“Carbs are bad for you” treats an entire macronutrient category — everything from lentils to lollies — as one thing. The large cohort evidence says the category doesn’t hold together as a single claim; quality and quantity both matter, and they cut in different directions.

The PURE study, a large observational cohort published in The Lancet in 2017 ^[1], followed over 135,000 people across 18 countries and found that very high carbohydrate intake (in the highest quintile of intake, which in this cohort skewed toward refined carbohydrate sources) was associated with higher mortality risk, while higher total fat intake was associated with lower mortality risk in the same analysis. This is the study most often cited by low-carb advocates — and it’s worth being precise about its limits: it’s observational, not a randomized trial, so it shows association, not proof of causation, and the authors themselves have cautioned against reading it as “eat less carbohydrate, more saturated fat” advice.

A more targeted 2019 series of systematic reviews and meta-analyses in The Lancet ^[2] looked specifically at carbohydrate quality rather than treating carbohydrate as one bucket. The consistent finding: higher fibre intake and whole-grain, lower-glycaemic-index carbohydrate sources associate with meaningfully lower risk of cardiovascular disease, type 2 diabetes, and colorectal cancer, compared with refined, low-fibre carbohydrate sources.

Put the two studies together and the honest answer isn’t “carbs are good” or “carbs are bad” — it’s “which carbs, and how much.”

This is exactly the kind of claim the It Depends verdict exists for: the “depends on” variable here is concrete and specific — carbohydrate quality (fibre content, degree of processing, glycaemic index) and total quantity relative to overall diet, not carbohydrate as an undifferentiated category. A diet built around vegetables, legumes, and whole grains and a diet built around white bread and soft drink are both “high carb,” and the outcome evidence treats them very differently.