Four common schedules

Intermittent fasting is a family of schedules, not one practice. This chapter compares four of them, all with the same amount of food each week, so that only the timing differs.

The last schedule is sometimes called a “rolling 48”; here it means one 48-hour fast each week, from dinner to dinner, and normal eating on the other days.

A small model of glycogen and ketones

The model follows three quantities through time: carbohydrate waiting to be absorbed from the gut, the liver’s glycogen $G$ (in grams), and blood β-hydroxybutyrate $K$ (in mM). A fourth variable, $a$, between 0 and 1, stands for the slow adaptations of fasting, mainly the release of fatty acids from fat and the liver’s capacity to make ketones, which build up over a day or more.

$$\frac{dG}{dt} = 0.3\,A – 5\,\frac{G}{G + 25}\,(1 – I)\,(1 – 0.6\,a), \qquad I = \frac{A}{A + 10},$$

$$\frac{da}{dt} = \frac{(1 – I)\,s(G) – a}{18 – 14\,I}, \qquad s(G) = \frac{1}{1 + (G/50)^4},$$

$$\frac{dK}{dt} = 2\,a\,(1 – I) + 0.03 – (0.9 + 2\,I)\,K .$$

Here $A$ is the rate at which carbohydrate is absorbed (grams per hour) and $I$ is a stand-in for insulin, near 1 just after a meal and near 0 when fasting. About 30 per cent of absorbed carbohydrate refills the liver, up to a maximum of 100 g. Between meals, the liver breaks glycogen down to hold blood glucose; as the fast lengthens, gluconeogenesis takes over, which the model represents by the factor $1 – 0.6\,a$. Ketone production switches on as glycogen falls ($s(G)$) and grows as the adaptations build up ($a$); insulin suppresses production and speeds clearance.

# A toy model of liver glycogen G (g) and blood beta-hydroxybutyrate K (mM), in 3-minute steps.
SCHEDULES = {   # meals as (clock hour, grams of carbohydrate); 250 g a day
    "12:12":      lambda day: [(8, 85), (13, 85), (19, 80)],
    "16:8":       lambda day: [(12, 125), (19, 125)],
    "OMAD":       lambda day: [(18, 250)],
    "48h weekly": lambda day: ([] if day % 7 == 1 else [(19, 250)] if day % 7 == 2
                               else [(8, 85), (13, 85), (19, 80)]),
}

def simulate(meals, days=14, dt=0.05):
    G, gut, K, a = 70.0, 0.0, 0.1, 0.0
    trace = []
    for i in range(int(days * 24 / dt)):
        t = i * dt
        for hour, grams in meals(int(t // 24)):
            if abs(t % 24 - hour) < dt / 2:
                gut += grams
        absorb = 0.7 * gut                       # carbohydrate absorbed (g/h)
        insulin = absorb / (absorb + 10.0)       # 0 = fasting, near 1 = just after a meal
        # glycogen broken down to hold blood glucose; gluconeogenesis takes over as a grows
        use = 5.0 * G / (G + 25.0) * (1 - insulin) * (1 - 0.6 * a)
        G = min(100.0, G + dt * (0.3 * absorb - use))    # 30% of absorbed carbohydrate
        gut -= dt * absorb                               # refills the liver
        switch = 1 / (1 + (G / 50.0) ** 4)       # ketogenesis turns on as glycogen runs low
        # slow build-up of fat release and ketogenesis, reset faster by food
        a += dt * ((1 - insulin) * switch - a) / (18.0 - 14.0 * insulin)
        K += dt * (2.0 * a * (1 - insulin) + 0.03 - (0.9 + 2.0 * insulin) * K)
        trace.append((t, G, K))
    return trace[len(trace) // 2:]               # report the second week only

tr = simulate(SCHEDULES["48h weekly"])
start = 7 * 24 + 19                              # last meal before the fast: 19:00, day 7
print("hours since last meal   liver glycogen (g)   BHB (mM)")
for t, G, K in tr:
    h = round(t - start, 2)
    if h in (0, 12, 16, 24, 36, 47.95):          # 47.95 = just before the meal that ends it
        print(f"{h:20.0f}   {G:18.0f}   {K:8.2f}")
print()
print("schedule    hours/week BHB > 0.5 mM   peak BHB (mM)   lowest liver glycogen (g)")
for name, meals in SCHEDULES.items():
    tr = simulate(meals)
    hours = sum(1 for _, _, K in tr if K > 0.5) * 0.05
    peak, low = max(K for _, _, K in tr), min(G for _, G, _ in tr)
    print(f"{name:10s}  {hours:24.0f}   {peak:13.2f}   {low:25.0f}")
# hours since last meal   liver glycogen (g)   BHB (mM)
#                    0                   95       0.09
#                   12                   72       0.17
#                   16                   58       0.25
#                   24                   35       0.61
#                   36                   14       1.34
#                   48                    5       1.78
#
# schedule    hours/week BHB > 0.5 mM   peak BHB (mM)   lowest liver glycogen (g)
# 12:12                              0            0.18                          68
# 16:8                               0            0.28                          57
# OMAD                               9            0.56                          41
# 48h weekly                        33            1.78                           5

Reading the output

Four results stand out.

The first day is glycogen and gluconeogenesis, not ketones. In the 48-hour timeline, β-hydroxybutyrate is still below 0.3 mM sixteen hours after the last meal and only passes 0.5 mM around the end of the first day. Most of the rise comes in the second day, after glycogen has fallen below about half its starting level.

16:8 barely touches ketosis. In the model, the 16:8 schedule raises the daily peak of β-hydroxybutyrate from about 0.2 to about 0.3 mM. By Chapter 3’s model, that changes the brain’s ketone share from about 6 to about 8 per cent. The schedule is a change in meal timing and in the length of the overnight fast; it is not a ketogenic diet.

One meal a day reaches the edge. Eating once a day brings the model into the lower edge of ketosis for a short time every afternoon, and empties the liver’s glycogen to about 40 per cent of capacity each day.

A 48-hour fast is a different kind of event. Once a week it takes glycogen nearly to empty and ketones to almost 2 mM, the range in which the signalling effects of Chapter 3 were seen in the laboratory. The model also shows that one meal after a long fast does not undo the adaptations at once: ketones rise again overnight. People with diabetes, and anyone taking medicines that lower blood sugar or SGLT2 inhibitors, face real danger at this depth (see the preface).

What the trials say

Model output is not evidence about health. For that, randomised trials are needed, and the best of them compare schedules with the same amount of food, so that any effect of timing is not confused with eating less.

The pattern across trials is consistent with the model. When the total amount of food is held equal, a modest change of schedule produces modest effects, and its advantages over ordinary calorie restriction are small or absent. Effects that do appear in humans seem to depend at least as much on when in the day people eat (early windows, aligned with the body clock) as on how long they fast.

Attention

None of the large trials above measured attention as a main outcome. The reviews of Chapter 3 found human cognitive data scarce and mixed. That gap is the reason for Chapter 9’s trial.