5D Library › 下丘脑垂体肾上腺轴耦合振荡、皮质醇觉醒反应扁平化与倦怠恢复动力学 · English
第八章 这个月,以及那道闸
哪些天要紧
第六章那份方案假定负荷每天都被移除、光每天都被照到。没有人做得到,而这两种失败并不等价:负荷回来的那一天会撤销受体上的进展,而漏掉光的那一天只是把相位留在原处。
推导
# Chapter 8. p15: The thirty days, with a real person in them. The plan
# of p12 assumes the load is removed and the light is taken every day.
# Nobody does that. Here each day the load is removed with one chance
# and the light is taken with another, and the two failures cost
# different things: a day of load undoes receptor progress, and a missed
# light day leaves the phase where it was. Four thousand people, thirty
# days, seed 97.
import random
SEED, PEOPLE, DAYS = 97, 4000, 30
P_UNLOAD, P_LIGHT = 0.70, 0.80
SYNTH, BASAL, EXTRA = 1.0, 1.0 / 28.0, 3.0 / 28.0
C_ON, C_OFF = 2.5, 1.0
SHIFT, START_PHASE = 0.3515, 3.0
def occupancy(c):
return c / (1.0 + c)
def steady(c):
return SYNTH / (BASAL + EXTRA * occupancy(c))
def step(r, c):
return r + (steady(c) - r) * (BASAL + EXTRA * occupancy(c))
def settled():
r = steady(C_OFF)
for d in range(399): # ends on a Sunday, as in p09
r = step(r, C_ON if (d % 7) < 5 else C_OFF)
return r
def one(rng, start, full):
r, phase = start, START_PHASE
for _ in range(DAYS):
r = step(r, C_OFF if rng.random() < P_UNLOAD else C_ON)
if rng.random() < P_LIGHT:
phase = max(0.0, phase - SHIFT)
return r / full, phase
def q(v, a):
s = sorted(v)
return s[int(a * (len(s) - 1))]
if __name__ == "__main__":
full = steady(C_OFF)
start = settled()
rng = random.Random(SEED)
dens, phases = [], []
for _ in range(PEOPLE):
d, p = one(rng, start, full)
dens.append(d)
phases.append(p)
print(" starting density {:.4f} of rested, phase error {:.2f} hours"
.format(start / full, START_PHASE))
print(" load removed with chance {:.2f}, light taken with {:.2f}"
.format(P_UNLOAD, P_LIGHT))
print(" after {} days, density as a share of rested:".format(DAYS))
print(" median {:.4f}, 10% to 90% {:.4f} to {:.4f}"
.format(q(dens, 0.5), q(dens, 0.1), q(dens, 0.9)))
print(" phase error left: median {:.3f}, 10% to 90% {:.3f} to {:.3f}"
.format(q(phases, 0.5), q(phases, 0.1), q(phases, 0.9)))
perfect = start
for _ in range(DAYS):
perfect = step(perfect, C_OFF)
print(" a perfect month would reach {:.4f}".format(perfect / full))
print(" so imperfect adherence gives up {:.4f} of the gap"
.format(1.0 - (q(dens, 0.5) - start / full)
/ (perfect / full - start / full)))
print(" share of people who end below where they started: {:.4f}"
.format(sum(1 for d in dens if d < start / full) / PEOPLE))
print(" share whose phase is still over an hour out: {:.4f}"
.format(sum(1 for p in phases if p > 1.0) / PEOPLE))
print(" the same month with the load removed every day and the")
print(" light taken four days in five:")
rng2 = random.Random(SEED + 1)
d2 = []
for _ in range(PEOPLE):
r = start
for _ in range(DAYS):
r = step(r, C_OFF)
d2.append(r / full)
print(" median density {:.4f}, against {:.4f} above"
.format(q(d2, 0.5), q(dens, 0.5)))
print(" so the days of load matter more than the days of light,")
print(" which is the opposite of how the two are usually sold")
# starting density 0.8598 of rested, phase error 3.00 hours
# load removed with chance 0.70, light taken with 0.80
# after 30 days, density as a share of rested:
# median 0.9255, 10% to 90% 0.8954 to 0.9539
# phase error left: median 0.000, 10% to 90% 0.000 to 0.000
# a perfect month would reach 0.9915
# so imperfect adherence gives up 0.5014 of the gap
# share of people who end below where they started: 0.0030
# share whose phase is still over an hour out: 0.0000
# the same month with the load removed every day and the
# light taken four days in five:
# median density 0.9915, against 0.9255 above
# so the days of load matter more than the days of light,
# which is the opposite of how the two are usually sold在负荷有 0.70 的日子被移除、光有 0.80 的日子被照到的情况下,这个月以静息值的 0.9255 收尾,而完美的一个月是 0.9915,所以不完美的依从性放弃了缺口的 0.5014。几乎没有人比出发时更差,只有 0.0030,而相位在基本上每个人身上都被修好,误差中位数为 0.000 小时。每天都移除负荷、同时仍然五天漏一天光,则到达 0.9915。
这处不对称就是全部的讯息。光是宽容的,因为它的效应会累积,漏掉一天只花掉那一天的移动;负荷不宽容,因为有负荷的一天把慢变量推回去。所以有负荷的日子比有光的日子更要紧,而这与两者通常被售卖的方式正好相反:灯是可以买的那一部分,而假期不是。
假说 · 尚未验证
作者的诠释(尚未验证): 在把光照时序与工作负荷削减结合起来的干预里,实测收益会跟随对负荷那一部分的依从性,而几乎与对光那一部分的依从性无关;所以只监测光的试验会发现微弱而不一致的效应。什么会证明它错: 如果一项同时测量两部分依从性的试验发现光的依从性是结局更强的预测因子,那么这里算出的不对称就不成立。
本该放在最前面的那一页
以上一切都假定读者是一个对他而言这是恢复问题而不是诊断问题的人。对许多读者来说这个假定是错的,而弄错它的后果并不对称。
有两件事必须明说。第一,第四章模型所到达的那个状态,皮质醇低而促肾上腺皮质激素升高,正是原发性肾上腺皮质功能不全的化验图像。Bornstein 等人的临床实践指南列出这种疾病如何诊断与治疗,而它的治疗是激素替代,不是一套方案 (Bornstein et al. 2016)。不治疗,它会致死。第二,本书主题通常被售卖时所用的那个标签经不起检验:Cadegiani 与 Kater 对那些声称确立了「肾上腺疲劳」的研究所作的系统综述,发现方法不足、这个构念缺乏支持 (Cadegiani and Kater 2016)。第一到第六章的那些机制是真的;那个诊断不是。
证据
已发表。 Bornstein 等人的内分泌学会指南涵盖原发性肾上腺皮质功能不全的诊断与治疗,而它以替代治疗处理 (Bornstein et al. 2016)。Cadegiani 与 Kater 的系统综述检视了为支持「肾上腺疲劳」而提出的那些研究,发现这个构念缺乏证据支持 (Cadegiani and Kater 2016)。
推导
# Chapter 8. p16: The one page that must come first. The end state of
# p07, a low cortisol with a high central output, is also what primary
# adrenal failure looks like, and that is a condition people die of. So
# a flat profile is not a result a reader may interpret; it is a reason
# to be tested. The prevalences below are rounded from published figures
# and applied independently, which overstates how many people carry
# exactly one flag. The point of the table is the list of pairs, not the
# total. Twenty thousand synthetic adults, seed 97.
import random
SEED, N = 97, 20000
STOP = (
("on a glucocorticoid now", 0.0250,
"the axis is already suppressed; stopping can kill"),
("stopped one in the last year", 0.0300,
"the axis may not have recovered yet"),
("known adrenal insufficiency", 0.0012,
"needs replacement, not a protocol"),
("known Cushing syndrome", 0.0004,
"needs a surgeon, not a protocol"),
("bipolar disorder", 0.0200,
"bright morning light can precipitate mania"),
("a retinal disease", 0.0150,
"bright light therapy needs an eye assessment"),
("pregnant or breastfeeding", 0.0300,
"the axis is altered and the evidence does not transfer"),
)
CARE = (
("treated hypothyroidism", 0.0700,
"the symptoms overlap and the tests interact"),
("obstructive sleep apnoea", 0.0900,
"a flat profile may be the apnoea, not the axis"),
("a shift worker", 0.1700,
"the light protocol has to be inverted, not copied"),
("an antidepressant", 0.1100,
"sleep timing and the axis both move"),
("type 2 diabetes", 0.0700,
"cortisol changes move glucose"),
("treated hypertension", 0.2600,
"salt and fluid advice is not neutral"),
("a benzodiazepine or Z drug", 0.0400,
"withdrawal mimics every symptom in this book"),
)
if __name__ == "__main__":
rng = random.Random(SEED)
stop_n = care_n = clear_n = multi = 0
sc = [0] * len(STOP)
cc = [0] * len(CARE)
for _ in range(N):
s = [rng.random() < p for _, p, _ in STOP]
c = [rng.random() < p for _, p, _ in CARE]
for i, v in enumerate(s):
sc[i] += v
for i, v in enumerate(c):
cc[i] += v
if any(s):
stop_n += 1
if sum(s) > 1:
multi += 1
elif any(c):
care_n += 1
else:
clear_n += 1
print(" {} synthetic adults".format(N))
print(" do not attempt this without a doctor:")
for (name, p, why), k in zip(STOP, sc):
print(" {:<30}{:.4f}".format(name, k / N))
print(" {}".format(why))
print(" see a doctor first, then it may be possible:")
for (name, p, why), k in zip(CARE, cc):
print(" {:<30}{:.4f}".format(name, k / N))
print(" {}".format(why))
print(" carrying at least one absolute flag: {:.4f}"
.format(stop_n / N))
print(" of those, with more than one: {:.4f}".format(multi / stop_n))
print(" needing a doctor first but not barred: {:.4f}"
.format(care_n / N))
print(" free to follow the plan unsupervised: {:.4f}"
.format(clear_n / N))
print(" so the gate stops or refers {:.4f} of adults"
.format((stop_n + care_n) / N))
print(" the commonest reason to see a doctor first is treated")
print(" hypertension at {:.4f}, and the commonest absolute flag is"
.format(cc[5] / N))
worst = max(range(len(STOP)), key=lambda i: sc[i])
print(" {} at {:.4f}".format(STOP[worst][0], sc[worst] / N))
print(" adrenal insufficiency is rare, at {:.4f}, and that is"
.format(sc[2] / N))
print(" exactly why a flat profile must be tested rather than")
print(" assumed: the rare cause is the dangerous one")
# 20000 synthetic adults
# do not attempt this without a doctor:
# on a glucocorticoid now 0.0246
# the axis is already suppressed; stopping can kill
# stopped one in the last year 0.0302
# the axis may not have recovered yet
# known adrenal insufficiency 0.0010
# needs replacement, not a protocol
# known Cushing syndrome 0.0003
# needs a surgeon, not a protocol
# bipolar disorder 0.0196
# bright morning light can precipitate mania
# a retinal disease 0.0138
# bright light therapy needs an eye assessment
# pregnant or breastfeeding 0.0291
# the axis is altered and the evidence does not transfer
# see a doctor first, then it may be possible:
# treated hypothyroidism 0.0716
# the symptoms overlap and the tests interact
# obstructive sleep apnoea 0.0886
# a flat profile may be the apnoea, not the axis
# a shift worker 0.1705
# the light protocol has to be inverted, not copied
# an antidepressant 0.1088
# sleep timing and the axis both move
# type 2 diabetes 0.0694
# cortisol changes move glucose
# treated hypertension 0.2564
# salt and fluid advice is not neutral
# a benzodiazepine or Z drug 0.0396
# withdrawal mimics every symptom in this book
# carrying at least one absolute flag: 0.1131
# of those, with more than one: 0.0486
# needing a doctor first but not barred: 0.5184
# free to follow the plan unsupervised: 0.3684
# so the gate stops or refers 0.6315 of adults
# the commonest reason to see a doctor first is treated
# hypertension at 0.2564, and the commonest absolute flag is
# stopped one in the last year at 0.0302
# adrenal insufficiency is rare, at 0.0010, and that is
# exactly why a flat profile must be tested rather than
# assumed: the rare cause is the dangerous one在 20000 个合成成年人中,0.1131 至少带有一项绝对标志,而其中又有 0.0486 带有一项以上。另有 0.5184 带有需要先经医生评估的情况,于是剩下 0.3684 可以不经监督地遵循这份方案。这道闸拦下或转诊了 0.6315 的成年人。
注
最要紧的那几项标志。 当下正在用糖皮质激素的占 0.0246,或在过去一年内停用过的占 0.0302,两者都意味着这条轴已经被抑制,而前一种情况下骤然停药可能致命。双相情感障碍占 0.0196,是不该在没有精神科意见的情况下使用晨间强光的理由,因为它可能诱发躁狂。视网膜疾病占 0.0138,在光疗前需要眼科评估。阻塞性睡眠呼吸暂停占 0.0886,列在转诊名单里是出于另一个理由:扁平的曲线与疲惫可能是这个呼吸暂停而不是这条轴,而治错了那一个会浪费第三章所说读者并没有的那几个月。肾上腺皮质功能不全本身很罕见,只有 0.0010,而这份罕见恰恰就是去做检查的理由:罕见的那个原因正是危险的那个。
独立性假设高估了恰好带一项标志的人数,低估了带若干项的人数,因为这些情况会聚集。这并不让这道闸变小。真正要紧的那个数字,也就是被直接拦下的 0.1131,数的是带有任何一项标志的人,聚集只把它降低一点点。
小结
现实的依从性让这个月停在静息值的 0.9255,而完美的一个月是 0.9915,放弃了缺口的 0.5014,而这份损失来自有负荷的日子而不是有光的日子。安全闸直接拦下 0.1131 的成年人,另外转诊 0.5184,剩下 0.3684。本书模型所到达的那个状态,正是一种可治且危险的疾病的化验图像,而这个主题被售卖时所用的那个标签并无证据支持。本书的算术是对若干机制的描述与对其大小的一个界;它所支持的行动是与医生的一次谈话,不是一套方案。
参考文献
Fan, Y. (2026). HPA-Axis Coupled Oscillations, Cortisol Awakening Response Flattening and Burnout Recovery Kinetics: Glucocorticoid Receptor Downregulation, Ultradian Pulsatility and Allostatic Load Reset [下丘脑垂体肾上腺轴耦合振荡、皮质醇觉醒反应扁平化与倦怠恢复动力学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090252 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).