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

第四章 三个阶段,以及一个自相矛盾的测量

方向为什么会翻转

关于慢性压力的流行说法有三个阶段:高皮质醇的警觉期、抵抗期,以及低皮质醇的衰竭期。它通常被当作已确立的生理学呈现,而且通常不带机制。不过,确实有一个扎实的经验理由让人认真对待这个形状。Miller、Chen 与 Zhou 对人体研究的荟萃分析发现,皮质醇变化的方向取决于压力源开始了多久:起病后不久皮质醇升高,而起病很久之后日节律变得扁平、晨间水平更低 (Miller et al. 2007)。方向确实会翻转,而翻转是时间的函数。

证据

已发表。 Miller、Chen 与 Zhou 的荟萃分析发现,压力源起病至今的时间可预测皮质醇变化的方向:起病后不久升高,很久之后节律变平且晨值更低 (Miller et al. 2007)。Adam 等人的荟萃分析发现,在许多研究中,更平的日间斜率与更差的心理与身体健康结局相关 (Adam et al. 2017)。

这个说法缺的是一个理由,而第三章提供了其中一半。受体下调削弱刹车,于是抬高皮质醇:这就是第一阶段,而它是一个正反馈。但第三章同时显示受体有一个下限,所以受体解释不了一场坍塌。另一半必须来自肾上腺,而模型所用的机制是:持续偏高的中枢输出会磨损肾上腺作答的能力。关键在于,那个输出本身又被削弱的刹车抬高,于是两个过程互相喂养。

推导

# Chapter 4. p07: Why the same load first raises cortisol and later
# lowers it. Three slow variables are followed day by day. Receptor
# density falls as occupancy rises, which weakens the brake and raises
# cortisol further. Adrenal responsiveness falls while the central
# output of corticotropin stays above a threshold, and that output is
# itself raised by the weakened brake, so the two feed each other: as
# the adrenal answers less, cortisol falls, the brake weakens more, the
# central output rises more, and the damage accelerates. The daily mean
# cortisol is solved each day from the balance between drive, brake and
# adrenal. The fast loop of p01 is averaged out at this timescale. The
# end state of this model, a low cortisol with a high central output,
# is also what real adrenal failure looks like, which is why Chapter 8
# refuses to treat a flat profile as a self-diagnosis.
SYNTH, BASAL, EXTRA = 1.0, 1.0 / 28.0, 3.0 / 28.0
CMAX, FEED_N, FEED_K = 6.0, 2.0, 1.0
TAU_A, DMG, ACTH_THR = 240.0, 0.08, 0.20
LOAD_FACTOR, LOAD_END, TOTAL = 2.2, 240, 720
HI, MID, LO = 1.30, 1.00, 0.80

def occupancy(c):
    return c / (1.0 + c)

def solve_c(drive, r, a):
    lo, hi = 0.0, 50.0
    for _ in range(80):
        m = 0.5 * (lo + hi)
        if CMAX * drive * a / (1.0 + r * (m / FEED_K) ** FEED_N) > m:
            lo = m
        else:
            hi = m
    return 0.5 * (lo + hi)

def acth(drive, r, c):
    return drive / (1.0 + r * (c / FEED_K) ** FEED_N)

if __name__ == "__main__":
    lo, hi = 0.0, 5.0
    for _ in range(80):
        m = 0.5 * (lo + hi)
        if solve_c(m, 1.0, 1.0) < 1.0:
            lo = m
        else:
            hi = m
    base_drive = 0.5 * (lo + hi)
    r_base = SYNTH / (BASAL + EXTRA * occupancy(1.0))
    print("   drive that makes the normal mean one: {:.4f}"
          .format(base_drive))
    print("   corticotropin output at that mean: {:.4f}"
          .format(acth(base_drive, 1.0, 1.0)))
    r, a = 1.0, 1.0
    traj = []
    for day in range(TOTAL):
        drive = base_drive * (LOAD_FACTOR if day < LOAD_END else 1.0)
        c = solve_c(drive, r, a)
        ac = acth(drive, r, c)
        traj.append((day, c, r, a, ac))
        rt = (SYNTH / (BASAL + EXTRA * occupancy(c))) / r_base
        r += (rt - r) * (BASAL + EXTRA * occupancy(c))
        a += (1.0 - a) / TAU_A - DMG * max(0.0, ac - ACTH_THR) * a
        a = max(0.02, min(1.0, a))
    print("   day    cortisol   receptors   adrenal   corticotropin")
    for day in (0, 15, 45, 90, 135, 180, 239, 300, 420, 540, 719):
        d, c, rr, aa, ac = traj[day]
        print("   {:<7d}{:<11.4f}{:<12.4f}{:<10.4f}{:.4f}"
              .format(d, c, rr, aa, ac))
    load = traj[:LOAD_END]
    peak = max(load, key=lambda t: t[1])
    print("   peak cortisol under load: {:.4f} on day {}"
          .format(peak[1], peak[0]))
    def first(pred):
        return next((t[0] for t in load if pred(t[1])), None)
    d_hi = first(lambda c: c < HI)
    d_mid = first(lambda c: c < MID)
    d_lo = first(lambda c: c < LO)
    print("   stage one, cortisol above {:.2f}: days 0 to {}"
          .format(HI, d_hi - 1))
    print("   stage two, between {:.2f} and {:.2f}: days {} to {}"
          .format(LO, HI, d_hi, d_lo - 1))
    print("   it crosses the normal mean on day {}".format(d_mid))
    print("   stage three, below {:.2f}: from day {} onward"
          .format(LO, d_lo))
    end = traj[LOAD_END - 1]
    print("   at the end of the load: cortisol {:.4f}, receptors {:.4f},"
          .format(end[1], end[2]))
    print("   adrenal {:.4f}, corticotropin {:.4f}".format(end[3], end[4]))
    print("   corticotropin has risen to {:.2f} times its normal value"
          .format(end[4] / acth(base_drive, 1.0, 1.0)))
    print("   while cortisol has fallen to {:.4f} of normal".format(end[1]))
    print("   that pair, low cortisol with high corticotropin, is the")
    print("   signature the model produces and the one a doctor reads")
    print("   after the load is lifted, days until each passes 0.80:")
    for name, idx in (("cortisol", 1), ("receptors", 2), ("adrenal", 3)):
        d = next((t[0] - LOAD_END for t in traj[LOAD_END:]
                  if t[idx] > 0.80), None)
        print("   {:<12}{}".format(name, d if d is not None
                                   else "not within this run"))
    last = traj[TOTAL - 1]
    print("   {} days after the load: cortisol {:.4f}, adrenal {:.4f}"
          .format(TOTAL - LOAD_END, last[1], last[3]))
    print("   so the receptor recovers first, in weeks, and the adrenal")
    print("   last, in many months: one axis, two very different clocks")
#    drive that makes the normal mean one: 0.3333
#    corticotropin output at that mean: 0.1667
#    day    cortisol   receptors   adrenal   corticotropin
#    0      1.4364     1.0000      1.0000    0.2394
#    15     1.4362     0.9227      0.9476    0.2526
#    45     1.3580     0.9124      0.8280    0.2734
#    90     1.1964     0.9405      0.6379    0.3126
#    135    1.0031     0.9854      0.4540    0.3682
#    180    0.7948     1.0524      0.3007    0.4405
#    239    0.5680     1.1705      0.1778    0.5323
#    300    0.4476     1.3225      0.2831    0.2635
#    420    0.6142     1.1774      0.4435    0.2308
#    540    0.7396     1.1048      0.5933    0.2078
#    719    0.8850     1.0403      0.8031    0.1837
#    peak cortisol under load: 1.4433 on day 7
#    stage one, cortisol above 1.30: days 0 to 62
#    stage two, between 0.80 and 1.30: days 63 to 178
#    it crosses the normal mean on day 136
#    stage three, below 0.80: from day 179 onward
#    at the end of the load: cortisol 0.5680, receptors 1.1705,
#    adrenal 0.1778, corticotropin 0.5323
#    corticotropin has risen to 3.19 times its normal value
#    while cortisol has fallen to 0.5680 of normal
#    that pair, low cortisol with high corticotropin, is the
#    signature the model produces and the one a doctor reads
#    after the load is lifted, days until each passes 0.80:
#    cortisol    360
#    receptors   0
#    adrenal     476
#    480 days after the load: cortisol 0.8850, adrenal 0.8031
#    so the receptor recovers first, in weeks, and the adrenal
#    last, in many months: one axis, two very different clocks

这条轨迹有着流行说法所描述的形状,而且带着日期。皮质醇在第 0 至 62 天高于常态的 1.30,在第 63 至 178 天介于两个阈值之间,第 136 天穿过常态均值,而自第 179 天起低于 0.80。负荷结束时皮质醇是常态的 0.5680,受体在 1.1705,肾上腺在 0.1778,而促肾上腺皮质激素已升到其常态值的 3.19 倍。

负荷期间与之后的皮质醇、受体密度与肾上腺反应性(p07)。

最后那一对就是本章存在的目的所要作出的更正。在模型的第三阶段,皮质醇低而促肾上腺皮质激素高。这不是一个被要求得更少的衰竭腺体的标志;这是一个答不出所被要求之物的腺体的标志,而它恰恰就是原发性肾上腺皮质功能不全的化验图像。所以这个模型产生了一个医生必须与一种疾病区分开的状态,而这就是第八章不允许读者独自解读扁平曲线的原因。

负荷解除之后的恢复把那两个时钟明明白白地显示出来。受体立刻越过 0.80,因为它们在坍塌期已经过冲;皮质醇要 360 天,肾上腺要 476 天。负荷之后四百八十天,皮质醇是 0.8850,肾上腺是 0.8031。

一个测量,两个定义

这条轴最常见的现场测量,是醒后半小时内皮质醇的上升。Fries、Dettenborn 与 Kirschbaum 综述了它反映与不反映什么 (Fries et al. 2009),而 Stalder 等人的专家共识指南列出结果有多依赖测量的细节,其中包括上升是以绝对值表示还是相对于觉醒值表示 (Stalder et al. 2016)。

证据

已发表。 Fries、Dettenborn 与 Kirschbaum 综述觉醒反应及其决定因素 (Fries et al. 2009)。Stalder 等人的专家共识指南规定取样、时序与指标的选择,并讨论绝对与相对测量之间的选择如何影响结论 (Stalder et al. 2016)。

推导

# Chapter 4. p08: The awakening rise, and two ways of measuring it that
# disagree. On waking, cortisol climbs for about half an hour and then
# falls. The size of that climb is the most used field measure of the
# axis. The three states below are the ones p07 reached on days 0, 90
# and 239 of its load: a baseline, the middle of the slide and the
# collapse. The burst's height is taken as proportional to adrenal
# responsiveness and the baseline as the day's mean. Two measures are
# computed for each: the absolute rise above waking, and the rise as a
# percentage of the waking value. They do not agree, and which one a
# study used decides what it reported.
STATES = (
    # label, daily mean cortisol, adrenal responsiveness
    ("baseline, day 0",      1.4364, 1.0000),
    ("mid slide, day 90",    1.1964, 0.6379),
    ("collapse, day 239",    0.5680, 0.1778),
    ("recovered, day 719",   0.8850, 0.8031),
)
BURST, TPEAK, MINUTES = 1.10, 30.0, 90
E = 2.718281828459045

def burst(t, a):
    # a gamma-like rise peaking at TPEAK minutes, scaled by responsiveness
    return BURST * a * (t / TPEAK) * E ** (1.0 - t / TPEAK)

def profile(mean, a):
    base = 0.62 * mean          # the waking value, below the daily mean
    return [base + burst(float(t), a) for t in range(MINUTES + 1)], base

if __name__ == "__main__":
    print("   state                 waking   peak     rise    rise pct")
    rows = []
    for label, mean, a in STATES:
        p, base = profile(mean, a)
        rise = max(p) - base
        rows.append((label, base, max(p), rise, rise / base,
                     sum(x - base for x in p) / 60.0))
        print("   {:<22}{:<9.4f}{:<9.4f}{:<8.4f}{:.4f}"
              .format(label, base, max(p), rise, rise / base))
    print("   state                 area above waking   relative area")
    first = rows[0]
    for label, base, peak, rise, pct, area in rows:
        print("   {:<22}{:<20.4f}{:.4f}"
              .format(label, area, area / first[5]))
    print("   the absolute rise falls to {:.4f} of its baseline value at"
          .format(rows[2][3] / first[3]))
    print("   the collapse, and to {:.4f} at the middle of the slide"
          .format(rows[1][3] / first[3]))
    print("   but the percentage rise falls only to {:.4f} and {:.4f}"
          .format(rows[2][4] / first[4], rows[1][4] / first[4]))
    print("   of their baseline values, because the waking value fell too")
    print("   ratio between what the two measures report at the collapse:")
    print("   {:.2f}".format((rows[2][4] / first[4])
                             / (rows[2][3] / first[3])))
    print("   so a study reporting the percentage rise sees a shallower")
    print("   flattening than one reporting the absolute rise, on exactly")
    print("   the same people, which is a measurement artefact and not a")
    print("   disagreement about biology")
    print("   minute by minute at the collapse against baseline:")
    print("   minute   baseline   collapse   ratio")
    pb, bb = profile(STATES[0][1], STATES[0][2])
    pc, bc = profile(STATES[2][1], STATES[2][2])
    for t in (0, 10, 20, 30, 45, 60, 90):
        print("   {:<9d}{:<11.4f}{:<11.4f}{:.4f}"
              .format(t, pb[t], pc[t], pc[t] / pb[t]))
#    state                 waking   peak     rise    rise pct
#    baseline, day 0       0.8906   1.9906   1.1000  1.2352
#    mid slide, day 90     0.7418   1.4435   0.7017  0.9460
#    collapse, day 239     0.3522   0.5477   0.1956  0.5554
#    recovered, day 719    0.5487   1.4321   0.8834  1.6100
#    state                 area above waking   relative area
#    baseline, day 0       1.2009              1.0000
#    mid slide, day 90     0.7660              0.6379
#    collapse, day 239     0.2135              0.1778
#    recovered, day 719    0.9644              0.8031
#    the absolute rise falls to 0.1778 of its baseline value at
#    the collapse, and to 0.6379 at the middle of the slide
#    but the percentage rise falls only to 0.4496 and 0.7659
#    of their baseline values, because the waking value fell too
#    ratio between what the two measures report at the collapse:
#    2.53
#    so a study reporting the percentage rise sees a shallower
#    flattening than one reporting the absolute rise, on exactly
#    the same people, which is a measurement artefact and not a
#    disagreement about biology
#    minute by minute at the collapse against baseline:
#    minute   baseline   collapse   ratio
#    0        0.8906     0.3522     0.3954
#    10       1.6047     0.4791     0.2986
#    20       1.9140     0.5341     0.2791
#    30       1.9906     0.5477     0.2752
#    45       1.8913     0.5301     0.2803
#    60       1.6999     0.4961     0.2918
#    90       1.3372     0.4316     0.3227

取 p07 的三个状态,并给每一个一次与肾上腺反应性成正比的觉醒爆发,绝对上升幅度从基线的 1.1000 落到坍塌期的 0.1956,也就是其基线值的 0.1778。但以觉醒值百分比表示的上升只从 1.2352 落到 0.5554,也就是其基线值的 0.4496,因为觉醒值本身也从 0.8906 落到了 0.3522。两种测量相差 2.53 倍。

这是一处测量假象,不是关于生物学的分歧,而它对阅读文献有实际后果。对同一批人的两项研究,一项报告绝对上升、一项报告百分比上升,所报告的扁平化程度会相差两倍以上,而两者都没有错。

假说 · 尚未验证

作者的诠释(尚未验证): 已发表的觉醒反应文献中相当一部分异质性,可归因于绝对与相对指标之间的选择,而不是人群之间的差异;用两种指标重新分析这些研究会缩小所报告效应量的离散。什么会证明它错: 如果一项把指标选择编码为调节变量的荟萃分析发现它只解释了很少的研究间异质性,那么这处假象就不是离散的驱动因素。

小结

把受体下调与肾上腺磨损耦合起来,复现了带日期的三个阶段:第 0 至 62 天高于 1.30,到第 178 天介于两者之间,自第 179 天起低于 0.80,终于皮质醇 0.5680 而促肾上腺皮质激素为常态的 3.19 倍。那一对正是肾上腺衰竭的图像,必须由医生与之区分。觉醒上升按绝对值测落到基线的 0.1778,按百分比测只落到 0.4496,相差 2.53 倍。第五章为人们实际会选的那些方案定价。

参考文献

Adam, Emma K., Meghan E. Quinn, Royette Tavernier, Mollie T. McQuillan, Katie A. Dahlke, and Kirsten E. Gilbert. 2017. “Diurnal Cortisol Slopes and Mental and Physical Health Outcomes: A Systematic Review and Meta-Analysis.” Psychoneuroendocrinology 83: 25 to 41.
Fries, Eva, Lucia Dettenborn, and Clemens Kirschbaum. 2009. “The Cortisol Awakening Response: Facts and Future Directions.” International Journal of Psychophysiology 72 (1): 67 to 73.
Miller, Gregory E., Edith Chen, and Eric S. Zhou. 2007. “If It Goes up, Must It Come down? Chronic Stress and the Hypothalamic-Pituitary-Adrenocortical Axis in Humans.” Psychological Bulletin 133 (1): 25 to 45.
Stalder, Tobias, Clemens Kirschbaum, Brigitte M. Kudielka, et al. 2016. “Assessment of the Cortisol Awakening Response: Expert Consensus Guidelines.” Psychoneuroendocrinology 63: 414 to 432.

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).