The one-year study
Researchers followed 262 adults with type 2 diabetes who chose a ketogenic diet delivered with continuous remote medical care, and 87 who chose usual care. After a year in the keto group, HbA1c had fallen from 7.6% to 6.3%, weight was down 13.8 kg, insulin had been reduced or stopped in 94% of the people using it, and sulfonylureas had been stopped entirely. The usual-care group did not change.
Now the limits. People picked their own group, so the two were not comparable to begin with. The company that sells the program funded the study. LDL cholesterol rose 10%. And the key ingredient was not in the food: physicians adjusted every participant's medication as their blood sugar fell. The paper's own first line says carbohydrate restriction necessitates prompt medication changes.
The pooled analysis
A BMJ analysis combined 23 randomized trials with 1,357 participants. At six months, 57% of people on low-carb diets had an HbA1c below 6.5%, against 31% on comparison diets. Weight, triglycerides and insulin sensitivity all improved by clinically important amounts.
By 12 months most of that advantage had gone, and there were signs of worse LDL cholesterol. The benefit was also smaller when remission was defined as being off medication, and much weaker in studies that included people on insulin.
Why it fades
The analysis points at adherence. Very-low-carb diets worked well among the people who stuck to them and poorly among those who did not. A diet you stop following stops working, which is not a deep finding, but it is the one that decides the result.
It is the reason these pages keep coming back to food you would choose to eat anyway.