Where the evidence is strongest: drug-resistant epilepsy

The ketogenic diet has been used to treat epilepsy for a century, and this is its best-supported use.

This use is a medical treatment carried out under close supervision by hospital dietitians and neurologists, with dedicated protocols for the composition of the diet, how it is started and how it is monitored; it is not the same thing as a “ketogenic diet” tried at home.

Type 2 diabetes

For people with diabetes on glucose-lowering drugs, any change of diet must be made together with a doctor adjusting the medication; cutting carbohydrate sharply without adjusting insulin or sulfonylureas can cause hypoglycaemia, and people on SGLT2 inhibitors risk ketoacidosis (see Chapter 1).

Weight: the insulin model and the energy-balance model

Two opposing frameworks explain obesity. The carbohydrate–insulin model holds that a high-carbohydrate diet raises insulin, which directs energy into storage in adipose tissue and so increases hunger and intake; cutting carbohydrate and insulin can promote fat loss even at the same calories (Ludwig and Ebbeling 2018). The energy-balance model holds that changes in body fat depend mainly on the difference between energy intake and expenditure, and that diet composition acts mainly through its effect on intake.

The two models can be compared directly in metabolic-ward studies with calories and protein strictly matched.

# Putting two well-controlled trials in proportion.
# (1) Hall et al. (2016): 17 men in metabolic chambers, a high-carbohydrate baseline diet, then an
#     isocaloric ketogenic diet. Daily energy expenditure rose by about 57 kcal/day in the
#     respiratory chamber (doubly labelled water gave a larger, less certain estimate, ~151 kcal/day).
#     What would a sustained 57 kcal/day mean for body fat? Two ways to compute it:
#     static: 7,700 kcal per kg of fat, no adaptation (overstates long-run change);
#     dynamic: Hall et al. (2011): eventual weight change ~ 1 kg per ~22-24 kcal/day (100 kJ/day per kg,
#     or 10 kcal/day per lb) of sustained
#     imbalance, approached with a time constant of roughly a year (half after ~1 year).
import math
for dE, label in ((57, "chamber estimate"), (151, "doubly labelled water estimate")):
    static_1y = dE * 365 / 7700
    final = dE / 22
    tau_years = 1 / math.log(2)                     # half the eventual change after one year
    dyn_1y = final * (1 - math.exp(-1 / tau_years))
    print(f"{dE:3d} kcal/day ({label}): static rule {static_1y:.1f} kg in a year;"
          f" dynamic {dyn_1y:.1f} kg in a year, {final:.1f} kg eventually")
# (2) Liu et al. (2022, NEJM): 139 adults with obesity, 12 months, calorie restriction with
#     time-restricted eating (08:00-16:00) versus the same calorie restriction alone.
#     Reported difference in weight change: -1.8 kg (95% CI -4.0 to 0.4).
d, lo, hi = -1.8, -4.0, 0.4
se = (hi - lo) / (2 * 1.96)
z = d / se
p = 2 * (1 - 0.5 * (1 + math.erf(abs(z) / math.sqrt(2))))
print(f"Liu 2022: difference {d} kg, SE {se:.2f}, z = {z:.2f}, two-sided p ~ {p:.2f}:"
      " no clear added effect of the eating window once calories are matched")
#  57 kcal/day (chamber estimate): static rule 2.7 kg in a year; dynamic 1.3 kg in a year, 2.6 kg eventually
# 151 kcal/day (doubly labelled water estimate): static rule 7.2 kg in a year; dynamic 3.4 kg in a year, 6.9 kg eventually
# Liu 2022: difference -1.8 kg, SE 1.12, z = -1.60, two-sided p ~ 0.11: no clear added effect of the eating window once calories are matched

The first part shows that a difference of 57 kcal a day, even if sustained, corresponds by a dynamic model allowing for metabolic adaptation to only about 1.3 kg in a year, and that only on the assumption that the difference persists (Hall et al. 2011). This is not negligible, but far smaller than the popular claim of “much more fat loss on the same calories”. The second part is the 12-month randomised trial of Liu and colleagues in Guangzhou: 139 adults with obesity all restricted calories, one group confining eating to 8:00–16:00 and the other eating at any time. The difference in weight between groups was −1.8 kg, with a 95 % confidence interval from −4.0 to 0.4, crossing zero (Liu et al. 2022). In other words, with calories matched, the eating window itself showed no clear additional weight loss.

A risk not to ignore: LDL cholesterol

LDL-C and apolipoprotein B are among the causal risk factors for atherosclerotic cardiovascular disease. Anyone planning to follow a low-carbohydrate or ketogenic diet for the long term should check blood lipids, including LDL-C and apolipoprotein B, with a doctor before starting and some weeks after, rather than watching only weight or blood ketones.

Summary

A ketogenic diet for drug-resistant childhood epilepsy is supported by reliable randomised evidence; in type 2 diabetes it brings short-term benefits that fade over time, and medication must be adjusted under a doctor’s guidance. In studies with strictly matched calories and protein, a ketogenic diet did not produce the extra fat loss the insulin model predicts; with calories matched, time-restricted eating showed no clear additional weight loss, though early time-restricted eating may improve some metabolic markers without any change in weight. A ketogenic diet can raise LDL-C markedly in some people. The next chapter takes up a more everyday question: why do many people feel tired, palpitations or cramps in the first days of a low-carbohydrate diet?