Short eating windows and heart deaths

In 2024 a widely reported conference abstract linked eating within less than eight hours a day — tighter than the popular 16:8 pattern — to a 91% higher risk of dying from cardiovascular disease (heart disease and stroke). The full paper found a bigger number, not a smaller one, and a second analysis of the same US survey points the same way — but both rest on people who were never shown to be fasting on purpose, and the signal is a contested association, not proof that time-restricted eating harms the heart.

Don't let it put you off a moderate overnight fast. If you have heart disease or diabetes, avoid eating windows of eight hours or less; for everyone else, eating across 8–12 hours a day remains the sensible default.

What the study is

The data come from the US National Health and Nutrition Examination Survey (NHANES), which asks a nationally representative sample of adults to report everything they ate over two separate days. Researchers took 19,831 adults surveyed between 2003 and 2018, estimated how many hours a day each one spread their eating across, and linked them to death records. Over a median of about eight years, 2,794 died: 833 from cardiovascular disease and 646 from cancer.^[1]

People eating within under eight hours were compared with those eating across 12–14 hours, a reference band the authors chose, by their own account arbitrarily, around the sample average of about 12.8 hours. The 2024 abstract had compared against 12–16 hours, so its 91% and the paper's figure are not strictly like for like. The analysis was first presented at an American Heart Association meeting in 2024, where the press release's "91% higher risk" drew heavy coverage,^[2] and was published in full in 2025.

What it found

Cardiovascular death was more than double in the under-eight-hour group (a hazard ratio of 2.35 — the ratio of death rates between the two groups). That held across all fourteen of the authors' sensitivity checks, including moving the reference window, excluding deaths in the first year, and adjusting for measured blood pressure, blood sugar and cholesterol.

Cancer deaths showed nothing — a 27% difference well within the range expected from chance.

Death from any cause was borderline. It was about 40% higher, with a range that only just clears no effect (hazard ratio 1.40, 95% confidence interval 1.01–1.94; the bracketed range is where the true effect most plausibly lies, and a range that reaches 1.0 is compatible with no effect at all). Unlike the cardiovascular result, it did not hold up under most of the sensitivity analyses.

The risk concentrated in people who were already ill. Among those with existing cardiovascular disease or diabetes, the cardiovascular hazard was more than three times higher — but on just 21 cardiovascular deaths among 79 people. No such link appeared in people with high blood pressure, high cholesterol or kidney disease.

In absolute terms, the paper puts the excess at about 0.7 percentage points of cardiovascular death over fifteen years from age 50 in the sample overall, rising to about 3.8 points in people with existing cardiovascular disease or diabetes, and higher at older starting ages.

In a subset scanned for body composition, the short-window group also carried about 2.7 kg less lean mass — another association, but consistent with the muscle-loss concern that runs through fasting trials.

Why the headline is probably too strong

It rests on a small group. The under-eight-hour group was 383 people, roughly 2% of the sample, contributing 33 cardiovascular deaths. Estimates built on that few events are fragile.

These were not deliberate fasters. The short eaters were younger, more often men, more often smokers and food-insecure, with lower incomes, poorer diets, less exercise, and more often skipped breakfast. They were never shown to be fasting for their health, and the authors say so themselves: a short eating duration in the real world "may not be comparable to 8-h TRE for the intentional pursuit of health". The sample was also too small to separate early from late eight-hour windows — the distinction the trial evidence suggests matters most (see Fasting).

Experts pushed back — at the earlier version. Thirty-four researchers who study time-restricted eating signed a joint letter answering the American Heart Association press release. They objected to publicising an unreviewed poster, and argued that leftover confounding, above all by smoking, could account for the signal, that the short-window group was too small, and that two days of dietary records cannot capture long-term eating habits.^[3] Several of those points still apply to the full paper, though it added adjustment for measured blood pressure, blood sugar and cholesterol. The letter came from researchers whose own work is on time-restricted eating, and it answered the 2024 abstract; no peer-reviewed rebuttal to the fuller 2025 publication has appeared, and the published version reported a larger cardiovascular hazard than the abstract did, not a smaller one.

The authors' own closing hedge is the useful one: even if the association vanished once all confounding were removed, the absence of a benefit would still not support eating in a short window for the long term.

The second analysis: a U-shape

A second team analysed the same survey years with a broader sample — 33,052 adults with two days of dietary records, followed for a median of about eight years. They found a U-shaped curve: the lowest mortality at an eating window of about 11–12 hours, with higher death rates at both ends — about a third higher for windows under eight hours, and about a quarter higher for windows of 15 hours or more. The cardiovascular signal for short windows was most pronounced in older adults, men and white participants. Excluding deaths in the first year did not remove it, and the authors argue that the crude two-day dietary measure would push their estimates toward zero, not away from it.^[4]

Because both analyses draw on the same national survey over the same years, many of the same people are likely in both. It is a second look at the same data that reaches a similar answer, not an independent replication.

What it can't tell you

Whether deliberate 16:8 is harmful. Neither analysis studied people choosing a window for health, and no trial has followed time-restricted eating long enough to measure deaths.
Anything about very short windows. Sustained eating windows under six hours have not been studied for long-term safety at all.
Cause. Survey data cannot show it, and the trials only measured risk markers, where the picture is mixed: windows under eight hours lowered fasting insulin but ranked worse than longer ones for fasting glucose, triglycerides and LDL cholesterol, and the rankings by window length were inconsistent overall.^[5]

What to do

Keep a moderate window as the default — a 12–14 hour overnight fast, eating across roughly 10–12 hours.
With heart disease or diabetes, avoid windows of eight hours or less. That is where both the risk and the uncertainty concentrate.
Don't read "91%" or "double" as a fact about deliberate 16:8. It describes a small, socially disadvantaged group of short eaters in a survey.
If you do eat in a short window, start it with breakfast and end it early, and keep your protein up.