Pancreatitis patients who started eating within two days went home sooner. The ones kept waiting were already sicker

For a long time, the standard response to acute pancreatitis was to stop food altogether and let the pancreas rest. That advice has been turning around for years, and a study published in Frontiers in Nutrition on 29 October adds a large set of everyday hospital records to the case for feeding people early. It also shows, quite clearly, why records like these can't settle the question on their own.

Researchers at the Second Affiliated Hospital of Hainan Medical University in Haikou went through four years of admissions, January 2022 to December 2025. After removing duplicates, people with severe disease, chronic pancreatitis, cancer, earlier pancreatic surgery and patchy notes, they had 650 adults with mild or moderately severe acute pancreatitis. Most cases came from gallstones or other bile duct disease (55%), followed by very high blood fats (19%) and alcohol (14%). Anyone who started eating or tube feeding within 48 hours of arriving counted as early (392 people). The other 258 counted as delayed.

One detail matters before anything else. The paper's term is enteral nutrition, which sounds like a feeding tube, but here it covers ordinary eating too. In 87% of patients the first nutrition was simply food or drink by mouth. Only 13% started with a tube. So this is mostly a study of how soon people were allowed to eat again.

On the face of it, early eating wins everywhere. Median hospital stay was 5.5 days against 7.4. Pain settled in 2.0 days against 2.7. Feeding problems such as vomiting, bloating or worsening pain after a meal hit 8.7% of early feeders and 18.6% of the delayed group. ICU transfers were 3.3% against 8.1%. Deaths (8 in all) and readmissions within 30 days did not differ significantly.

Now the catch. The delayed group arrived sicker. They were more often moderately severe (25% against 17%), more often vomiting (68% against 53%), more often in moderate or severe pain, with higher inflammation (median CRP 65 against 45 mg/L), lower albumin and higher nutritional risk scores. And the reasons written in their notes for waiting were persistent vomiting (32%), pain (26%), a gut that had stopped moving (16%) and a doctor's instruction to fast (31%). People who feel well enough to eat get fed sooner, and people who feel well also tend to go home sooner. The authors put it plainly in their discussion: the differences 'may reflect disease severity rather than feeding time'.

They did try to separate the two. After adjusting for 12 factors, including severity, CRP, albumin and vomiting, early feeding was still linked to a stay 1.21 days shorter, down from the raw gap of about 1.9 days in the medians. Each extra 12 hours of waiting went with about 0.28 more days in hospital. Adjustment helps, but it can only correct for what was written down, and a nurse's sense that a patient 'isn't ready' rarely is. The rarer outcomes (ICU, procedures, deaths) had too few events for the adjusted models to be trusted, which the authors flag, and the many secondary results were not corrected for multiple testing.

What does this add? Not proof, but a real-world picture that agrees with randomised trials and guidelines that already favour feeding people with mild pancreatitis as soon as they can tolerate it, rather than fasting by habit. The useful number may be the 31% of the delayed group whose notes record a doctor's order to fast. Those patients, not the ones who couldn't keep food down, are where a feed-early policy could make a difference. That's for hospital teams to test properly, ideally in a trial.