Didn't the low fat diet fail?

The largest test of it did. 48,835 postmenopausal women were randomised for a mean 8.1 years and coronary heart disease, stroke, total cardiovascular disease and invasive breast cancer all came back null, with LDL down only 3.55 mg/dL. What it did not do is cause harm or weight gain, which was a live question at the time. And a different intervention, cutting saturated fat specifically rather than total fat, did reduce combined cardiovascular events by 17 percent across 15 trials.

But at the same calories, is low fat better than low carb?

For body fat loss, yes, and by an amount too small to plan around. In a metabolic ward with intake fixed, fat restriction produced 89 g/day of body fat loss against 53 g/day on carbohydrate restriction. That is a real, statistically significant difference of about 36 grams of fat a day. It is also the opposite direction to what the carbohydrate-insulin model predicts, which is why it gets cited so often.

What about the Ornish reversal trial?

It is real and it is not a diet trial. 48 patients were randomised to a package of a 10 percent fat vegetarian diet plus aerobic exercise, stress management, smoking cessation and group psychosocial support. 35 completed five-year imaging. Stenosis regressed in the experimental group and progressed substantially in controls, with roughly half the cardiac event rate. Which of the five components produced that is unmeasured, the sample is small, and the trial ran from 1986 to 1992, before current statin and revascularisation care.

Will I lose weight on it?

Against no intervention, yes, about 5.4 kg in the pooled trials. Against another active weight loss diet of similar intensity, no difference at all, and a low carbohydrate diet beats it by 1.15 kg. In the four-arm trial that varied fat from 20 to 40 percent of energy at matched calories, both groups lost 3.3 kg at two years and attendance at group sessions predicted the result better than the macronutrient split did.

Who does this pattern suit least?

People whose lipid problem is high triglycerides with low HDL. At constant weight this pattern produced little or no triglyceride effect, and in the pooled trials the higher-fat comparators won by the largest margin precisely in the comparisons where the triglyceride difference at follow-up was widest. That is the clearest signal anywhere in this section that the low fat and carbohydrate-restricted patterns suit different people.

Macronutrient-defined12 human studies cited

Low fat

Written by Aaron CuhaReviewed Sep 2026

Also called: Low fat diet, Ornish, Lifestyle Heart programme, Plant-based low fat

The largest dietary trial ever run put 48,835 women on it for eight years and found no reduction in coronary heart disease, stroke or breast cancer. At fixed calories it beats carbohydrate restriction for fat loss. Against other active diets it beats none of them.

What it actually is

Restriction of total dietary fat, conventionally to 20 to 30 percent of energy. The version most people mean when they say the diet failed is the public health version, around 20 percent of energy with more grains, fruit and vegetables. The version with the famous imaging result is different in kind: the Ornish Lifestyle Heart programme runs about 10 percent of energy from fat on a whole foods vegetarian diet, and it was never tested as a diet. It was tested as a five-part package that also included aerobic exercise, stress management training, smoking cessation and group psychosocial support. Nobody has separated the diet from the other four, and every claim made for the Ornish result is a claim about the package.

The mechanism its advocates propose

Two arguments, and they should be kept apart because the evidence treats them differently. The first is about energy: fat carries nine kilocalories per gram against four for carbohydrate and protein, so restricting it lowers the energy density of the diet and should reduce intake without portion control. The second is about lipids: saturated fat raises LDL, LDL causes atherosclerosis, and lowering dietary fat lowers LDL and therefore events. The second argument is specific and has been tested at the level of hard outcomes, which is more than most dietary hypotheses can say. The result depends heavily on whether the trial reduced total fat or targeted saturated fat specifically, and those two came out differently.

What happened when calories were controlled

Strongest design available: controlled

Energy intake was matched between groups, so a difference in the result is attributable to the composition of the diet rather than to how much was eaten.

At matched energy, fat restriction produces more body fat loss than carbohydrate restriction, which is the opposite of the popular claim and the effect is tiny in both directions. Against other active diets at similar intensity there is no weight advantage at all. The pattern's real calorie finding is a free-living one: in a metabolic ward, people offered a low fat plant-based diet ate 689 kcal a day less than on a ketogenic diet.

Hall 2015, NIH metabolic ward, 19 adults with obesity, isocaloric, each participant on both diets. Fat restriction produced 89 g/day of body fat loss against 53 g/day on carbohydrate restriction, p = 0.002. This is the single cleanest calorie-controlled comparison in the low fat literature and it favours low fat, by an amount measured in tens of grams a day.
Hall 2021 fed 20 adults inpatient and let them eat freely. The low fat plant-based diet led to 689 kcal/day less energy intake than the animal-based ketogenic diet over two weeks, p < 0.0001. That is a large effect and it is an appetite effect rather than a metabolic one, which is the same shape of finding the ketogenic pattern produces in the other direction in free-living trials.
POUNDS LOST randomised 811 adults to four diets for two years, all calorie reduced, with fat at 20 or 40 percent of energy and protein at 15 or 25 percent. At two years the 20 percent fat arms and the 40 percent fat arms had both lost 3.3 kg. Protein made no difference either, 3.0 against 3.6 kg. Attendance at group sessions predicted weight loss at 0.2 kg per session, which predicted the result better than any macronutrient did.
Tobias 2015 pooled 53 randomised trials and 68,128 participants comparing low fat against something else for at least a year. Low carbohydrate interventions beat low fat by 1.15 kg (95% CI 0.52 to 1.79). Low fat did not differ from other higher-fat weight loss interventions at all (0.36 kg, 95% CI minus 0.66 to 1.37). Low fat beat a usual diet by 5.41 kg. The authors' own interpretation is the useful one: the apparent effect of a low fat diet depends almost entirely on how intensive the comparison group's intervention was.
The saturated fat question is separate and comes out differently. A 2020 Cochrane review of 15 randomised trials and 56,675 participants found that reducing saturated fat cut combined cardiovascular events by 17 percent (RR 0.83, 95% CI 0.70 to 0.98, moderate certainty), with a number needed to treat of 56 over about four years in primary prevention. In the same review there was little or no effect on all-cause mortality (RR 0.96) or cardiovascular mortality (RR 0.95), and greater reductions in saturated fat produced greater reductions in events, which explained most of the heterogeneity. Reducing saturated fat is not the same intervention as reducing total fat, and conflating them is how this literature gets misreported in both directions.

What happened when people just ate it

The Women's Health Initiative Dietary Modification trial is the largest test of this pattern anyone will ever run and it is the reason the low fat era ended. 48,835 postmenopausal women, 40 percent randomised to intensive behavioural counselling aimed at 20 percent of energy from fat with more vegetables, fruit and grains, 60 percent to a comparison group receiving education materials, mean 8.1 years of follow-up. The intervention worked as a behaviour change: fat intake fell 8.2 percent of energy by year six and vegetable and fruit intake rose. The biological effect was small: LDL down 3.55 mg/dL, diastolic pressure down 0.31 mmHg, factor VIIc down 4.29 percent, with no significant change in HDL, triglycerides, glucose or insulin. The clinical result was null across the board. Coronary heart disease hazard ratio 0.97 (95% CI 0.90 to 1.06), stroke 1.02, total cardiovascular disease 0.98. Invasive breast cancer, which was a primary endpoint, 0.91 (0.83 to 1.01), not significant. Weight did not rebound, which answered a live question at the time: the intervention group lost 2.2 kg in the first year and was still 0.4 kg lighter at 7.5 years, p = 0.01. A 2017 analysis found the coronary result varied sharply by baseline status: in women without prior cardiovascular disease who were normotensive the hazard ratio was 0.70 (0.56 to 0.87), in hypertensive women 1.04, p for interaction 0.003, and the authors note that the coronary benefit in the healthy normotensive group was partially offset by an increase in ischaemic stroke risk. In the 3.4 percent with cardiovascular disease at baseline the coronary hazard ratio ran at 1.47 during intervention and 1.61 afterwards, which the authors attribute to confounding by post-randomisation statin use. Every one of those is a post hoc subgroup of a null primary result and should be read as hypothesis generating. By 2019, over a median 19.6 years, the primary outcomes were still null and two secondary signals had emerged: breast cancer followed by death at 0.84 (0.74 to 0.96) and diabetes requiring insulin at 0.87 (0.77 to 0.98). Look AHEAD is the other large free-living test, in 5,145 adults with type 2 diabetes given a calorie-reduced lower fat diet plus activity. It was stopped for futility at a median 9.6 years. Weight loss was 8.6 against 0.7 percent at one year and 6.0 against 3.5 percent at the end, and the cardiovascular composite hazard ratio was 0.95 (0.83 to 1.09), p = 0.51.

Protein, and whether it confounds the result

Low fat diets define themselves by what comes out and are silent on protein, and in practice the energy removed from fat gets replaced by carbohydrate rather than protein: in the Women's Health Initiative the intervention arm ran 8 to 10 percent of energy lower in fat and 8 to 10 percent higher in carbohydrate. Where protein was tested directly it did nothing on its own terms. In POUNDS LOST, 15 percent against 25 percent of energy from protein produced 3.0 and 3.6 kg of weight loss at two years, not significantly different. No large low fat trial has measured lean mass, and the Ornish programme bundles resistance-free aerobic exercise rather than resistance training, so the body composition question has not been asked here at all.

The measured intakes behind the protein figures are on the protein page.

What reliably moves

Marker	Direction	From
Body fat at fixed calories	Down, and by more than on carbohydrate restriction	89 g/day against 53 g/day in a metabolic ward with intake fixed, p = 0.002. Real, replicated in a pooled analysis of controlled feeding studies, and trivially small in absolute terms.
Energy intake eaten freely	Down sharply against a ketogenic comparator	689 kcal/day less over two weeks in an inpatient crossover, p < 0.0001. This is the pattern's strongest single result and it is an appetite finding.
LDL cholesterol	Down, and only slightly in free-living conditions	3.55 mg/dL in the Women's Health Initiative across nearly 49,000 women, against a fat reduction of 8.2 percent of energy. Larger reductions appear when food is provided and the fat target is severe.
Combined cardiovascular events when saturated fat specifically is cut	Down 17 percent	RR 0.83 across 15 trials and 56,675 participants, with a dose response: greater saturated fat reduction gave greater event reduction. All-cause and cardiovascular mortality showed little or no effect.
Coronary heart disease, stroke and breast cancer when total fat is cut	No effect	Hazard ratios 0.97, 1.02 and 0.91 over a mean 8.1 years in 48,835 women, all crossing 1. Still null on the primary outcomes at 19.6 years median follow-up.
Body weight against a usual diet	Down	5.41 kg in the pooled trials. Against another active weight loss diet of similar intensity, no difference at all.
Triglycerides	Little or no effect, and the direction matters for who this suits	The saturated fat Cochrane review found little or no effect on serum triglycerides. In the pooled low fat weight loss trials, higher-fat arms won by 1.38 kg specifically in the comparisons where the triglyceride gap at follow-up was at least 0.06 mmol/L, which is the signal that this pattern suits a different person than carbohydrate restriction does.

Long term, and hard outcomes

This pattern has the longest and largest randomised follow-up of any diet on this site and the answer it gives is mostly no. The Women's Health Initiative ran an 8.5-year median intervention and has followed participants for a median 19.6 years, and its primary endpoints of breast cancer, colorectal cancer and coronary heart disease were null over the intervention period and over cumulative follow-up. The Ornish evidence sits at the opposite extreme of size and the opposite extreme of intervention intensity, and it has to be reported with both halves. 48 patients with moderate to severe coronary disease were randomised and 35 completed five-year quantitative coronary arteriography. Average percent diameter stenosis in the experimental group fell 1.75 percentage points at one year and 3.1 points at five. In the control group it rose 2.3 points at one year and 11.8 points at five, p = 0.001 between groups. Cardiac events numbered 25 in 28 experimental patients against 45 in 20 control patients, risk ratio for the control group 2.47 (95% CI 1.48 to 4.20). That is a striking result and it carries four caveats that are not optional: the intervention was a five-component lifestyle package and the diet cannot be isolated from the exercise, stress management, smoking cessation and group support; 48 people were randomised and 35 finished; the trial ran from 1986 to 1992, before current statin and revascularisation standards, so the control group's trajectory is not today's control group's trajectory; and the same investigators' later work, a 93-man randomised trial in early prostate cancer finding PSA down 4 percent against up 6 percent (p = 0.016), is the same package tested again rather than an independent confirmation of the diet. Where the free-living evidence is large it is null, and where it is dramatic it is small and bundled.

Citations

Human2006
Low-fat dietary pattern and risk of cardiovascular disease: the Women's Health Initiative Randomized Controlled Dietary Modification Trial.
JAMA
48,835 postmenopausal women, mean 8.1 years. Fat intake fell 8.2 percent of energy by year six. LDL down 3.55 mg/dL, diastolic pressure down 0.31 mmHg. Coronary heart disease HR 0.97 (0.90 to 1.06), stroke 1.02, cardiovascular disease 0.98. All null.
Human2006
Low-fat dietary pattern and risk of invasive breast cancer: the Women's Health Initiative Randomized Controlled Dietary Modification Trial.
JAMA
Same trial, the primary cancer endpoint. 655 cases in the intervention group against 1,072 in the comparison group, hazard ratio 0.91 (95% CI 0.83 to 1.01), not significant over 8.1 years.
Human2006
Low-fat dietary pattern and weight change over 7 years: the Women's Health Initiative Dietary Modification Trial.
JAMA
Intervention women lost 2.2 kg in the first year (p < 0.001) and remained lighter than comparison women by 1.9 kg at one year and 0.4 kg at 7.5 years (p = 0.01), with no tendency to weight gain in any subgroup. No calorie restriction goal was set.
Human2017
Low-fat dietary pattern and cardiovascular disease: results from the Women's Health Initiative randomized controlled trial.
American Journal of Clinical Nutrition
Incidence of coronary heart disease and total cardiovascular disease did not differ overall. In women without prior cardiovascular disease the coronary hazard ratio was 0.70 (0.56 to 0.87) if normotensive and 1.04 if hypertensive, p for interaction 0.003, with the benefit partially offset by increased ischaemic stroke risk. Post hoc subgroups of a null trial.
Human2019
Low-Fat Dietary Pattern among Postmenopausal Women Influences Long-Term Cancer, Cardiovascular Disease, and Diabetes Outcomes.
Journal of Nutrition
The long-term overview. Primary outcomes remained null over the intervention period and over cumulative follow-up. Over a median 19.6 years, breast cancer followed by death HR 0.84 (0.74 to 0.96) and diabetes requiring insulin 0.87 (0.77 to 0.98).
Human1990
Can lifestyle changes reverse coronary heart disease? The Lifestyle Heart Trial.
The Lancet
The original Ornish trial. No abstract is indexed in PubMed, so no numbers are quoted from it here; the five-year results in the 1998 report below carry the figures. The intervention was a package of a 10 percent fat vegetarian diet, aerobic exercise, stress management, smoking cessation and group support.
Human1998
Intensive lifestyle changes for reversal of coronary heart disease.
JAMA
48 patients randomised, 35 completed five-year arteriography. Percent diameter stenosis fell 1.75 points at one year and 3.1 at five in the experimental group; it rose 2.3 and 11.8 points in controls, p = 0.001. Cardiac events 25 in 28 experimental patients against 45 in 20 controls, control risk ratio 2.47 (1.48 to 4.20). A five-component lifestyle package, not a diet trial.
Human2005
Intensive lifestyle changes may affect the progression of prostate cancer.
Journal of Urology
93 men with early prostate cancer who had declined conventional treatment, randomised to the same lifestyle package or usual care for one year. PSA fell 4 percent against a 6 percent rise (p = 0.016) and serum-stimulated LNCaP cell growth was inhibited 70 percent against 9 percent (p < 0.001). Again a package, not a diet.
Review2015
Effect of low-fat diet interventions versus other diet interventions on long-term weight change in adults: a systematic review and meta-analysis.
Lancet Diabetes and Endocrinology
53 trials, 68,128 participants, at least one year. Low carbohydrate beat low fat by 1.15 kg (0.52 to 1.79). Low fat did not differ from other higher-fat weight loss interventions (0.36 kg, minus 0.66 to 1.37). Low fat beat a usual diet by 5.41 kg. The apparent effect depends on the intensity of the comparison group's intervention.
Review2020
Reduction in saturated fat intake for cardiovascular disease.
Cochrane Database of Systematic Reviews
15 randomised trials, 56,675 participants. Reducing saturated fat cut combined cardiovascular events 17 percent (RR 0.83, 0.70 to 0.98, moderate certainty), number needed to treat 56 in primary prevention. Little or no effect on all-cause mortality (0.96) or cardiovascular mortality (0.95). Greater saturated fat reduction produced greater event reduction.
Human2013
Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes.
New England Journal of Medicine
Look AHEAD. 5,145 adults with type 2 diabetes on a calorie-reduced lower fat diet plus activity, stopped for futility at median 9.6 years. Weight loss 8.6 against 0.7 percent at one year and 6.0 against 3.5 percent at the end. Cardiovascular composite HR 0.95 (0.83 to 1.09), p = 0.51.
Human2009
Comparison of weight-loss diets with different compositions of fat, protein, and carbohydrates.
New England Journal of Medicine
POUNDS LOST. 811 adults, four calorie-reduced diets, two years. 20 percent fat and 40 percent fat arms both lost 3.3 kg; 15 percent and 25 percent protein arms lost 3.0 and 3.6 kg; 65 percent and 35 percent carbohydrate arms 2.9 and 3.4 kg, all p > 0.20. Session attendance predicted weight loss at 0.2 kg per session.
Human2015
Calorie for Calorie, Dietary Fat Restriction Results in More Body Fat Loss than Carbohydrate Restriction in People with Obesity.
Cell Metabolism
NIH metabolic ward, 19 adults, isocaloric, each participant on both diets. Fat restriction produced 89 g/day of body fat loss against 53 g/day on carbohydrate restriction, p = 0.002. The cleanest calorie-controlled comparison, and the effect size is tens of grams.
Human2021
Effect of a plant-based, low-fat diet versus an animal-based, ketogenic diet on ad libitum energy intake.
Nature Medicine
20 adults, inpatient, ad libitum crossover. The low fat plant-based diet led to 689 kcal/day less energy intake over two weeks, p < 0.0001. An appetite result, in a ward, over two weeks.

What people report

These are uncontrolled self-reports, not evidence. They are here because they tell you what to expect and what to watch for, which the trial literature does not. They cannot tell you whether anything works.

Large food volume for the calories, described as the main reason it feels sustainable, and consistent with the 689 kcal/day intake difference measured in a ward.
Strong early weight loss followed by a plateau, which matches the trial data precisely: 2.2 kg in year one and 0.4 kg of difference remaining at 7.5 years.
Hunger returning once the initial novelty of the food changes wears off, reported often and not distinguishable in self-report from ordinary adherence drift.
No change in cholesterol, reported by a substantial minority and consistent with the 3.55 mg/dL average LDL reduction achieved in the largest trial.
Digestive complaints and gas as fibre rises, usually described as settling within weeks.
Difficulty with satiety and with cooking without fat, and difficulty eating out, the two most cited practical reasons for stopping.
A vocal community around the Ornish and whole-food plant-based versions reporting reversal of angina and reduced medication, which is what the trial reported in 28 people and which the trial cannot attribute to the diet alone.

Sources: Whole-food plant-based and Ornish programme communities, general weight loss forums, and the adherence data inside the trials, which for the Women's Health Initiative is unusually well documented because intake was measured repeatedly across eight years. Uncontrolled self-report otherwise.

Who this is wrong for

Anyone reading the Ornish arteriography result as a diet result. Five components were changed at once in 48 randomised patients, 35 of whom completed, in a pre-statin comparison. The package produced the result. Which part did is unmeasured.
Anyone whose lipid problem is high triglycerides with low HDL. At constant weight this pattern produced little or no triglyceride effect, and in the pooled trials the higher-fat comparators won by more precisely where the triglyceride gap was widest.
Anyone expecting the fat reduction to do the work by itself. Against another active diet of similar intensity there is no weight difference, and in the largest trial ever run the difference against a comparison group at 7.5 years was 0.4 kg.
Anyone who takes the Women's Health Initiative subgroups as guidance. The coronary hazard ratio of 0.70 in healthy normotensive women and the ratios above 1 in women with prior cardiovascular disease are both post hoc subgroups of a null primary result, and the authors attribute the latter to confounding by statin use.
Anyone conflating cutting total fat with cutting saturated fat. The saturated fat trials reduced combined cardiovascular events by 17 percent. The total fat trial did not reduce them at all.
Anyone attempting a 10 percent fat diet without support. The only trials that achieved and sustained it supplied intensive group sessions, and the free-living trial that aimed at 20 percent achieved a reduction of 8.2 percent of energy across eight years with intensive counselling.

Questions

Didn't the low fat diet fail?: The largest test of it did. 48,835 postmenopausal women were randomised for a mean 8.1 years and coronary heart disease, stroke, total cardiovascular disease and invasive breast cancer all came back null, with LDL down only 3.55 mg/dL. What it did not do is cause harm or weight gain, which was a live question at the time. And a different intervention, cutting saturated fat specifically rather than total fat, did reduce combined cardiovascular events by 17 percent across 15 trials.
But at the same calories, is low fat better than low carb?: For body fat loss, yes, and by an amount too small to plan around. In a metabolic ward with intake fixed, fat restriction produced 89 g/day of body fat loss against 53 g/day on carbohydrate restriction. That is a real, statistically significant difference of about 36 grams of fat a day. It is also the opposite direction to what the carbohydrate-insulin model predicts, which is why it gets cited so often.
What about the Ornish reversal trial?: It is real and it is not a diet trial. 48 patients were randomised to a package of a 10 percent fat vegetarian diet plus aerobic exercise, stress management, smoking cessation and group psychosocial support. 35 completed five-year imaging. Stenosis regressed in the experimental group and progressed substantially in controls, with roughly half the cardiac event rate. Which of the five components produced that is unmeasured, the sample is small, and the trial ran from 1986 to 1992, before current statin and revascularisation care.
Will I lose weight on it?: Against no intervention, yes, about 5.4 kg in the pooled trials. Against another active weight loss diet of similar intensity, no difference at all, and a low carbohydrate diet beats it by 1.15 kg. In the four-arm trial that varied fat from 20 to 40 percent of energy at matched calories, both groups lost 3.3 kg at two years and attendance at group sessions predicted the result better than the macronutrient split did.
Who does this pattern suit least?: People whose lipid problem is high triglycerides with low HDL. At constant weight this pattern produced little or no triglyceride effect, and in the pooled trials the higher-fat comparators won by the largest margin precisely in the comparisons where the triglyceride difference at follow-up was widest. That is the clearest signal anywhere in this section that the low fat and carbohydrate-restricted patterns suit different people.