5D Library › Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization · 中文

Chapter 3. Same Food, Two Patterns

What the trials already settle

The claim to be tested is narrow and should be stated narrowly: holding energy and macronutrients constant, does compressing the day’s food into a short window remove more visceral fat than spreading it out? Two trials bear on it directly. Trepanowski and colleagues randomised metabolically healthy obese adults to alternate-day fasting or to daily restriction at matched energy and found no advantage for the fasting arm in weight or in the cardiometabolic measures, with higher dropout in the fasting group (Trepanowski et al. 2017). Lowe and colleagues randomised adults with overweight to sixteen-hour time-restricted eating or to three structured meals and found a small weight difference that was not accompanied by an advantage in fat mass, with a loss of lean mass in the restricted arm (Lowe et al. 2020).

Evidence

Published. Trepanowski and colleagues found alternate-day fasting no better than matched daily restriction over a year, with greater dropout (Trepanowski et al. 2017). Lowe and colleagues found sixteen-hour time-restricted eating produced no advantage in fat mass against three meals a day, and a decrease in lean mass (Lowe et al. 2020). Hall and colleagues set out why a model of energy balance, and not a rule of thumb, is needed to turn a deficit into a predicted change in weight (Hall et al. 2011).

So the trials do not support a large effect. The interesting question is then whether the mechanism is absent or merely small, and that is a question arithmetic can answer. p05 gives both patterns the same energy and the same carbohydrate and differs only in when it arrives.

Calculation

# Chapter 3. p05: Two ways to eat the same food. Each meal raises insulin
# along a curve that peaks about fifty minutes after it and decays; the
# height of the peak is taken to be proportional to the meal's
# carbohydrate. Both patterns below carry the same energy and the same
# carbohydrate: one eats it in two meals inside an eight-hour window,
# the other spreads it over six meals across fourteen hours. The depot
# release curve is the one p04 fits. The grid is one minute wide.
BASE, KA, TP = 35.0, 3.12, 0.80        # pmol/L, pmol/L per g, hours
CARB_DAY = 202.0                       # g of carbohydrate a day
VMAX = {"visceral": 2.30, "subcut": 1.00}
IC50 = {"visceral": 110.0, "subcut": 55.0}
MASS = {"visceral": 2.2, "subcut": 12.0}
HILL, PER_UNIT = 1.6, 0.84
LOWINS = 50.0                          # pmol/L, below which release is free
E = 2.718281828459045

WINDOW = (12.0, 19.0)                  # two meals, eight-hour window
GRAZE = (7.0, 10.0, 13.0, 16.0, 19.0, 21.0)

def insulin(t, meals, carb_each):
    v = BASE
    for tm in meals:
        d = t - tm
        if d > 0.0:
            v += KA * carb_each * (d / TP) * E ** (1.0 - d / TP)
    return v

def release(depot, ins):
    return MASS[depot] * VMAX[depot] / (1.0 + (ins / IC50[depot]) ** HILL)

def day(meals):
    carb_each = CARB_DAY / len(meals)
    n = 24 * 60
    vis = sub = low = tot_ins = 0.0
    peak = 0.0
    for i in range(n):
        t = 24.0 * i / n
        ins = insulin(t, meals, carb_each)
        peak = max(peak, ins)
        tot_ins += ins
        if ins < LOWINS:
            low += 1.0
        vis += release("visceral", ins)
        sub += release("subcut", ins)
    h = 24.0 / n
    return {"peak": peak, "mean": tot_ins / n, "low_h": low * h,
            "vis": vis * h * PER_UNIT, "sub": sub * h * PER_UNIT}

if __name__ == "__main__":
    a, b = day(WINDOW), day(GRAZE)
    print("   both patterns: {:.0f} g of carbohydrate a day"
          .format(CARB_DAY))
    print("   pattern        meals   peak   mean   hours below {:.0f}"
          .format(LOWINS))
    for name, meals, d in (("eight-hour", WINDOW, a),
                           ("spread out", GRAZE, b)):
        print("   {:<15}{:<8d}{:<7.0f}{:<7.1f}{:.2f}"
              .format(name, len(meals), d["peak"], d["mean"], d["low_h"]))
    print("   fatty acid released a day, grams:")
    print("   pattern        visceral   subcutaneous   together   share")
    for name, d in (("eight-hour", a), ("spread out", b)):
        print("   {:<15}{:<11.1f}{:<15.1f}{:<11.1f}{:.3f}"
              .format(name, d["vis"], d["sub"], d["vis"] + d["sub"],
                      d["vis"] / (d["vis"] + d["sub"])))
    print("   visceral release, eight-hour minus spread out: {:+.1f} g a day"
          .format(a["vis"] - b["vis"]))
    print("   as a share of the spread-out figure: {:+.3f}"
          .format(a["vis"] / b["vis"] - 1.0))
    print("   extra hours below {:.0f} pmol/L: {:+.2f}"
          .format(LOWINS, a["low_h"] - b["low_h"]))
    print("   mean insulin, eight-hour minus spread out: {:+.1f} pmol/L"
          .format(a["mean"] - b["mean"]))
#    both patterns: 202 g of carbohydrate a day
#    pattern        meals   peak   mean   hours below 50
#    eight-hour     2       351    91.7   14.73
#    spread out     6       174    90.9   7.05
#    fatty acid released a day, grams:
#    pattern        visceral   subcutaneous   together   share
#    eight-hour     69.4       118.3          187.8      0.370
#    spread out     61.9       92.0           153.9      0.402
#    visceral release, eight-hour minus spread out: +7.6 g a day
#    as a share of the spread-out figure: +0.122
#    extra hours below 50 pmol/L: +7.68
#    mean insulin, eight-hour minus spread out: +0.8 pmol/L

The two patterns carry 202 grams of carbohydrate a day. The compressed day peaks at an insulin of 351 and the spread-out day at 174, but their twenty-four hour means are 91.7 and 90.9, a difference of 0.8 pmol/L. By the measure that matters for the lipase, however, they are not close at all: the compressed day spends 14.73 hours below the threshold against 7.05, which is 7.68 extra hours. Visceral release comes out at 69.4 grams against 61.9, a difference of 7.6 grams a day, or 0.122 as a share. The mechanism is real, and it is of this size.

Insulin through the day under the two patterns, with the threshold marked (p05).

Notice what the same table says about the visceral share. It is 0.370 in the compressed day and 0.402 in the spread-out day: the compressed day releases more fat in total but a slightly smaller fraction of it from the abdomen, because when insulin is low the larger subcutaneous mass is free to release too. Chapter 2 predicted exactly this.

From a day to a month, and a disagreement

A day’s release is not a month’s loss. Hall and colleagues make the general point that turning an energy imbalance into a change in body composition requires a model, because the body’s own adjustments are part of the answer (Hall et al. 2011). p06 takes the narrow version of that: the deficit sets how much fat is lost, and the only thing the eating pattern is allowed to change is which depot it comes from. Two defensible ways of apportioning that are computed side by side.

Calculation

# Chapter 3. p06: From a day to twenty-eight days. Release is not loss:
# fatty acids that are not burnt go back into the depot. What a deficit
# fixes is the total amount of fat lost; what the eating pattern changes
# is which depot it comes out of, through the share of the day's release
# that the visceral depot contributes. The same insulin curve as p05 is
# used, with the day's carbohydrate split over one to six meals spread
# evenly through the eating window. Adipose tissue is taken at 7.7 kcal
# a gram, and the deficit at 500 kcal a day for twenty-eight days.
BASE, KA, TP = 35.0, 3.12, 0.80
CARB_DAY = 202.0
VMAX = {"visceral": 2.30, "subcut": 1.00}
IC50 = {"visceral": 110.0, "subcut": 55.0}
MASS = {"visceral": 2.2, "subcut": 12.0}
HILL, LOWINS, E = 1.6, 50.0, 2.718281828459045
DEFICIT, KCAL_PER_G, DAYS = 500.0, 7.7, 28
FOXMAX, IC_OX, PER_UNIT = 0.62, 70.0, 0.84

def patterns(n):
    # n meals, evenly spaced inside a window that widens with n
    span = {1: 0.0, 2: 7.0, 3: 9.0, 4: 11.0, 5: 12.5, 6: 14.0}[n]
    start = 19.0 - span
    if n == 1:
        return (19.0,)
    return tuple(start + span * i / (n - 1) for i in range(n))

def insulin(t, meals, carb_each):
    v = BASE
    for tm in meals:
        d = t - tm
        if d > 0.0:
            v += KA * carb_each * (d / TP) * E ** (1.0 - d / TP)
    return v

def release(depot, ins):
    return MASS[depot] * VMAX[depot] / (1.0 + (ins / IC50[depot]) ** HILL)

def portal_oxidised(ins):
    # share of the visceral depot's output the liver burns on first pass;
    # insulin suppresses hepatic fat oxidation, so this rises when fasted
    return FOXMAX / (1.0 + (ins / IC_OX) ** HILL)

def day(n):
    meals = patterns(n)
    carb_each = CARB_DAY / n
    steps = 24 * 60
    vis = sub = low = burnt = 0.0
    for i in range(steps):
        ins = insulin(24.0 * i / steps, meals, carb_each)
        if ins < LOWINS:
            low += 1.0
        v = release("visceral", ins)
        vis += v
        sub += release("subcut", ins)
        burnt += v * portal_oxidised(ins)
    share = vis / (vis + sub)
    total_loss = DEFICIT / KCAL_PER_G
    h = 24.0 / steps
    return {"low_h": low * h, "share": share,
            "vis_day": total_loss * share,
            "vis_28": total_loss * share * DAYS,
            "burnt_day": burnt * h * PER_UNIT,
            "burnt_28": burnt * h * PER_UNIT * DAYS}

if __name__ == "__main__":
    print("   deficit {:.0f} kcal a day, {:.0f} days, {:.1f} kcal a gram"
          .format(DEFICIT, DAYS, KCAL_PER_G))
    print("   total fat lost either way: {:.0f} g in {:.0f} days"
          .format(DEFICIT / KCAL_PER_G * DAYS, DAYS))
    print("   meals   hours low   visceral share   split g   burnt g")
    out = {}
    for n in range(1, 7):
        d = day(n)
        out[n] = d
        print("   {:<8d}{:<12.2f}{:<17.3f}{:<10.1f}{:.1f}"
              .format(n, d["low_h"], d["share"], d["vis_28"],
                      d["burnt_28"]))
    best, worst = out[1], out[6]
    print("   one meal against six, visceral fat in {:.0f} days:"
          " {:.1f} g against {:.1f} g".format(DAYS, best["vis_28"],
                                              worst["vis_28"]))
    print("   difference {:.1f} g, a ratio of {:.3f}"
          .format(best["vis_28"] - worst["vis_28"],
                  best["vis_28"] / worst["vis_28"]))
    two = out[2]
    print("   two meals against six: {:.1f} g against {:.1f} g, ratio {:.3f}"
          .format(two["vis_28"], worst["vis_28"],
                  two["vis_28"] / worst["vis_28"]))
    print("   the deficit sets the total; the pattern only moves"
          " {:.1f} g of it".format(best["vis_28"] - worst["vis_28"]))
    print("   visceral fat burnt on first pass in the liver, {:.0f} days:"
          .format(DAYS))
    print("   one meal {:.0f} g, two meals {:.0f} g, six meals {:.0f} g"
          .format(best["burnt_28"], two["burnt_28"], worst["burnt_28"]))
    print("   two meals against six by that route: ratio {:.3f}"
          .format(two["burnt_28"] / worst["burnt_28"]))
    print("   the two accountings disagree in direction, so the honest")
    print("   answer is that the pattern matters little either way")
#    deficit 500 kcal a day, 28 days, 7.7 kcal a gram
#    total fat lost either way: 1818 g in 28 days
#    meals   hours low   visceral share   split g   burnt g
#    1       19.02       0.357            649.7     946.4
#    2       14.73       0.370            672.4     794.2
#    3       11.28       0.381            693.4     693.1
#    4       9.05        0.391            711.3     624.1
#    5       7.77        0.399            725.6     576.7
#    6       6.43        0.406            739.0     538.2
#    one meal against six, visceral fat in 28 days: 649.7 g against 739.0 g
#    difference -89.4 g, a ratio of 0.879
#    two meals against six: 672.4 g against 739.0 g, ratio 0.910
#    the deficit sets the total; the pattern only moves -89.4 g of it
#    visceral fat burnt on first pass in the liver, 28 days:
#    one meal 946 g, two meals 794 g, six meals 538 g
#    two meals against six by that route: ratio 1.476
#    the two accountings disagree in direction, so the honest
#    answer is that the pattern matters little either way

At a deficit of 500 kcal a day for 28 days and 7.7 kcal a gram of adipose tissue, 1818 grams are lost either way. Apportioned by the share of release, one meal a day yields 649.7 grams of visceral fat against 739.0 for six meals, a difference of 89.4 grams and a ratio of 0.879; two meals against six gives 672.4 against 739.0, a ratio of 0.910. That route favours the many-meal day. Apportioned instead by what the liver burns on first pass from the portal vein, one meal yields 946 grams, two meals 794 and six meals 538, and two meals against six is a ratio of 1.476 the other way.

Both accountings are defensible and they disagree in direction. That disagreement is the finding, and it is consistent with the trials: when two reasonable models of the same mechanism put the effect on opposite sides of zero, and the effect in either case is under a fifth of the total, the honest conclusion is that the pattern matters little and the deficit matters almost entirely. Everything the marketing attributes to the window, the deficit was already doing.

Hypothesis · not established

The author’s interpretation (not yet tested): the eating window changes where fat is lost from by less than a fifth either way at matched energy, and its real clinical value is adherence, not partitioning: a short window lowers intake in people who do not count, and does nothing extra in people who do. What would show it wrong: in a trial with energy intake measured rather than prescribed, if a short window produced a larger fall in visceral fat by imaging than matched continuous restriction, with intake equal in both arms, the partitioning effect would be larger than this arithmetic allows.

Summary

At the same energy and carbohydrate the compressed day and the spread-out day differ by 0.8 pmol/L in mean insulin, yet the compressed day spends 7.68 more hours below the lipase threshold and releases 0.122 more visceral fatty acid. Over 28 days at a fixed deficit the two routes for apportioning the loss disagree in sign, 0.910 one way and 1.476 the other, which agrees with trials that found no advantage. Chapter 4 turns to measurement: how well the two cheapest markers separate the people this arithmetic is about.

References

Hall, Kevin D., Gary Sacks, Dhruva Chandramohan, et al. 2011. “Quantification of the Effect of Energy Imbalance on Bodyweight.” The Lancet 378 (9793): 826 to 837.
Lowe, Dylan A., Nancy Wu, Linnea Rohdin-Bibby, et al. 2020. “Effects of Time-Restricted Eating on Weight Loss and Other Metabolic Parameters in Women and Men with Overweight and Obesity.” JAMA Internal Medicine 180 (11): 1491 to 1499.
Trepanowski, John F., Cynthia M. Kroeger, Adrienne Barnosky, et al. 2017. “Effect of Alternate-Day Fasting on Weight Loss, Weight Maintenance, and Cardioprotection among Metabolically Healthy Obese Adults.” JAMA Internal Medicine 177 (7): 930 to 938.

Fan, Y. (2026). Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization: Hepatic Glycogen Depletion Times, Triglyceride-Glucose Indices and Micronutrient-Constrained Menus [胰岛素胰高血糖素双稳态、内脏脂解动力学与线性规划配餐] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090248 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).