5D Library › 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 · English
第五章 这套饮食拿走的那种矿物质
为什么低碳水的一周要花掉盐
低碳水饮食第一周出现的那些抱怨,头痛、起身时眼前发白、夜里小腿抽筋,既不是排毒,也不是必须忍过去的适应。它们是钠流失可以预见的后果,而机制正是整套方案所围绕的那一次胰岛素下降。Boulter、Hoffman 与 Arky 在断食受试者身上直接测过这个形态,发现头几天有一次迅速的排钠,此后排泄量稳定下来 (Boulter et al. 1973)。Cahill 的综述把同一个观察放进断食适应的次序里 (Cahill 2006)。在肾小管,胰岛素是一种保钠的激素;它一下降,钠就走,水跟着它走,而肾脏对钾与镁的处理也随之改变。
证据
已发表。 Boulter、Hoffman 与 Arky 测量了饥饿过程中的钠排泄,描述了早期的排钠及其后来的稳定 (Boulter et al. 1973)。Cahill 的综述把肾脏对钠的处理放进断食适应的更大次序之中 (Cahill 2006)。
所以一套刻意压低胰岛素的方案造成了一项它很少承认的义务:它必须把矿物质放回去。这就变成一个最优化问题,因为这项义务要与能量预算、食物成本,以及任何人实际愿意吃的东西竞争。
推导
# Chapter 5. p09: Meeting the electrolyte floors at the lowest cost. A
# low-carbohydrate week lowers insulin, and low insulin makes the kidney
# excrete sodium, which drags potassium and magnesium with it, so the
# floors below are the ones a careful plan holds. Nine items, nought to
# three portions of each, every one of the 262,144 combinations priced
# by exhaustive search. Composition figures are rounded from published
# food tables; the prices are illustrative. The whole-food-only run
# repeats the search with the three added salts and capsules removed.
ITEMS = (
# name, sodium, potassium, magnesium (mg), kcal, cost
("avocado 150 g", 11, 727, 44, 240, 1.20),
("spinach 180 g", 126, 839, 157, 41, 0.90),
("almonds 28 g", 0, 208, 76, 164, 0.45),
("pumpkin seeds 28 g", 5, 223, 156, 158, 0.50),
("salmon 150 g", 75, 534, 44, 310, 3.50),
("bone broth 250 ml", 860, 110, 10, 40, 1.10),
("table salt 2.5 g", 970, 0, 0, 0, 0.01),
("lite salt 2 g", 0, 1050, 0, 0, 0.05),
("magnesium capsule", 0, 0, 100, 0, 0.12),
)
WHOLE = 6 # the first six items are whole foods
FLOOR = (3000, 3500, 400) # sodium, potassium, magnesium, mg a day
KCAL_CAP = 900 # kcal these items may take of the day
MAXP = 3
def search(items):
n = len(items)
best = None
counts = [0] * n
while True:
na = k = mg = kcal = 0
cost = 0.0
for i, c in enumerate(counts):
if c:
_, a, b, m, e, p = items[i]
na += a * c
k += b * c
mg += m * c
kcal += e * c
cost += p * c
if (na >= FLOOR[0] and k >= FLOOR[1] and mg >= FLOOR[2]
and kcal <= KCAL_CAP):
if best is None or cost < best[0] - 1e-12:
best = (cost, tuple(counts), na, k, mg, kcal)
i = 0
while i < n:
counts[i] += 1
if counts[i] <= MAXP:
break
counts[i] = 0
i += 1
if i == n:
break
return best
if __name__ == "__main__":
print(" floors: sodium {} mg, potassium {} mg, magnesium {} mg"
.format(*FLOOR))
print(" energy these items may take: {} kcal".format(KCAL_CAP))
for label, items in (("all nine items", ITEMS),
("whole foods only", ITEMS[:WHOLE])):
best = search(items)
if best is None:
print(" {}: no combination meets the floors".format(label))
continue
cost, counts, na, k, mg, kcal = best
print(" {}: cheapest plan costs {:.2f}".format(label, cost))
for i, c in enumerate(counts):
if c:
print(" {} x {}".format(c, items[i][0]))
print(" sodium {} mg, potassium {} mg, magnesium {} mg,"
" {} kcal".format(na, k, mg, kcal))
print(" slack: sodium {} mg, potassium {} mg,"
" magnesium {} mg".format(na - FLOOR[0], k - FLOOR[1],
mg - FLOOR[2]))
a = search(ITEMS)
b = search(ITEMS[:WHOLE])
print(" whole foods cost {:.2f} more and bring {} kcal more"
.format(b[0] - a[0], b[5] - a[5]))
print(" ratio of the two costs: {:.2f}".format(b[0] / a[0]))
print(" the cheap plan spends {:.3f} of its cost on the three salts"
" and capsules".format(sum(a[1][i] * ITEMS[i][5]
for i in range(WHOLE, len(ITEMS)))
/ a[0]))
# floors: sodium 3000 mg, potassium 3500 mg, magnesium 400 mg
# energy these items may take: 900 kcal
# all nine items: cheapest plan costs 1.44
# 1 x spinach 180 g
# 3 x table salt 2.5 g
# 3 x lite salt 2 g
# 3 x magnesium capsule
# sodium 3036 mg, potassium 3989 mg, magnesium 457 mg, 41 kcal
# slack: sodium 36 mg, potassium 489 mg, magnesium 57 mg
# whole foods only: cheapest plan costs 9.95
# 3 x spinach 180 g
# 1 x almonds 28 g
# 1 x salmon 150 g
# 3 x bone broth 250 ml
# sodium 3033 mg, potassium 3589 mg, magnesium 621 mg, 717 kcal
# slack: sodium 33 mg, potassium 89 mg, magnesium 221 mg
# whole foods cost 8.51 more and bring 676 kcal more
# ratio of the two costs: 6.91
# the cheap plan spends 0.375 of its cost on the three salts and capsules在允许用那三种盐与胶囊时,清过全部三条下界的最便宜一天花 1.44,占 41 kcal:一份菠菜,加三份食盐、三份低钠盐与三粒镁胶囊,交付 3036、3989 与 457 毫克,对应的下界是 3000、3500 与 400。限制在全食物内时,最便宜的一天花 9.95,占 717 kcal,成本比 6.91,并且多出 676 kcal。对全部 262144 种组合的穷举搜索,才使这些数字成为论断而不是猜测:在给定的清单与上限之内,不存在更便宜的组合。
那 676 kcal 才是要紧的部分,因为第六章的方案总共只有一千八百上下可以花。把其中 717 花在矿物质上,留给蛋白质的就不多了。这就是一条否则听起来像补剂营销的建议背后的算术:在一套刻意低胰岛素的饮食上,那些盐不是加价项,而是让能量预算根本还能解得出来的东西。
哪些食物带矿物质而不带能量
同一张清单换一种排法,就能看出压力从哪里来。
推导
# Chapter 5. p10: Which foods carry electrolytes without carrying energy.
# For every item of p09 this prints the milligrams of each electrolyte a
# hundred kilocalories buy, and the money a gram of each electrolyte
# costs from that item. An item with no energy is listed as unlimited on
# the first measure, which is the point of a salt: it adds the mineral
# without adding a single kilocalorie to the day's budget.
ITEMS = (
("avocado 150 g", 11, 727, 44, 240, 1.20),
("spinach 180 g", 126, 839, 157, 41, 0.90),
("almonds 28 g", 0, 208, 76, 164, 0.45),
("pumpkin seeds 28 g", 5, 223, 156, 158, 0.50),
("salmon 150 g", 75, 534, 44, 310, 3.50),
("bone broth 250 ml", 860, 110, 10, 40, 1.10),
("table salt 2.5 g", 970, 0, 0, 0, 0.01),
("lite salt 2 g", 0, 1050, 0, 0, 0.05),
("magnesium capsule", 0, 0, 100, 0, 0.12),
)
def per_100kcal(mg, kcal):
return None if kcal == 0 else 100.0 * mg / kcal
def cost_per_g(cost, mg):
return None if mg == 0 else 1000.0 * cost / mg
def show(v, width, digits, blank="no energy"):
if v is None:
return "{:<{}}".format(blank, width)
return "{:<{}.{}f}".format(v, width, digits)
if __name__ == "__main__":
print(" milligrams per 100 kcal")
print(" item sodium potassium magnesium")
for name, na, k, mg, kcal, cost in ITEMS:
print(" {:<22}{}{}{}"
.format(name, show(per_100kcal(na, kcal), 11, 1),
show(per_100kcal(k, kcal), 11, 1),
show(per_100kcal(mg, kcal), 11, 1)))
print(" money per gram of mineral")
print(" item sodium potassium magnesium")
for name, na, k, mg, kcal, cost in ITEMS:
print(" {:<22}{}{}{}"
.format(name, show(cost_per_g(cost, na), 11, 3, "none"),
show(cost_per_g(cost, k), 11, 3, "none"),
show(cost_per_g(cost, mg), 11, 3, "none")))
best_k = min((cost_per_g(c, k), n) for n, _, k, _, _, c in ITEMS
if k > 0)
best_mg = min((cost_per_g(c, m), n) for n, _, _, m, _, c in ITEMS
if m > 0)
best_na = min((cost_per_g(c, a), n) for n, a, _, _, _, c in ITEMS
if a > 0)
print(" cheapest gram of sodium: {} at {:.3f}"
.format(best_na[1], best_na[0]))
print(" cheapest gram of potassium: {} at {:.3f}"
.format(best_k[1], best_k[0]))
print(" cheapest gram of magnesium: {} at {:.3f}"
.format(best_mg[1], best_mg[0]))
dense = max((per_100kcal(k, kc) + per_100kcal(m, kc), n)
for n, _, k, m, kc, _ in ITEMS if kc > 0)
print(" densest whole food per 100 kcal: {}".format(dense[1]))
print(" it carries {:.1f} mg of potassium and magnesium together"
.format(dense[0]))
worst = min((per_100kcal(k, kc) + per_100kcal(m, kc), n)
for n, _, k, m, kc, _ in ITEMS if kc > 0)
print(" least dense: {} at {:.1f} mg".format(worst[1], worst[0]))
print(" ratio between the densest and the least dense: {:.1f}"
.format(dense[0] / worst[0]))
# milligrams per 100 kcal
# item sodium potassium magnesium
# avocado 150 g 4.6 302.9 18.3
# spinach 180 g 307.3 2046.3 382.9
# almonds 28 g 0.0 126.8 46.3
# pumpkin seeds 28 g 3.2 141.1 98.7
# salmon 150 g 24.2 172.3 14.2
# bone broth 250 ml 2150.0 275.0 25.0
# table salt 2.5 g no energy no energy no energy
# lite salt 2 g no energy no energy no energy
# magnesium capsule no energy no energy no energy
# money per gram of mineral
# item sodium potassium magnesium
# avocado 150 g 109.091 1.651 27.273
# spinach 180 g 7.143 1.073 5.732
# almonds 28 g none 2.163 5.921
# pumpkin seeds 28 g 100.000 2.242 3.205
# salmon 150 g 46.667 6.554 79.545
# bone broth 250 ml 1.279 10.000 110.000
# table salt 2.5 g 0.010 none none
# lite salt 2 g none 0.048 none
# magnesium capsule none none 1.200
# cheapest gram of sodium: table salt 2.5 g at 0.010
# cheapest gram of potassium: lite salt 2 g at 0.048
# cheapest gram of magnesium: magnesium capsule at 1.200
# densest whole food per 100 kcal: spinach 180 g
# it carries 2429.3 mg of potassium and magnesium together
# least dense: almonds 28 g at 173.2 mg
# ratio between the densest and the least dense: 14.0按每一百千卡计,菠菜带 2046.3 毫克钾与 382.9 毫克镁,这就是它出现在那份便宜方案里的原因;杏仁带的是 126.8 与 46.3。在密度最高与最低的全食物之间,比值是 14.0。改用金钱来量,最便宜的一克钠是食盐的 0.010,最便宜的一克钾是低钠盐的 0.048,最便宜的一克镁是胶囊的 1.200,而最好的全食物来源分别是 1.279、1.073 与 3.205。就钠而言,普通的盐罐把最好的食物甩开一百倍以上。
注
下界不是目标,也不是上限。 p09 里的三个数字是一套方案应当清过的下界,不是任何人应当力求超过的量。钠尤其有一个对高血压、心脏病或肾病患者要紧的上限;而本章的钾数字对任何肾功能下降者,或在服用保钾利尿剂、血管紧张素转换酶抑制剂或血管紧张素受体阻滞剂的人,都是危险的,在这些人身上额外补钾可能致命。第八章算出这一群人有多大。本章是一道最优化习题,不是一张补剂时刻表。
假说 · 尚未验证
作者的诠释(尚未验证): 被描述为低碳水饮食适应期的那些症状,大部分是钠耗竭;一套从第一天起就清过钠下界的方案,会显示那组症状基本不出现,而不只是被缩短。什么会证明它错: 在一项把同一套低碳水饮食随机分为补钠到下界与不补钠两组的比较里,如果第一周的头痛、起身时的眼前发白与抽筋在两组中的发生率相近,那么耗竭这一解释就是错的。
小结
刻意压低胰岛素会让肾脏排钠,而方案必须把它放回去。对 262144 种组合的穷举搜索把最便宜的合规一天定在允许用盐时的 1.44 与只用全食物时的 9.95,比值 6.91,而全食物这条路要多花 676 千卡,可预算里并没有这些。第六章接下同样的约束,把整整一天解成一个线性规划。
参考文献
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).