndrome: “A lot of times, these hide in plain sight. You have to really look deep and try to figure things out and understand that maybe they have a hereditary form of kidney cancer.”

It is important to consider early genetic testing, Shuch said. Red flags for hereditary syndromes include early-onset RCC (age ≤ 45 years), multifocal tumors, bilateral tumors (especially in younger individuals), or a relevant family personal history, he said. 

Unusual skin conditions are also potential signs, Shuch said. These can include leiomyomas, fibrofolliculomas, and angiofibromas: “Patients have lots of lumps or bumps.”

“When I look at a patient, I go head to toe and ask if there any issues with your vision, any issues with your hearing, any issues swallowing,” he explained at the meeting. “Do you have any problems with heart issues, adrenal issues? You’ve got to go through each organ, and it can lead you to different things.”

Shuch highlighted Von Hippel-Lindau (VHL) syndrome, which affects 1 in 25,000 people. About 80% to 90% of these patients have a family history, Shuch said.

But the others do not. “Unfortunately, some get diagnosed later in life because they don’t get cascade testing starting at aged 2, which is recommended. These are the patients who might be coming into the ER with a hemangioblastoma or picking up the phone and all of a sudden being deaf in one ear due to an endolymphatic sac tumor.

“We want to limit metastatic spread and preserve the kidneys,” Shuch said. “We don’t want to be doing radical nephrectomies. We want to avoid chronic kidney disease, prevent end-stage renal disease, and maximize quality of life.”

It’s a good idea to avoid surgical removal unless a patient’s tumor grows to be > 3 cm, a line that indicates risk of metastases, he said. 

In terms of treatment, Shuch highlighted a 2021 study that showed benefit in VHL from belzutifan (Welireg), an oral HIF-2 α inhibitor approved by the US Food and Drug Administration. The medication significantly reduced the need for surgical intervention. 

“Patients go on this drug, and surgeons are putting their scalpels down,” said Shuch, who worked on the 2021 study. 

Other hereditary syndromes include the rare hereditary papillary RCC, and Birt-Hogg-Dubé syndrome, believed to affect 1 in 200,000 people but may be more common, he said. 

Birt-Hogg-Dubé syndrome is linked to lung cysts, lung collapse, and skin manifestations. The 3 cm surgery rule is appropriate in these cases, Shuch said, and metastases are rare.

Another condition, hereditary leiomyomatosis and RCC, is the most dangerous hereditary form. Originally thought to affect 1 in 200,000 people, hereditary leiomyomatosis and RCC is similar to Birt-Hogg-Dubé syndrome in that it is believed to be more common.

“You will see this,” Shuch predicted. 

Shuch advised colleagues to intervene early and take a large margin during surgery.

He also highlighted familial paraganglioma syndrome, which is associated with gastrointestinal stromal tumors, and Cowden syndrome, which is linked to skin manifestations and breast, thyroid, and endometrial cancer. 

Shuch reported that he had no disclosures.

The role of hereditary syndromes in renal cell carcinoma (RCC) might be easily missed, a kidney cancer specialist said during a recent Association of VA Hematology and Oncology (AVAHO) seminar in Long Beach, California, though careful clinical evaluation can uncover genetic traits that may affect treatment and familial risk.

“The importance of finding or identifying hereditary forms of kidney cancer really should not be underestimated,” said urologist Brian Shuch, MD, director of the UCLA Kidney Cancer Program, on treating veterans with kidney cancer. 

According to Shuch, recent data suggest that about 4.5% of patients with RCC have a hereditary syndrome: “A lot of times, these hide in plain sight. You have to really look deep and try to figure things out and understand that maybe they have a hereditary form of kidney cancer.”

It is important to consider early genetic testing, Shuch said. Red flags for hereditary syndromes include early-onset RCC (age ≤ 45 years), multifocal tumors, bilateral tumors (especially in younger individuals), or a relevant family personal history, he said. 

Unusual skin conditions are also potential signs, Shuch said. These can include leiomyomas, fibrofolliculomas, and angiofibromas: “Patients have lots of lumps or bumps.”

“When I look at a patient, I go head to toe and ask if there any issues with your vision, any issues with your hearing, any issues swallowing,” he explained at the meeting. “Do you have any problems with heart issues, adrenal issues? You’ve got to go through each organ, and it can lead you to different things.”

Shuch highlighted Von Hippel-Lindau (VHL) syndrome, which affects 1 in 25,000 people. About 80% to 90% of these patients have a family history, Shuch said.

But the others do not. “Unfortunately, some get diagnosed later in life because they don’t get cascade testing starting at aged 2, which is recommended. These are the patients who might be coming into the ER with a hemangioblastoma or picking up the phone and all of a sudden being deaf in one ear due to an endolymphatic sac tumor.

“We want to limit metastatic spread and preserve the kidneys,” Shuch said. “We don’t want to be doing radical nephrectomies. We want to avoid chronic kidney disease, prevent end-stage renal disease, and maximize quality of life.”

It’s a good idea to avoid surgical removal unless a patient’s tumor grows to be > 3 cm, a line that indicates risk of metastases, he said. 

In terms of treatment, Shuch highlighted a 2021 study that showed benefit in VHL from belzutifan (Welireg), an oral HIF-2 α inhibitor approved by the US Food and Drug Administration. The medication significantly reduced the need for surgical intervention. 

“Patients go on this drug, and surgeons are putting their scalpels down,” said Shuch, who worked on the 2021 study. 

Other hereditary syndromes include the rare hereditary papillary RCC, and Birt-Hogg-Dubé syndrome, believed to affect 1 in 200,000 people but may be more common, he said. 

Birt-Hogg-Dubé syndrome is linked to lung cysts, lung collapse, and skin manifestations. The 3 cm surgery rule is appropriate in these cases, Shuch said, and metastases are rare.

Another condition, hereditary leiomyomatosis and RCC, is the most dangerous hereditary form. Originally thought to affect 1 in 200,000 people, hereditary leiomyomatosis and RCC is similar to Birt-Hogg-Dubé syndrome in that it is believed to be more common.

“You will see this,” Shuch predicted. 

Shuch advised colleagues to intervene early and take a large margin during surgery.

He also highlighted familial paraganglioma syndrome, which is associated with gastrointestinal stromal tumors, and Cowden syndrome, which is linked to skin manifestations and breast, thyroid, and endometrial cancer. 

Shuch reported that he had no disclosures.

The role of hereditary syndromes in renal cell carcinoma (RCC) might be easily missed, a kidney cancer specialist said during a recent Association of VA Hematology and Oncology (AVAHO) seminar in Long Beach, California, though careful clinical evaluation can uncover genetic traits that may affect treatment and familial risk.

“The importance of finding or identifying hereditary forms of kidney cancer really should not be underestimated,” said urologist Brian Shuch, MD, director of the UCLA Kidney Cancer Program, on treating veterans with kidney cancer. 

According to Shuch, recent data suggest that about 4.5% of patients with RCC have a hereditary syndrome: “A lot of times, these hide in plain sight. You have to really look deep and try to figure things out and understand that maybe they have a hereditary form of kidney cancer.”

It is important to consider early genetic testing, Shuch said. Red flags for hereditary syndromes include early-onset RCC (age ≤ 45 years), multifocal tumors, bilateral tumors (especially in younger individuals), or a relevant family personal history, he said. 

Unusual skin conditions are also potential signs, Shuch said. These can include leiomyomas, fibrofolliculomas, and angiofibromas: “Patients have lots of lumps or bumps.”

“When I look at a patient, I go head to toe and ask if there any issues with your vision, any issues with your hearing, any issues swallowing,” he explained at the meeting. “Do you have any problems with heart issues, adrenal issues? You’ve got to go through each organ, and it can lead you to different things.”

Shuch highlighted Von Hippel-Lindau (VHL) syndrome, which affects 1 in 25,000 people. About 80% to 90% of these patients have a family history, Shuch said.

But the others do not. “Unfortunately, some get diagnosed later in life because they don’t get cascade testing starting at aged 2, which is recommended. These are the patients who might be coming into the ER with a hemangioblastoma or picking up the phone and all of a sudden being deaf in one ear due to an endolymphatic sac tumor.

“We want to limit metastatic spread and preserve the kidneys,” Shuch said. “We don’t want to be doing radical nephrectomies. We want to avoid chronic kidney disease, prevent end-stage renal disease, and maximize quality of life.”

It’s a good idea to avoid surgical removal unless a patient’s tumor grows to be > 3 cm, a line that indicates risk of metastases, he said. 

In terms of treatment, Shuch highlighted a 2021 study that showed benefit in VHL from belzutifan (Welireg), an oral HIF-2 α inhibitor approved by the US Food and Drug Administration. The medication significantly reduced the need for surgical intervention. 

“Patients go on this drug, and surgeons are putting their scalpels down,” said Shuch, who worked on the 2021 study. 

Other hereditary syndromes include the rare hereditary papillary RCC, and Birt-Hogg-Dubé syndrome, believed to affect 1 in 200,000 people but may be more common, he said. 

Birt-Hogg-Dubé syndrome is linked to lung cysts, lung collapse, and skin manifestations. The 3 cm surgery rule is appropriate in these cases, Shuch said, and metastases are rare.

Another condition, hereditary leiomyomatosis and RCC, is the most dangerous hereditary form. Originally thought to affect 1 in 200,000 people, hereditary leiomyomatosis and RCC is similar to Birt-Hogg-Dubé syndrome in that it is believed to be more common.

“You will see this,” Shuch predicted. 

Shuch advised colleagues to intervene early and take a large margin during surgery.

He also highlighted familial paraganglioma syndrome, which is associated with gastrointestinal stromal tumors, and Cowden syndrome, which is linked to skin manifestations and breast, thyroid, and endometrial cancer. 

Shuch reported that he had no disclosures.

WASHINGTON — For years, oncologist Jonathan Peled, MD, PhD, and his colleagues at Memorial Sloan Kettering Cancer Center (MSKCC) in New York City have been documenting gut microbiota disruption during allogeneic hematopoietic stem cell transplantation (allo-HSCT) and its role in frequent and potentially fatal bloodstream infections (BSIs) in the first 100 days after transplant.

Modulating microbiome composition to improve outcomes after allo-HSCT for hematological malignancies is a prime goal, and at the Gut Microbiota for Health (GMFH) World Summit 2025, Peled shared two new findings.

In one study, his team found that sucrose can exacerbate antibiotic-induced microbiome injury in patients undergoing allo-HSCT — a finding that “raises the question of whether our dietary recommendations [for] allo-HSCT patients are correct,” said Peled, assistant attending at MSKCC, during a session on the gut microbiome and oncology.

And in another study, they found that a rationally designed probiotic formulation may help lower the incidence of bacterial BSIs. In December 2024, the probiotic formulation (SER-155, Seres Therapeutics, Inc.) was granted breakthrough therapy designation by the FDA.

With immunotherapies more broadly, researchers are increasingly looking at diet and modulation of the microbiome to improve both treatment tolerance and efficacy, experts said at the meeting convened by the AGA and the European Society of Neurogastroenterology and Motility.

“Cancer patients and caregivers are asking, ‘What should I eat?’” said Carrie Daniel-MacDougall, PhD, MPH, a nutritional epidemiologist at the University of Texas MD Anderson Cancer Center in Houston. “They’re not just focused on side effects — they want a good outcome for their treatment, and they’re exploring a lot of dietary strategies [for which there] is not a lot of evidence.”

Clinicians are challenged by the fact that “we don’t typically collect dietary data in clinical trials of cancer drugs,” leaving them to extrapolate from evidence-based diet guidelines for cancer prevention, Daniel-MacDougall said.

But “I think that’s starting to shift,” she said, with the microbiome being increasingly recognized for its potential influences on therapeutic response and clinical trials underway looking at “a healthy dietary pattern not just for prevention but survival.”

 

Diet and Probiotics After allo-HSCT

The patterns of microbiota disruption during allo-HSCT — a procedure that includes antibiotic administration, chemotherapy, and sometimes irradiation — are characterized by loss of diversity and the expansion of potentially pathogenic organisms, most commonly Enterococcus, said Peled.

This has been demonstrated across transplantation centers. In a multicenter, international study published in 2020, the patterns of microbiota disruption and their impact on mortality were similar across MSK and other transplantation centers, with higher diversity of intestinal microbiota associated with lower mortality.

Other studies have shown that Enterococcus domination alone (defined arbitrarily as > 30% of fecal microbial composition) is associated with graft vs host disease and higher mortality after allo-HSCT and that intestinal domination by Proteobacteria coincides temporally with BSIs, he said.

Autologous fecal microbiota transplantation (FMT) has been shown to largely restore the microbiota composition the patient had before antibiotic treatment and allo-HSCT, he said, making fecal sample banking and posttreatment FMT a potential approach for reconstituting the gut microbiome and improving outcomes.

But “lately we’ve been very interested in diet for modulating [harmful] patterns” in the microbiome composition, Peled said.

In the new study suggesting a role for sugar avoidance, published last year as a bioRxiv preprint, Peled and his colleagues collected real-time dietary intake data (40,702 food entries) from 173 patients hospitalized for several weeks for allo-HSCT at MSK and analyzed it alongside longitudinally collected fecal samples. They used a Bayesian mixed-effects model to identify dietary components that may correlate with microbial disruption.

“What jumped out as very predictive of a low diversity fecal sample [and expansion of Enterococcus] in the 2 days prior to collection was the interaction between antibiotics and the consumption of sweets” — foods rich in simple sugars, Peled said. The relationship between sugar and the microbiome occurred only during periods of antibiotic exposure.

“And it was particularly perplexing because the foods that fall into the ‘sweets’ category are foods we encourage people to eat clinically when they’re not feeling well and food intake drops dramatically,” he said. This includes foods like nutritional drinks or shakes, Italian ice, gelatin dessert, and sports drinks.

(In a mouse model of post-antibiotic Enterococcus expansion, Peled and his co-investigators then validated the findings and ruled out the impact of any reductions in fiber.)

In addition to possibly revising dietary recommendations for patients undergoing allo-HSCT, the findings raise the question of whether avoiding sugar intake while on antibiotics, in general, is a way to mitigate antibiotic-induced dysbiosis, he said.

To test the role of probiotics, Peled and colleagues collaborated with Seres Therapeutics on a phase 1b trial of an oral combination (SER-155) of 16 fermented strains “selected rationally,” he said, for their ability to decolonize gut pathogens, improve gut barrier function (in vitro), and reduce gut inflammation and local immune activation.

After a safety lead-in, patients were randomized to receive SER-155 (20) or placebo (14) three times — prior to transplant, upon neutrophil engraftment (with vancomycin “conditioning”), and after transplant. “The strains succeeded in grafting in the [gastrointestinal] GI tract…and some of them persisted all the way through to day 100,” Peled said.

The incidence of pathogen domination was substantially lower in the probiotic recipients compared to an MSK historical control cohort, and the incidence of BSIs was significantly lower compared to the placebo arm (10% vs 43%, respectively, representing a 77% relative risk reduction), he said.

 

Diet and Immunotherapy Response: Trials at MD Anderson

One of the first trials Daniel-MacDougall launched at MD Anderson on diet and the microbiome randomized 55 patients who were obese and had a history of colorectal cancer or precancerous polyps to add a cup of beans to their usual diet or to continue their usual diet without beans. There was a crossover at 8 weeks in the 16-week BE GONE trial; stool and fasting blood were collected every 4 weeks.

“Beans are a prebiotic super-house in my opinion, and they’re also something this population would avoid,” said Daniel-MacDougall, associate professor in the department of epidemiology at MD Anderson and faculty director of the Bionutrition Research Core and Research Kitchen.

“We saw a modest increase in alpha diversity [in the intervention group] and similar trends with microbiota-derived metabolites” that regressed when patients returned to their usual diet, she said. The researchers also documented decreases in proteomic biomarkers of intestinal and systemic immune and inflammatory response.

The impact of diet on cancer survival was shown in subsequent research, including an observational study published in Science in 2021 of patients with melanoma receiving immune checkpoint blockade (ICB) treatment. “Patients who consumed insufficient dietary fiber at the start of therapy tended to do worse [than those reporting sufficient fiber intake],” with significantly lower progression-free survival, Daniel-MacDougall said.

“And interestingly, when we looked at dietary fiber [with and without] probiotic use, patients who had sufficient fiber but did not take probiotics did the best,” she said. [The probiotics were not endorsed or selected by their physicians.]

Now, the researchers at MD Anderson are moving into “precision nutrition” research, Daniel-MacDougall said, with a phase 2 randomized, double-blind trial of high dietary fiber intake (a target of 50 g/d from whole foods) vs a healthy control diet (20 g/d of fiber) in patients with melanoma receiving ICB.

The study, which is underway, is a fully controlled feeding study, with all meals and snacks provided by MD Anderson and macronutrients controlled. Researchers are collecting blood, stool, and tumor tissue (if available) to answer questions about the microbiome, changes in systemic and tissue immunity, disease response and immunotherapy toxicity, and other issues.

Peled disclosed IP licensing and research support from Seres Therapeutics; consulting with Da Volterra, MaaT Pharma, and CSL Behring; and advisory/equity with Postbiotics + Research LLC and Prodigy Biosciences. Daniel-MacDougall reported having no disclosures.

A version of this article appeared on Medscape.com.

WASHINGTON — For years, oncologist Jonathan Peled, MD, PhD, and his colleagues at Memorial Sloan Kettering Cancer Center (MSKCC) in New York City have been documenting gut microbiota disruption during allogeneic hematopoietic stem cell transplantation (allo-HSCT) and its role in frequent and potentially fatal bloodstream infections (BSIs) in the first 100 days after transplant.

Modulating microbiome composition to improve outcomes after allo-HSCT for hematological malignancies is a prime goal, and at the Gut Microbiota for Health (GMFH) World Summit 2025, Peled shared two new findings.

In one study, his team found that sucrose can exacerbate antibiotic-induced microbiome injury in patients undergoing allo-HSCT — a finding that “raises the question of whether our dietary recommendations [for] allo-HSCT patients are correct,” said Peled, assistant attending at MSKCC, during a session on the gut microbiome and oncology.

And in another study, they found that a rationally designed probiotic formulation may help lower the incidence of bacterial BSIs. In December 2024, the probiotic formulation (SER-155, Seres Therapeutics, Inc.) was granted breakthrough therapy designation by the FDA.

With immunotherapies more broadly, researchers are increasingly looking at diet and modulation of the microbiome to improve both treatment tolerance and efficacy, experts said at the meeting convened by the AGA and the European Society of Neurogastroenterology and Motility.

“Cancer patients and caregivers are asking, ‘What should I eat?’” said Carrie Daniel-MacDougall, PhD, MPH, a nutritional epidemiologist at the University of Texas MD Anderson Cancer Center in Houston. “They’re not just focused on side effects — they want a good outcome for their treatment, and they’re exploring a lot of dietary strategies [for which there] is not a lot of evidence.”

Clinicians are challenged by the fact that “we don’t typically collect dietary data in clinical trials of cancer drugs,” leaving them to extrapolate from evidence-based diet guidelines for cancer prevention, Daniel-MacDougall said.

But “I think that’s starting to shift,” she said, with the microbiome being increasingly recognized for its potential influences on therapeutic response and clinical trials underway looking at “a healthy dietary pattern not just for prevention but survival.”

 

Diet and Probiotics After allo-HSCT

The patterns of microbiota disruption during allo-HSCT — a procedure that includes antibiotic administration, chemotherapy, and sometimes irradiation — are characterized by loss of diversity and the expansion of potentially pathogenic organisms, most commonly Enterococcus, said Peled.

This has been demonstrated across transplantation centers. In a multicenter, international study published in 2020, the patterns of microbiota disruption and their impact on mortality were similar across MSK and other transplantation centers, with higher diversity of intestinal microbiota associated with lower mortality.

Other studies have shown that Enterococcus domination alone (defined arbitrarily as > 30% of fecal microbial composition) is associated with graft vs host disease and higher mortality after allo-HSCT and that intestinal domination by Proteobacteria coincides temporally with BSIs, he said.

Autologous fecal microbiota transplantation (FMT) has been shown to largely restore the microbiota composition the patient had before antibiotic treatment and allo-HSCT, he said, making fecal sample banking and posttreatment FMT a potential approach for reconstituting the gut microbiome and improving outcomes.

But “lately we’ve been very interested in diet for modulating [harmful] patterns” in the microbiome composition, Peled said.

In the new study suggesting a role for sugar avoidance, published last year as a bioRxiv preprint, Peled and his colleagues collected real-time dietary intake data (40,702 food entries) from 173 patients hospitalized for several weeks for allo-HSCT at MSK and analyzed it alongside longitudinally collected fecal samples. They used a Bayesian mixed-effects model to identify dietary components that may correlate with microbial disruption.

“What jumped out as very predictive of a low diversity fecal sample [and expansion of Enterococcus] in the 2 days prior to collection was the interaction between antibiotics and the consumption of sweets” — foods rich in simple sugars, Peled said. The relationship between sugar and the microbiome occurred only during periods of antibiotic exposure.

“And it was particularly perplexing because the foods that fall into the ‘sweets’ category are foods we encourage people to eat clinically when they’re not feeling well and food intake drops dramatically,” he said. This includes foods like nutritional drinks or shakes, Italian ice, gelatin dessert, and sports drinks.

(In a mouse model of post-antibiotic Enterococcus expansion, Peled and his co-investigators then validated the findings and ruled out the impact of any reductions in fiber.)

In addition to possibly revising dietary recommendations for patients undergoing allo-HSCT, the findings raise the question of whether avoiding sugar intake while on antibiotics, in general, is a way to mitigate antibiotic-induced dysbiosis, he said.

To test the role of probiotics, Peled and colleagues collaborated with Seres Therapeutics on a phase 1b trial of an oral combination (SER-155) of 16 fermented strains “selected rationally,” he said, for their ability to decolonize gut pathogens, improve gut barrier function (in vitro), and reduce gut inflammation and local immune activation.

After a safety lead-in, patients were randomized to receive SER-155 (20) or placebo (14) three times — prior to transplant, upon neutrophil engraftment (with vancomycin “conditioning”), and after transplant. “The strains succeeded in grafting in the [gastrointestinal] GI tract…and some of them persisted all the way through to day 100,” Peled said.

The incidence of pathogen domination was substantially lower in the probiotic recipients compared to an MSK historical control cohort, and the incidence of BSIs was significantly lower compared to the placebo arm (10% vs 43%, respectively, representing a 77% relative risk reduction), he said.

 

Diet and Immunotherapy Response: Trials at MD Anderson

One of the first trials Daniel-MacDougall launched at MD Anderson on diet and the microbiome randomized 55 patients who were obese and had a history of colorectal cancer or precancerous polyps to add a cup of beans to their usual diet or to continue their usual diet without beans. There was a crossover at 8 weeks in the 16-week BE GONE trial; stool and fasting blood were collected every 4 weeks.

“Beans are a prebiotic super-house in my opinion, and they’re also something this population would avoid,” said Daniel-MacDougall, associate professor in the department of epidemiology at MD Anderson and faculty director of the Bionutrition Research Core and Research Kitchen.

“We saw a modest increase in alpha diversity [in the intervention group] and similar trends with microbiota-derived metabolites” that regressed when patients returned to their usual diet, she said. The researchers also documented decreases in proteomic biomarkers of intestinal and systemic immune and inflammatory response.

The impact of diet on cancer survival was shown in subsequent research, including an observational study published in Science in 2021 of patients with melanoma receiving immune checkpoint blockade (ICB) treatment. “Patients who consumed insufficient dietary fiber at the start of therapy tended to do worse [than those reporting sufficient fiber intake],” with significantly lower progression-free survival, Daniel-MacDougall said.

“And interestingly, when we looked at dietary fiber [with and without] probiotic use, patients who had sufficient fiber but did not take probiotics did the best,” she said. [The probiotics were not endorsed or selected by their physicians.]

Now, the researchers at MD Anderson are moving into “precision nutrition” research, Daniel-MacDougall said, with a phase 2 randomized, double-blind trial of high dietary fiber intake (a target of 50 g/d from whole foods) vs a healthy control diet (20 g/d of fiber) in patients with melanoma receiving ICB.

The study, which is underway, is a fully controlled feeding study, with all meals and snacks provided by MD Anderson and macronutrients controlled. Researchers are collecting blood, stool, and tumor tissue (if available) to answer questions about the microbiome, changes in systemic and tissue immunity, disease response and immunotherapy toxicity, and other issues.

Peled disclosed IP licensing and research support from Seres Therapeutics; consulting with Da Volterra, MaaT Pharma, and CSL Behring; and advisory/equity with Postbiotics + Research LLC and Prodigy Biosciences. Daniel-MacDougall reported having no disclosures.

A version of this article appeared on Medscape.com.

WASHINGTON — For years, oncologist Jonathan Peled, MD, PhD, and his colleagues at Memorial Sloan Kettering Cancer Center (MSKCC) in New York City have been documenting gut microbiota disruption during allogeneic hematopoietic stem cell transplantation (allo-HSCT) and its role in frequent and potentially fatal bloodstream infections (BSIs) in the first 100 days after transplant.

Modulating microbiome composition to improve outcomes after allo-HSCT for hematological malignancies is a prime goal, and at the Gut Microbiota for Health (GMFH) World Summit 2025, Peled shared two new findings.

In one study, his team found that sucrose can exacerbate antibiotic-induced microbiome injury in patients undergoing allo-HSCT — a finding that “raises the question of whether our dietary recommendations [for] allo-HSCT patients are correct,” said Peled, assistant attending at MSKCC, during a session on the gut microbiome and oncology.

And in another study, they found that a rationally designed probiotic formulation may help lower the incidence of bacterial BSIs. In December 2024, the probiotic formulation (SER-155, Seres Therapeutics, Inc.) was granted breakthrough therapy designation by the FDA.

With immunotherapies more broadly, researchers are increasingly looking at diet and modulation of the microbiome to improve both treatment tolerance and efficacy, experts said at the meeting convened by the AGA and the European Society of Neurogastroenterology and Motility.

“Cancer patients and caregivers are asking, ‘What should I eat?’” said Carrie Daniel-MacDougall, PhD, MPH, a nutritional epidemiologist at the University of Texas MD Anderson Cancer Center in Houston. “They’re not just focused on side effects — they want a good outcome for their treatment, and they’re exploring a lot of dietary strategies [for which there] is not a lot of evidence.”

Clinicians are challenged by the fact that “we don’t typically collect dietary data in clinical trials of cancer drugs,” leaving them to extrapolate from evidence-based diet guidelines for cancer prevention, Daniel-MacDougall said.

But “I think that’s starting to shift,” she said, with the microbiome being increasingly recognized for its potential influences on therapeutic response and clinical trials underway looking at “a healthy dietary pattern not just for prevention but survival.”

 

Diet and Probiotics After allo-HSCT

The patterns of microbiota disruption during allo-HSCT — a procedure that includes antibiotic administration, chemotherapy, and sometimes irradiation — are characterized by loss of diversity and the expansion of potentially pathogenic organisms, most commonly Enterococcus, said Peled.

This has been demonstrated across transplantation centers. In a multicenter, international study published in 2020, the patterns of microbiota disruption and their impact on mortality were similar across MSK and other transplantation centers, with higher diversity of intestinal microbiota associated with lower mortality.

Other studies have shown that Enterococcus domination alone (defined arbitrarily as > 30% of fecal microbial composition) is associated with graft vs host disease and higher mortality after allo-HSCT and that intestinal domination by Proteobacteria coincides temporally with BSIs, he said.

Autologous fecal microbiota transplantation (FMT) has been shown to largely restore the microbiota composition the patient had before antibiotic treatment and allo-HSCT, he said, making fecal sample banking and posttreatment FMT a potential approach for reconstituting the gut microbiome and improving outcomes.

But “lately we’ve been very interested in diet for modulating [harmful] patterns” in the microbiome composition, Peled said.

In the new study suggesting a role for sugar avoidance, published last year as a bioRxiv preprint, Peled and his colleagues collected real-time dietary intake data (40,702 food entries) from 173 patients hospitalized for several weeks for allo-HSCT at MSK and analyzed it alongside longitudinally collected fecal samples. They used a Bayesian mixed-effects model to identify dietary components that may correlate with microbial disruption.

“What jumped out as very predictive of a low diversity fecal sample [and expansion of Enterococcus] in the 2 days prior to collection was the interaction between antibiotics and the consumption of sweets” — foods rich in simple sugars, Peled said. The relationship between sugar and the microbiome occurred only during periods of antibiotic exposure.

“And it was particularly perplexing because the foods that fall into the ‘sweets’ category are foods we encourage people to eat clinically when they’re not feeling well and food intake drops dramatically,” he said. This includes foods like nutritional drinks or shakes, Italian ice, gelatin dessert, and sports drinks.

(In a mouse model of post-antibiotic Enterococcus expansion, Peled and his co-investigators then validated the findings and ruled out the impact of any reductions in fiber.)

In addition to possibly revising dietary recommendations for patients undergoing allo-HSCT, the findings raise the question of whether avoiding sugar intake while on antibiotics, in general, is a way to mitigate antibiotic-induced dysbiosis, he said.

To test the role of probiotics, Peled and colleagues collaborated with Seres Therapeutics on a phase 1b trial of an oral combination (SER-155) of 16 fermented strains “selected rationally,” he said, for their ability to decolonize gut pathogens, improve gut barrier function (in vitro), and reduce gut inflammation and local immune activation.

After a safety lead-in, patients were randomized to receive SER-155 (20) or placebo (14) three times — prior to transplant, upon neutrophil engraftment (with vancomycin “conditioning”), and after transplant. “The strains succeeded in grafting in the [gastrointestinal] GI tract…and some of them persisted all the way through to day 100,” Peled said.

The incidence of pathogen domination was substantially lower in the probiotic recipients compared to an MSK historical control cohort, and the incidence of BSIs was significantly lower compared to the placebo arm (10% vs 43%, respectively, representing a 77% relative risk reduction), he said.

 

Diet and Immunotherapy Response: Trials at MD Anderson

One of the first trials Daniel-MacDougall launched at MD Anderson on diet and the microbiome randomized 55 patients who were obese and had a history of colorectal cancer or precancerous polyps to add a cup of beans to their usual diet or to continue their usual diet without beans. There was a crossover at 8 weeks in the 16-week BE GONE trial; stool and fasting blood were collected every 4 weeks.

“Beans are a prebiotic super-house in my opinion, and they’re also something this population would avoid,” said Daniel-MacDougall, associate professor in the department of epidemiology at MD Anderson and faculty director of the Bionutrition Research Core and Research Kitchen.

“We saw a modest increase in alpha diversity [in the intervention group] and similar trends with microbiota-derived metabolites” that regressed when patients returned to their usual diet, she said. The researchers also documented decreases in proteomic biomarkers of intestinal and systemic immune and inflammatory response.

The impact of diet on cancer survival was shown in subsequent research, including an observational study published in Science in 2021 of patients with melanoma receiving immune checkpoint blockade (ICB) treatment. “Patients who consumed insufficient dietary fiber at the start of therapy tended to do worse [than those reporting sufficient fiber intake],” with significantly lower progression-free survival, Daniel-MacDougall said.

“And interestingly, when we looked at dietary fiber [with and without] probiotic use, patients who had sufficient fiber but did not take probiotics did the best,” she said. [The probiotics were not endorsed or selected by their physicians.]

Now, the researchers at MD Anderson are moving into “precision nutrition” research, Daniel-MacDougall said, with a phase 2 randomized, double-blind trial of high dietary fiber intake (a target of 50 g/d from whole foods) vs a healthy control diet (20 g/d of fiber) in patients with melanoma receiving ICB.

The study, which is underway, is a fully controlled feeding study, with all meals and snacks provided by MD Anderson and macronutrients controlled. Researchers are collecting blood, stool, and tumor tissue (if available) to answer questions about the microbiome, changes in systemic and tissue immunity, disease response and immunotherapy toxicity, and other issues.

Peled disclosed IP licensing and research support from Seres Therapeutics; consulting with Da Volterra, MaaT Pharma, and CSL Behring; and advisory/equity with Postbiotics + Research LLC and Prodigy Biosciences. Daniel-MacDougall reported having no disclosures.

A version of this article appeared on Medscape.com.

SAN DIEGO — A novel generative artificial intelligence (AI) framework known as trajectory flow matching (TFM) can predict the need for red blood cell transfusion and mortality risk in intensive care unit (ICU) patients with acute gastrointestinal (GI) bleeding, researchers reported at Digestive Disease Week^® (DDW) 2025.

Acute GI bleeding is the most common cause of digestive disease–related hospitalization, with an estimated 500,000 hospital admissions annually. It’s known that predicting the need for red blood cell transfusion in the first 24 hours may improve resuscitation and decrease both morbidity and mortality.

However, an existing clinical score known as the Rockall Score does not perform well for predicting mortality, Xi (Nicole) Zhang, an MD-PhD student at McGill University, Montreal, Quebec, Canada, told attendees at DDW. With an area under the curve of 0.65-0.75, better prediction is needed, said Zhang, whose coresearchers included Dennis Shung, MD, MHS, PhD, director of Applied Artificial Intelligence at Yale University School of Medicine, New Haven, Connecticut.

“We’d like to predict multiple outcomes in addition to mortality,” said Zhang, who is also a student at the Mila-Quebec Artificial Intelligence Institute.

As a result, the researchers turned to the TFM approach, applying it to ICU patients with acute GI bleeding to predict both the need for transfusion and in-hospital mortality risk. The all-cause mortality rate is up to 11%, according to a 2020 study by James Y. W. Lau, MD, and colleagues. The rebleeding rate of nonvariceal upper GI bleeds is up to 10.4%. Zhang said the rebleeding rate for variceal upper gastrointestinal bleeding is up to 65%.

The AI method the researchers used outperformed a standard deep learning model at predicting the need for transfusion and estimating mortality risk.

 

Defining the AI Framework

“Probabilistic flow matching is a class of generative artificial intelligence that learns how a simple distribution becomes a more complex distribution with ordinary differential equations,” Zhang told GI & Hepatology News. “For example, if you had a few lines and shapes you could learn how it could become a detailed portrait of a face. In our case, we start with a few blood pressure and heart rate measurements and learn the pattern of blood pressures and heart rates over time, particularly if they reflect clinical deterioration with hemodynamic instability.”

Another way to think about the underlying algorithm, Zhang said, is to think about a river with boats where the river flow determines where the boats end up. “We are trying to direct the boat to the correct dock by adjusting the flow of water in the canal. In this case we are mapping the distribution with the first few data points to the distribution with the entire patient trajectory.”

The information gained, she said, could be helpful in timing endoscopic evaluation or allocating red blood cell products for emergent transfusion.

 

Study Details

The researchers evaluated a cohort of 2602 patients admitted to the ICU, identified from the publicly available MIMIC-III database. They divided the patients into a training set of 2342 patients and an internal validation set of 260 patients. Input variables were severe liver disease comorbidity, administration of vasopressor medications, mean arterial blood pressure, and heart rate over the first 24 hours.

Excluded was hemoglobin, since the point was to test the trajectory of hemodynamic parameters independent of hemoglobin thresholds used to guide red blood cell transfusion.

The outcome measures were administration of packed red blood cell transfusion within 24 hours and all-cause hospital mortality.

The TFM was more accurate than a standard deep learning model in predicting red blood cell transfusion, with an accuracy of 93.6% vs 43.2%; P ≤ .001. It was also more accurate at predicting all-cause in-hospital mortality, with an accuracy of 89.5% vs 42.5%, P = .01.

The researchers concluded that the TFM approach was able to predict the hemodynamic trajectories of patients with acute GI bleeding defined as deviation and outperformed the baseline from the measured mean arterial pressure and heart rate.

 

Expert Perspective

“This is an exciting proof-of-concept study that shows generative AI methods may be applied to complex datasets in order to improve on our current predictive models and improve patient care,” said Jeremy Glissen Brown, MD, MSc, an assistant professor of medicine and a practicing gastroenterologist at Duke University who has published research on the use of AI in clinical practice. He reviewed the study for GI & Hepatology News but was not involved in the research.

“Future work will likely look into the implementation of a version of this model on real-time data.” he said. “We are at an exciting inflection point in predictive models within GI and clinical medicine. Predictive models based on deep learning and generative AI hold the promise of improving how we predict and treat disease states, but the excitement being generated with studies such as this needs to be balanced with the trade-offs inherent to the current paradigm of deep learning and generative models compared to more traditional regression-based models. These include many of the same ‘black box’ explainability questions that have risen in the age of convolutional neural networks as well as some method-specific questions due to the continuous and implicit nature of TFM.”

Elaborating on that, Glissen Brown said: “TFM, like many deep learning techniques, raises concerns about explainability that we’ve long seen with convolutional neural networks — the ‘black box’ problem, where it’s difficult to interpret exactly how and why the model arrives at a particular decision. But TFM also introduces unique challenges due to its continuous and implicit formulation. Since it often learns flows without explicitly defining intermediate representations or steps, it can be harder to trace the logic or pathways it uses to connect inputs to outputs. This makes standard interpretability tools less effective and calls for new techniques tailored to these continuous architectures.”

“This approach could have a real clinical impact,” said Robert Hirten, MD, associate professor of medicine and artificial intelligence, Icahn School of Medicine at Mount Sinai, New York City, who also reviewed the study. “Accurately predicting transfusion needs and mortality risk in real time could support earlier, more targeted interventions for high-risk patients. While these findings still need to be validated in prospective studies, it could enhance ICU decision-making and resource allocation.”

“For the practicing gastroenterologist, we envision this system could help them figure out when to perform endoscopy in a patient admitted with acute gastrointestinal bleeding in the ICU at very high risk of exsanguination,” Zhang told GI & Hepatology News.

The approach, the researchers said, will be useful in identifying unique patient characteristics, make possible the identification of high-risk patients and lead to more personalized medicine.

Hirten, Zhang, and Shung had no disclosures. Glissen Brown reported consulting relationships with Medtronic, OdinVision, Doximity, and Olympus. The National Institutes of Health funded this study.

A version of this article appeared on Medscape.com.

SAN DIEGO — A novel generative artificial intelligence (AI) framework known as trajectory flow matching (TFM) can predict the need for red blood cell transfusion and mortality risk in intensive care unit (ICU) patients with acute gastrointestinal (GI) bleeding, researchers reported at Digestive Disease Week^® (DDW) 2025.

Acute GI bleeding is the most common cause of digestive disease–related hospitalization, with an estimated 500,000 hospital admissions annually. It’s known that predicting the need for red blood cell transfusion in the first 24 hours may improve resuscitation and decrease both morbidity and mortality.

However, an existing clinical score known as the Rockall Score does not perform well for predicting mortality, Xi (Nicole) Zhang, an MD-PhD student at McGill University, Montreal, Quebec, Canada, told attendees at DDW. With an area under the curve of 0.65-0.75, better prediction is needed, said Zhang, whose coresearchers included Dennis Shung, MD, MHS, PhD, director of Applied Artificial Intelligence at Yale University School of Medicine, New Haven, Connecticut.

“We’d like to predict multiple outcomes in addition to mortality,” said Zhang, who is also a student at the Mila-Quebec Artificial Intelligence Institute.

As a result, the researchers turned to the TFM approach, applying it to ICU patients with acute GI bleeding to predict both the need for transfusion and in-hospital mortality risk. The all-cause mortality rate is up to 11%, according to a 2020 study by James Y. W. Lau, MD, and colleagues. The rebleeding rate of nonvariceal upper GI bleeds is up to 10.4%. Zhang said the rebleeding rate for variceal upper gastrointestinal bleeding is up to 65%.

The AI method the researchers used outperformed a standard deep learning model at predicting the need for transfusion and estimating mortality risk.

 

Defining the AI Framework

“Probabilistic flow matching is a class of generative artificial intelligence that learns how a simple distribution becomes a more complex distribution with ordinary differential equations,” Zhang told GI & Hepatology News. “For example, if you had a few lines and shapes you could learn how it could become a detailed portrait of a face. In our case, we start with a few blood pressure and heart rate measurements and learn the pattern of blood pressures and heart rates over time, particularly if they reflect clinical deterioration with hemodynamic instability.”

Another way to think about the underlying algorithm, Zhang said, is to think about a river with boats where the river flow determines where the boats end up. “We are trying to direct the boat to the correct dock by adjusting the flow of water in the canal. In this case we are mapping the distribution with the first few data points to the distribution with the entire patient trajectory.”

The information gained, she said, could be helpful in timing endoscopic evaluation or allocating red blood cell products for emergent transfusion.

 

Study Details

The researchers evaluated a cohort of 2602 patients admitted to the ICU, identified from the publicly available MIMIC-III database. They divided the patients into a training set of 2342 patients and an internal validation set of 260 patients. Input variables were severe liver disease comorbidity, administration of vasopressor medications, mean arterial blood pressure, and heart rate over the first 24 hours.

Excluded was hemoglobin, since the point was to test the trajectory of hemodynamic parameters independent of hemoglobin thresholds used to guide red blood cell transfusion.

The outcome measures were administration of packed red blood cell transfusion within 24 hours and all-cause hospital mortality.

The TFM was more accurate than a standard deep learning model in predicting red blood cell transfusion, with an accuracy of 93.6% vs 43.2%; P ≤ .001. It was also more accurate at predicting all-cause in-hospital mortality, with an accuracy of 89.5% vs 42.5%, P = .01.

The researchers concluded that the TFM approach was able to predict the hemodynamic trajectories of patients with acute GI bleeding defined as deviation and outperformed the baseline from the measured mean arterial pressure and heart rate.

 

Expert Perspective

“This is an exciting proof-of-concept study that shows generative AI methods may be applied to complex datasets in order to improve on our current predictive models and improve patient care,” said Jeremy Glissen Brown, MD, MSc, an assistant professor of medicine and a practicing gastroenterologist at Duke University who has published research on the use of AI in clinical practice. He reviewed the study for GI & Hepatology News but was not involved in the research.

“Future work will likely look into the implementation of a version of this model on real-time data.” he said. “We are at an exciting inflection point in predictive models within GI and clinical medicine. Predictive models based on deep learning and generative AI hold the promise of improving how we predict and treat disease states, but the excitement being generated with studies such as this needs to be balanced with the trade-offs inherent to the current paradigm of deep learning and generative models compared to more traditional regression-based models. These include many of the same ‘black box’ explainability questions that have risen in the age of convolutional neural networks as well as some method-specific questions due to the continuous and implicit nature of TFM.”

Elaborating on that, Glissen Brown said: “TFM, like many deep learning techniques, raises concerns about explainability that we’ve long seen with convolutional neural networks — the ‘black box’ problem, where it’s difficult to interpret exactly how and why the model arrives at a particular decision. But TFM also introduces unique challenges due to its continuous and implicit formulation. Since it often learns flows without explicitly defining intermediate representations or steps, it can be harder to trace the logic or pathways it uses to connect inputs to outputs. This makes standard interpretability tools less effective and calls for new techniques tailored to these continuous architectures.”

“This approach could have a real clinical impact,” said Robert Hirten, MD, associate professor of medicine and artificial intelligence, Icahn School of Medicine at Mount Sinai, New York City, who also reviewed the study. “Accurately predicting transfusion needs and mortality risk in real time could support earlier, more targeted interventions for high-risk patients. While these findings still need to be validated in prospective studies, it could enhance ICU decision-making and resource allocation.”

“For the practicing gastroenterologist, we envision this system could help them figure out when to perform endoscopy in a patient admitted with acute gastrointestinal bleeding in the ICU at very high risk of exsanguination,” Zhang told GI & Hepatology News.

The approach, the researchers said, will be useful in identifying unique patient characteristics, make possible the identification of high-risk patients and lead to more personalized medicine.

Hirten, Zhang, and Shung had no disclosures. Glissen Brown reported consulting relationships with Medtronic, OdinVision, Doximity, and Olympus. The National Institutes of Health funded this study.

A version of this article appeared on Medscape.com.

SAN DIEGO — A novel generative artificial intelligence (AI) framework known as trajectory flow matching (TFM) can predict the need for red blood cell transfusion and mortality risk in intensive care unit (ICU) patients with acute gastrointestinal (GI) bleeding, researchers reported at Digestive Disease Week^® (DDW) 2025.

Acute GI bleeding is the most common cause of digestive disease–related hospitalization, with an estimated 500,000 hospital admissions annually. It’s known that predicting the need for red blood cell transfusion in the first 24 hours may improve resuscitation and decrease both morbidity and mortality.

However, an existing clinical score known as the Rockall Score does not perform well for predicting mortality, Xi (Nicole) Zhang, an MD-PhD student at McGill University, Montreal, Quebec, Canada, told attendees at DDW. With an area under the curve of 0.65-0.75, better prediction is needed, said Zhang, whose coresearchers included Dennis Shung, MD, MHS, PhD, director of Applied Artificial Intelligence at Yale University School of Medicine, New Haven, Connecticut.

“We’d like to predict multiple outcomes in addition to mortality,” said Zhang, who is also a student at the Mila-Quebec Artificial Intelligence Institute.

As a result, the researchers turned to the TFM approach, applying it to ICU patients with acute GI bleeding to predict both the need for transfusion and in-hospital mortality risk. The all-cause mortality rate is up to 11%, according to a 2020 study by James Y. W. Lau, MD, and colleagues. The rebleeding rate of nonvariceal upper GI bleeds is up to 10.4%. Zhang said the rebleeding rate for variceal upper gastrointestinal bleeding is up to 65%.

The AI method the researchers used outperformed a standard deep learning model at predicting the need for transfusion and estimating mortality risk.

 

Defining the AI Framework

“Probabilistic flow matching is a class of generative artificial intelligence that learns how a simple distribution becomes a more complex distribution with ordinary differential equations,” Zhang told GI & Hepatology News. “For example, if you had a few lines and shapes you could learn how it could become a detailed portrait of a face. In our case, we start with a few blood pressure and heart rate measurements and learn the pattern of blood pressures and heart rates over time, particularly if they reflect clinical deterioration with hemodynamic instability.”

Another way to think about the underlying algorithm, Zhang said, is to think about a river with boats where the river flow determines where the boats end up. “We are trying to direct the boat to the correct dock by adjusting the flow of water in the canal. In this case we are mapping the distribution with the first few data points to the distribution with the entire patient trajectory.”

The information gained, she said, could be helpful in timing endoscopic evaluation or allocating red blood cell products for emergent transfusion.

 

Study Details

The researchers evaluated a cohort of 2602 patients admitted to the ICU, identified from the publicly available MIMIC-III database. They divided the patients into a training set of 2342 patients and an internal validation set of 260 patients. Input variables were severe liver disease comorbidity, administration of vasopressor medications, mean arterial blood pressure, and heart rate over the first 24 hours.

Excluded was hemoglobin, since the point was to test the trajectory of hemodynamic parameters independent of hemoglobin thresholds used to guide red blood cell transfusion.

The outcome measures were administration of packed red blood cell transfusion within 24 hours and all-cause hospital mortality.

The TFM was more accurate than a standard deep learning model in predicting red blood cell transfusion, with an accuracy of 93.6% vs 43.2%; P ≤ .001. It was also more accurate at predicting all-cause in-hospital mortality, with an accuracy of 89.5% vs 42.5%, P = .01.

The researchers concluded that the TFM approach was able to predict the hemodynamic trajectories of patients with acute GI bleeding defined as deviation and outperformed the baseline from the measured mean arterial pressure and heart rate.

 

Expert Perspective

“This is an exciting proof-of-concept study that shows generative AI methods may be applied to complex datasets in order to improve on our current predictive models and improve patient care,” said Jeremy Glissen Brown, MD, MSc, an assistant professor of medicine and a practicing gastroenterologist at Duke University who has published research on the use of AI in clinical practice. He reviewed the study for GI & Hepatology News but was not involved in the research.

“Future work will likely look into the implementation of a version of this model on real-time data.” he said. “We are at an exciting inflection point in predictive models within GI and clinical medicine. Predictive models based on deep learning and generative AI hold the promise of improving how we predict and treat disease states, but the excitement being generated with studies such as this needs to be balanced with the trade-offs inherent to the current paradigm of deep learning and generative models compared to more traditional regression-based models. These include many of the same ‘black box’ explainability questions that have risen in the age of convolutional neural networks as well as some method-specific questions due to the continuous and implicit nature of TFM.”

Elaborating on that, Glissen Brown said: “TFM, like many deep learning techniques, raises concerns about explainability that we’ve long seen with convolutional neural networks — the ‘black box’ problem, where it’s difficult to interpret exactly how and why the model arrives at a particular decision. But TFM also introduces unique challenges due to its continuous and implicit formulation. Since it often learns flows without explicitly defining intermediate representations or steps, it can be harder to trace the logic or pathways it uses to connect inputs to outputs. This makes standard interpretability tools less effective and calls for new techniques tailored to these continuous architectures.”

“This approach could have a real clinical impact,” said Robert Hirten, MD, associate professor of medicine and artificial intelligence, Icahn School of Medicine at Mount Sinai, New York City, who also reviewed the study. “Accurately predicting transfusion needs and mortality risk in real time could support earlier, more targeted interventions for high-risk patients. While these findings still need to be validated in prospective studies, it could enhance ICU decision-making and resource allocation.”

“For the practicing gastroenterologist, we envision this system could help them figure out when to perform endoscopy in a patient admitted with acute gastrointestinal bleeding in the ICU at very high risk of exsanguination,” Zhang told GI & Hepatology News.

The approach, the researchers said, will be useful in identifying unique patient characteristics, make possible the identification of high-risk patients and lead to more personalized medicine.

Hirten, Zhang, and Shung had no disclosures. Glissen Brown reported consulting relationships with Medtronic, OdinVision, Doximity, and Olympus. The National Institutes of Health funded this study.

A version of this article appeared on Medscape.com.