Low-Carb vs Low-Fat, Finally

Thirty years of diet wars produced one genuinely useful answer, and it is boring: when adherence is equal, the diets work about equally, because adherence is the whole game. The head-to-head randomized trials — DIETFITS above all — are the scorecard this page reads slowly, so you can pick a diet without joining a tribe.

The Scorecard: What the Trials Actually Found

DIETFITS, Read Slowly

DIETFITS is the trial this debate had been waiting for: 609 overweight adults randomized to a healthy low-fat diet (around 20 g of fat to start) or a healthy low-carb diet (around 20 g of carbs to start), both coached toward whole foods, vegetables, and satiety — with no calorie targets, on purpose. At 12 months, low-carb lost 6.0 kg and low-fat 5.3 kg; the difference was not meaningful (JAMA, 2018). The trial also tested two popular claims directly: that DNA variants predict who thrives on which diet, and that insulin secretion does the same. Neither panned out — genotype-matched diets lost no more weight than mismatched ones, and baseline insulin secretion predicted nothing. The within-arm spread was the real story: some participants lost 30 kg, others gained, on both diets. When the same intervention produces that range, the question "which diet wins?" is asking about a few hundred grams of group mean while ignoring kilograms of individual variation.

One more detail matters for reading the null: the arms converged. By month twelve the low-carb group was eating around 130 g of carbohydrate a day and the low-fat group around 60 g of fat — both a long way from their starting extremes and, practically, not far from each other. When "low-carb" and "low-fat" both drift toward a whole-food middle, a null result is not evidence that diet composition is irrelevant; it is evidence that ordinary humans moderate extreme prescriptions into the same livable diet. The trials compare what people actually do with a label, not what the label promises.

Why Adherence Beats Identity

POUNDS LOST ran 811 people on four diets spanning 20% to 40% fat and 15% to 25% protein. At two years the arms had all lost about 3–4 kg — no meaningful differences — and the strongest predictor of success was attendance and self-reported adherence, not which diet had been assigned (NEJM, 2009). The pattern repeats across the trial literature: diets differ less in their biology than in their livability. Both low-carb and low-fat work in the studies mainly by simplifying choice — cutting a macro removes whole aisles of hyperpalatable food and, in practice, lowers energy intake. The environment design protocol explains why: the diet that survives is the one whose defaults you can live with, and that property belongs to the person, not the macro split.

Where Low-Carb Holds an Edge

Honest accounting gives low-carb a modest, consistent edge in the trials. DIRECT found the low-carb arm lost 4.7 kg versus 2.9 kg for low-fat at two years — the one meaningful difference in the big head-to-heads (NEJM, 2008). A randomized trial in obese adults found a very low-carb diet (under 40 g/day) beat low-fat for weight and several lipid markers over a year (Annals of Internal Medicine, 2014). And a meta-analysis of long-term trials found higher-fat, lower-carb approaches ahead of low-fat by about 1.15 kg at 12 months (Lancet Diabetes & Endocrinology, 2015). For insulin resistance and type 2 diabetes, carbohydrate restriction is a genuine clinical tool with trial support — the sugar topic owns that evidence. The edge is small, real, and partly mechanical: protein and fat are more satiating, so low-carb dieters often eat less without being told to.

Where Low-Fat Holds an Edge

The counter-evidence is also real. In a tightly controlled inpatient trial, adults offered an ad libitum plant-based low-fat diet ate roughly 700 kcal less per day than on an animal-based ketogenic diet, and lost more fat over two weeks (Nature Medicine, 2021) — a result that says something about the satiating power of bulky, low-calorie plants. Low-fat's broader problem is historical, not biological: the 1990s low-fat era replaced fat with refined carbohydrate, which the substitution evidence says is the one swap that buys nothing (see saturated fat, re-examined). Done well — legumes, whole grains, vegetables, lean protein — low-fat diets are simply a Mediterranean-adjacent pattern, and they perform like one. The diet wars made "low-fat" mean rice cakes and fat-free cookies; the trials that actually fed people real food tell a different story.

Choosing Without a War

• 🎯 Start with the one you can sustain. The trials say adherence is the deciding variable, so preference is a legitimate starting datum, not a cop-out.
• 🥩 Keep the protein floor either way. Protein adequacy holds whichever macro you cut — see the protein topic.
• 🫒 Watch the quality override. Low-carb with olive oil, nuts, and fish, or low-fat with legumes and whole grains — never the refined-carb or butter-and-bacon versions.
• 🧪 Run a three-month trial. Same protein, same activity, one macro cut; check weight, energy, and lipids before and after.
• 🫙 The middle path is legitimate. The Mediterranean pattern is what the trials keep pointing at when both extremes underperform.

Questions, Answered Briefly

• 🤷 Which should I pick? The one you can keep. If you love the pattern, adherence follows, and adherence is what the trials say predicts the scale.
• 🩸 Does low-carb wreck cholesterol? Not automatically. LDL response varies by person, and saturated-fat quality matters more than carb grams; a lipid panel before and after answers it for you.
• 🍬 What about diabetes specifically? That is where low-carb's trial support is strongest — but medication adjustments around carbohydrate restriction belong to a clinician, not a website. The glucose topic covers the physiology.
• 🍞 Isn't keto special for hunger? Some trials and the inpatient data suggest high-fat ketogenic eating is less satiating per calorie than bulky low-fat food — the opposite of the folklore. Individual experience varies more than either claim.
• 🧬 Should I pay for the DNA diet tests? No. The genotype-matching hypothesis was tested directly in DIETFITS and failed; the commercial tests were selling the hypothesis before the data existed.
• 📉 What actually predicts who keeps it off? Self-monitoring, support, and environment — the variables the habit-formation protocol engineers. Diet identity does not.

Related Topics