Butter: the answer depends on what replaces it

Butter on its own barely registers. Nine cohorts covering 636 thousand people and 6.5 million person-years found that each extra tablespoon a day went with an all-cause mortality risk of 1.01, and the same pooling found no randomised trials of butter at all. The picture changes as soon as something takes its place. Across 54 trials, every oil tested lowered LDL cholesterol more than butter did when it replaced a tenth of the calories, by 0.23 to 0.42 mmol/l, and so did beef fat. The reviewers who ran that network rated most of their own cholesterol comparisons low or moderate quality, downgraded for few trials, unclear risk of bias and imprecision. One trial did feed people a real butter product rather than modelling the swap, and it used ghee. Against the same calories of olive oil, the ghee diet raised fasting apolipoprotein B by 0.09 g/l and non-HDL cholesterol by 0.53 mmol/l in thirty healthy adults over four weeks.

The numbers, per 100 g

How much is actually in it

• One reason studies of dairy fat disagree is that the cows were fed differently. — 16 studies of high-fat dairy and adiposity plus further studies of metabolic health, diabetes and cardiovascular disease (n = 16 studies on adiposity) Evidence C source
  11 of 16 observational studies of high-fat dairy foods, not of butter itself, found an inverse association with measures of adiposity; metabolic health studies showed inverse or no association; diabetes and cardiovascular studies were inconsistent; the authors name residual confounding, which high-fat dairy foods were eaten, and pasture against grain feeding as sources of the variation

What cooking and storage change

• Every oil tested lowered LDL by 0.23 to 0.42 mmol/l when it took the place of a tenth of the calories from butter across 54 randomised trials, and so did beef fat. — 54 randomised trials (n = 54 trials) Evidence A source
  Safflower, sunflower, rapeseed, flaxseed, corn, olive, soybean, palm and coconut oil and beef fat all reduced LDL cholesterol more than butter, by 0.23 to 0.42 mmol/l per 10% isocaloric exchange; safflower oil ranked first for LDL and total cholesterol, palm oil first for triglycerides and coconut oil first for HDL
• Clarified butter raised two blood markers of heart risk compared with the same calories of olive oil: in 30 healthy adults over four weeks, fasting apolipoprotein B by 0.09 g/l and non-HDL cholesterol by 0.53 mmol/l. — 30 healthy men and women (n = 30) Evidence B source
  Against olive oil, the ghee diet raised fasting apolipoprotein B by 0.09 g/l (95% CI 0.02 to 0.17, P = 0.018) and non-HDL cholesterol by 0.53 mmol/l (0.01 to 1.05, P = 0.046); LDL cholesterol did not differ significantly (0.29, -0.05 to 0.63, P = 0.092) and neither did the total to HDL cholesterol ratio, triglycerides, glucose, insulin or PAI-1; body weight was unchanged

What it does to your health

• Over 33 years the highest quarter of butter intake was associated with 15 percent higher total mortality, and the highest quarter of plant oil intake with 16 percent lower. — 221 054 US adults free of cancer, cardiovascular disease, diabetes and neurodegenerative disease at baseline, followed up to 33 years, 50 932 deaths (n = 221 054) Evidence C source
  Highest against lowest quartile of butter gave hazard ratio 1.15 for total mortality (95% CI 1.08 to 1.22, P for trend < .001); highest against lowest total plant-based oil gave 0.84 (0.79 to 0.90); canola, soybean and olive oils each showed lower total mortality per 5 g/day increment

What people believe that the data does not support

• A label saying reduced fat butter is measured against a number written into US statute, not against the tub beside it. — US butter labels Evidence E source
  Any percent fat reduction claim on butter must be calculated against the 80 percent milkfat requirement set by the statutory standard for butter at 21 U.S.C. 321a, and the claim must also satisfy the general rules of 21 CFR 101.13

Who this works differently for

• Cutting saturated fat bought almost nothing for people at low heart risk and a real amount for those at high risk. — 17 randomised trials with 66 337 adults, with and without existing cardiovascular disease (n = 66 337) Evidence A source
  Low to moderate certainty of reduced all-cause mortality (RR 0.96, 95% CI 0.88 to 1.06), cardiovascular mortality (0.93, 0.77 to 1.11), nonfatal myocardial infarction (0.86, 0.70 to 1.06) and stroke (0.83, 0.58 to 1.19); absolute reductions fell below the thresholds for low-risk people and above them for high-risk people; replacing saturated with polyunsaturated fat gave nonfatal myocardial infarction RR 0.75 (0.58 to 0.99, P for interaction 0.05)

What is still unknown

• The same review that ranked butter last rated most of its own LDL cholesterol comparisons, drawn from 54 trials, low or moderate in quality. — the 54 trials pooled in the network (n = 54 trials) Evidence A source
  Quality for LDL cholesterol was rated low or moderate for most comparisons based on mixed evidence and mainly low or very low for indirect comparisons; total cholesterol was mostly moderate, HDL mainly moderate, triglycerides moderate or high; the downgrades came from few trials, unclear risk of bias, imprecision and inconsistency
• Asked whether butter eaters die sooner, nine cohorts covering 636 thousand people answered with almost nothing. — 9 publications covering 15 country-specific cohorts, 636 151 participants, 6.5 million person-years, 28 271 deaths (n = 636 151) Evidence C source
  Per 14 g (one tablespoon) a day, all-cause mortality RR 1.01 (95% CI 1.00 to 1.03, P = 0.045); cardiovascular disease 1.00 (0.98 to 1.02); coronary heart disease 0.99 (0.96 to 1.03); stroke 1.01 (0.98 to 1.03); diabetes incidence 0.96 (0.93 to 0.99, P = 0.021); no randomised trials were found and no heterogeneity or publication bias was detected

The bottom line

How much butter matters depends on your own heart risk, and that is the finding with the clearest shape. A review that split 17 randomised trials in 66 thousand adults by risk concluded that for people at low cardiovascular risk, cutting or modifying saturated fat brings little or no benefit over five years. At high risk the same reductions crossed the thresholds that matter, and replacing saturated fat with polyunsaturated fat gave a relative risk of 0.75 for a nonfatal heart attack. The practical question is what you cook with. Over 33 years in 221 thousand American adults, the highest quarter of butter intake was associated with 15 percent higher total mortality than the lowest, and the highest quarter of plant oil intake with 16 percent lower. That is a cohort and not a trial, so it links the two without showing that butter caused anything. Canola, soybean and olive oil each pointed the same way per 5 g a day. Read the observational side with some patience. Studies of high-fat dairy disagree with each other, and one reason the reviewers name is that the cows were fed differently, pasture against grain. Last, a label detail that trips people up. A claim of reduced fat on butter is calculated against the 80 percent milkfat that United States law requires butter to contain, so it tells you nothing about the tub beside it on the shelf. What we cannot tell you is who should be careful with butter. We hold no card on butter in pregnancy, in milk allergy or lactose intolerance, or alongside any medicine, and none on an amount above which harm has been measured, so this page has no safety section at all.

Sources

1. but-r5-07 · systematic review of observational studies
   We discuss factors that may have contributed to the variability between studies, including differences in (1) the potential for residual confounding; (2) the types of high-fat dairy foods consumed; and (3) bovine feeding practices (pasture- vs. grain-based) known to influence the composition of dairy fat.
   Eur J Nutr, 2013 · checked 2026-09-22
2. but-r5-02 · systematic review and network meta-analysis of randomised trials
   Safflower, sunflower, rapeseed, flaxseed, corn, olive, soybean, palm, and coconut oil as well beef fat were more effective in reducing LDL-C (-0.42 to -0.23 mmol/l) as compared with butter.
   J Lipid Res, 2018 · checked 2026-09-22
3. but-r5-06 · randomised crossover trial
   Compared with the olive oil, the diet with ghee increased fasting plasma apo-B (apo B) (0·09, 95 % CI 0·02, 0·17 g/l, P = 0·018), non-HDL-cholesterol (non-HDL-cholesterol) (0·53, 95 % CI 0·01, 1·05 mmol/l, P = 0·046)
   Br J Nutr, 2022 · checked 2026-09-22
4. but-r5-05 · prospective cohort study
   After adjusting for potential confounders, the highest butter intake was associated with a 15% higher risk of total mortality compared to the lowest intake (hazard ratio [HR], 1.15; 95% CI, 1.08-1.22; P for trend < .001).
   JAMA Intern Med, 2025 · checked 2026-09-22
5. but-r5-08 · codified federal regulation
   In determining whether a claim is appropriate, the calculation of the percent fat reduction in milkfat shall be based on the 80 percent milkfat requirement provided by the statutory standard for butter (21 U.S.C. 321a);
   21 CFR 101.67, Use of nutrient content claims for butter · checked 2026-09-22
6. but-r5-01 · risk stratified systematic review of randomised trials
   For persons at low cardiovascular risk, reducing or modifying saturated fat intake has little or no benefit over a period of 5 years.
   Ann Intern Med, 2026 · checked 2026-09-22
7. but-r5-03 · GRADE assessment within a network meta-analysis
   Low quality of evidence judgments were mainly driven by the low number of trials, unclear risk of bias, imprecision, and inconsistency for several comparisons.
   J Lipid Res, 2018 · checked 2026-09-22
8. but-r5-04 · systematic review and meta-analysis of prospective cohorts
   Butter consumption was weakly associated with all-cause mortality (N = 9 country-specific cohorts; per 14g(1 tablespoon)/day: RR = 1.01, 95%CI = 1.00, 1.03, P = 0.045); was not significantly associated with any cardiovascular disease (N = 4; RR = 1.00, 95%CI = 0.98, 1.02; P = 0.704)
   PLoS One, 2016 · checked 2026-09-22