5D Library › 复合补剂相互作用、时间疗法给药窗口与代谢酶竞争 · English

第五章 肝脏里的两个酶

清除咖啡因的那一个

咖啡因几乎是单个酶的完美探针,因为它的清除几乎全部走一条通路。Carrillo 与 Benitez 编目了改变那条通路的药物以及改变的幅度,从吸烟带来的适度诱导,到某些抗抑郁药带来的数倍抑制 (Carrillo and Benitez 2000)。在问题的另一侧,Drake 等人在睡前零、三与六小时给予咖啡因并客观测量睡眠扰动,发现即便在六小时也有效应 (Drake et al. 2013)。

证据

已发表。 Carrillo 与 Benitez 综述膳食咖啡因与药物之间具有临床意义的相互作用,并给出每一项的方向与大致幅度 (Carrillo and Benitez 2000)。Drake 等人测量了睡前六小时以内摄入咖啡因所造成的客观睡眠扰动 (Drake et al. 2013)。

把半衰期的倍数放进指数,实际问题就自己回答了。

推导

# Chapter 5. p09: Why the same coffee ruins one person's sleep and not
# another's. Caffeine is cleared almost entirely by one liver enzyme,
# and the activity of that enzyme varies by an order of magnitude
# between people and is changed by common medicines and by smoking. The
# half-life multipliers below are rounded from the published ranges.
# From a dose at a given hour the residual at a fixed bedtime follows,
# and from the residual a sleep-onset latency, modelled as the baseline
# latency multiplied by one plus the residual over a reference amount.
DOSE_MG = 100.0
BASE_HALFLIFE = 5.0
BEDTIME = 23.0
BASE_LATENCY = 14.0        # minutes to fall asleep with no caffeine
REF_MG = 40.0              # residual that doubles the latency
THRESHOLD = 25.0           # mg at bedtime a careful plan stays under
STATES = (
    ("no modifier",              1.00),
    ("a smoker",                 0.60),
    ("oral contraceptive",       1.60),
    ("ciprofloxacin",            2.10),
    ("late pregnancy",           3.50),
    ("fluvoxamine",              4.50),
)
HOURS = (8.0, 10.0, 12.0, 14.0, 16.0, 18.0)
LN2 = 0.6931471805599453
E = 2.718281828459045

def residual(dose, hour, halflife):
    dt = BEDTIME - hour
    if dt <= 0.0:
        return dose
    return dose * E ** (-LN2 * dt / halflife)

def latency(mg):
    return BASE_LATENCY * (1.0 + mg / REF_MG)

if __name__ == "__main__":
    print("   {:.0f} mg of caffeine, bedtime at {:.0f}:00".format(DOSE_MG,
                                                                  BEDTIME))
    print("   state                 half-life   mg left   minutes to sleep")
    for name, mult in STATES:
        hl = BASE_HALFLIFE * mult
        mg = residual(DOSE_MG, 14.0, hl)
        print("   {:<22}{:<12.2f}{:<10.2f}{:.1f}"
              .format(name, hl, mg, latency(mg)))
    print("   that column is for a dose at 14:00")
    print("   milligrams left at bedtime, by hour of the dose:")
    print("   state                 " + "".join("{:>8.0f}".format(h)
                                                for h in HOURS))
    for name, mult in STATES:
        hl = BASE_HALFLIFE * mult
        print("   {:<22}".format(name)
              + "".join("{:>8.2f}".format(residual(DOSE_MG, h, hl))
                        for h in HOURS))
    print("   latest hour at which {:.0f} mg keeps the bedtime residual"
          .format(DOSE_MG))
    print("   under {:.0f} mg:".format(THRESHOLD))
    for name, mult in STATES:
        hl = BASE_HALFLIFE * mult
        if residual(DOSE_MG, 0.0, hl) >= THRESHOLD:
            print("   {:<22}{}".format(name, "no hour in the day works"))
            continue
        lo, hi = 0.0, BEDTIME
        for _ in range(60):
            mid = 0.5 * (lo + hi)
            if residual(DOSE_MG, mid, hl) < THRESHOLD:
                lo = mid
            else:
                hi = mid
        mins = int(round(0.5 * (lo + hi) * 60.0))
        print("   {:<22}{:02d}:{:02d}".format(name, mins // 60, mins % 60))
    hl_a = BASE_HALFLIFE * 0.60
    hl_b = BASE_HALFLIFE * 4.50
    print("   between the fastest and the slowest state the half-life")
    print("   differs by {:.2f} times".format(hl_b / hl_a))
    print("   and the bedtime residual from a 14:00 dose by {:.2f} times"
          .format(residual(DOSE_MG, 14.0, hl_b)
                  / residual(DOSE_MG, 14.0, hl_a)))
    print("   with the slowest clearance the latency rises from {:.1f}"
          .format(latency(residual(DOSE_MG, 14.0, hl_a))))
    print("   to {:.1f} minutes, a difference of {:.1f}"
          .format(latency(residual(DOSE_MG, 14.0, hl_b)),
                  latency(residual(DOSE_MG, 14.0, hl_b))
                  - latency(residual(DOSE_MG, 14.0, hl_a))))
#    100 mg of caffeine, bedtime at 23:00
#    state                 half-life   mg left   minutes to sleep
#    no modifier           5.00        28.72     24.1
#    a smoker              3.00        12.50     18.4
#    oral contraceptive    8.00        45.85     30.0
#    ciprofloxacin         10.50       55.20     33.3
#    late pregnancy        17.50       70.01     38.5
#    fluvoxamine           22.50       75.79     40.5
#    that column is for a dose at 14:00
#    milligrams left at bedtime, by hour of the dose:
#    state                        8      10      12      14      16      18
#    no modifier              12.50   16.49   21.76   28.72   37.89   50.00
#    a smoker                  3.13    4.96    7.87   12.50   19.84   31.50
#    oral contraceptive       27.26   32.42   38.56   45.85   54.53   64.84
#    ciprofloxacin            37.15   42.39   48.38   55.20   63.00   71.89
#    late pregnancy           55.20   59.76   64.68   70.01   75.79   82.03
#    fluvoxamine              63.00   67.00   71.26   75.79   80.60   85.72
#    latest hour at which 100 mg keeps the bedtime residual
#    under 25 mg:
#    no modifier           13:00
#    a smoker              17:00
#    oral contraceptive    07:00
#    ciprofloxacin         02:00
#    late pregnancy        no hour in the day works
#    fluvoxamine           no hour in the day works
#    between the fastest and the slowest state the half-life
#    differs by 7.50 times
#    and the bedtime residual from a 14:00 dose by 6.06 times
#    with the slowest clearance the latency rises from 18.4
#    to 40.5 minutes, a difference of 22.2

下午两点的一百毫克,在十一点就寝时,对没有任何修饰因素的人剩下 28.72 毫克,对吸烟者剩下 12.50 毫克,而对服用氟伏沙明的人剩下 75.79 毫克;半衰期相差 7.50 倍,两端之间就相差 6.06 倍。建模的入睡潜伏期从最快状态的 18.4 分钟升到最慢状态的 40.5 分钟,同一杯咖啡造成 22.2 分钟的差别,而对完全没有修饰因素的人是 24.1 分钟。

最后那张表里带着一条指令。对没有修饰因素的人,一百毫克能让就寝残留低于阈值的最晚时辰是 13:00;吸烟者是 17:00;服用口服避孕药的是 07:00;服用环丙沙星的是 02:00。而对晚期妊娠与服用氟伏沙明的人,一天之中没有任何时辰可行。所以「下午两点以后不喝咖啡」这类规则,对中位数的人是对的,而对相当一部分少数人在两个方向上各错六小时,这正是任何只引用一个钟点、却不问这个人还吃什么的建议的一般形状。

清除其余大部分东西的那一个

第二个酶之所以要紧,不是因为某一种物质用它,而是因为太多东西都用它。Ito 等人陈述了这样一个方程:给定某底物的清除中有多少比例由这个共享酶承担,就能预测一个竞争性抑制剂把该底物的暴露提高多少 (Ito et al. 1998)。这个方程有一个性质,比它所产生的任何单一数字都更有价值。

推导

# Chapter 5. p10: What an inhibitor does to everything else in the
# cupboard. If a fraction fm of a substance's clearance runs through one
# enzyme and an inhibitor raises that enzyme's apparent affinity by one
# plus the ratio of inhibitor to its own constant, then the exposure to
# the substance is multiplied by one over fm divided by that factor plus
# one minus fm. This is the standard interaction equation, and it has a
# property worth seeing: the multiplier is bounded by one over one minus
# fm however strong the inhibitor is, so a substance cleared by several
# routes is safe and one cleared by a single route is not.
import math

FM = (0.30, 0.50, 0.70, 0.90, 0.95, 0.99)
RATIO = (0.0, 0.5, 1.0, 2.0, 5.0, 20.0)
PIPERINE_RATIO = 2.4       # illustrative, at a 10 mg dose with food
HALFLIFE_BASE = 6.0        # hours, the substance in the example
WASHOUT_FRACTION = 0.05    # when the interaction is effectively over
INHIB_HALFLIFE = 2.5       # hours, the inhibitor itself
LN2 = 0.6931471805599453
E = 2.718281828459045

def exposure(fm, ratio):
    return 1.0 / (fm / (1.0 + ratio) + (1.0 - fm))

if __name__ == "__main__":
    print("   exposure multiplier, inhibitor ratio by cleared fraction")
    print("   ratio " + "".join("{:>9.2f}".format(f) for f in FM))
    for r in RATIO:
        print("   {:<6.1f}".format(r)
              + "".join("{:>9.3f}".format(exposure(f, r)) for f in FM))
    print("   ceiling at an infinitely strong inhibitor:")
    print("   fraction cleared   ceiling")
    for f in FM:
        print("   {:<19.2f}{:.3f}".format(f, 1.0 / (1.0 - f)))
    print("   at the piperine ratio of {:.1f} the multipliers are:"
          .format(PIPERINE_RATIO))
    for f in FM:
        print("   fraction {:.2f}: {:.3f}, which is {:.3f} of the ceiling"
              .format(f, exposure(f, PIPERINE_RATIO),
                      exposure(f, PIPERINE_RATIO) / (1.0 / (1.0 - f))))
    print("   the half-life of the substance is stretched by the same")
    print("   factor, so for a {:.1f} hour half-life and a fraction of"
          .format(HALFLIFE_BASE))
    for f in (0.70, 0.90, 0.99):
        hl = HALFLIFE_BASE * exposure(f, PIPERINE_RATIO)
        print("   {:.2f} it becomes {:.2f} hours".format(f, hl))
    print("   time for the inhibitor itself to fall to {:.2f} of its peak:"
          .format(WASHOUT_FRACTION))
    hours = INHIB_HALFLIFE * math.log(1.0 / WASHOUT_FRACTION) / LN2
    print("   {:.2f} hours".format(hours))
    print("   so a once-daily inhibitor taken in the morning is still")
    print("   present in the evening, and the schedule has to allow for")
    print("   it rather than treat each dose as independent")
    print("   worked example: a substance cleared 0.90 by this enzyme,")
    print("   taken with piperine, reaches {:.3f} times its usual"
          .format(exposure(0.90, PIPERINE_RATIO)))
    print("   exposure, against a ceiling of {:.3f}"
          .format(1.0 / (1.0 - 0.90)))
    print("   and a substance cleared 0.30 reaches only {:.3f}"
          .format(exposure(0.30, PIPERINE_RATIO)))
    print("   ratio between those two: {:.2f}"
          .format(exposure(0.90, PIPERINE_RATIO)
                  / exposure(0.30, PIPERINE_RATIO)))
#    exposure multiplier, inhibitor ratio by cleared fraction
#    ratio      0.30     0.50     0.70     0.90     0.95     0.99
#    0.0       1.000    1.000    1.000    1.000    1.000    1.000
#    0.5       1.111    1.200    1.304    1.429    1.463    1.493
#    1.0       1.176    1.333    1.538    1.818    1.905    1.980
#    2.0       1.250    1.500    1.875    2.500    2.727    2.941
#    5.0       1.333    1.714    2.400    4.000    4.800    5.714
#    20.0      1.400    1.909    3.000    7.000   10.500   17.500
#    ceiling at an infinitely strong inhibitor:
#    fraction cleared   ceiling
#    0.30               1.429
#    0.50               2.000
#    0.70               3.333
#    0.90               10.000
#    0.95               20.000
#    0.99               100.000
#    at the piperine ratio of 2.4 the multipliers are:
#    fraction 0.30: 1.269, which is 0.888 of the ceiling
#    fraction 0.50: 1.545, which is 0.773 of the ceiling
#    fraction 0.70: 1.977, which is 0.593 of the ceiling
#    fraction 0.90: 2.742, which is 0.274 of the ceiling
#    fraction 0.95: 3.036, which is 0.152 of the ceiling
#    fraction 0.99: 3.320, which is 0.033 of the ceiling
#    the half-life of the substance is stretched by the same
#    factor, so for a 6.0 hour half-life and a fraction of
#    0.70 it becomes 11.86 hours
#    0.90 it becomes 16.45 hours
#    0.99 it becomes 19.92 hours
#    time for the inhibitor itself to fall to 0.05 of its peak:
#    10.80 hours
#    so a once-daily inhibitor taken in the morning is still
#    present in the evening, and the schedule has to allow for
#    it rather than treat each dose as independent
#    worked example: a substance cleared 0.90 by this enzyme,
#    taken with piperine, reaches 2.742 times its usual
#    exposure, against a ceiling of 10.000
#    and a substance cleared 0.30 reaches only 1.269
#    ratio between those two: 2.16

无论抑制剂变得多强,暴露倍数都被一除以一减清除分数所封顶。一种有 0.30 由该酶清除的物质,永远不会超过其通常暴露的 1.429 倍;而有 0.99 由该酶清除的,可以达到 100.000 倍。在 2.4 这个示意的胡椒碱比值下,清除分数 0.30 的倍数是 1.269,0.90 的是 2.742,两者相差 2.16 倍,而它们分别是各自上限的 0.888 与 0.274。

那个上界就是本章的实际内容。它说的是,关于一个抑制剂要问的不是它有多强,而是这个人还吃着什么、而那样东西的清除是否依赖单一通路。一个中等强度的抑制剂旁边放着一种几乎全靠被抑制的酶清除的物质,比一个强力抑制剂旁边放着一种有三条出路的物质更危险。而由于抑制剂自身的半衰期决定效应持续多久,一次晨服的每日一次抑制剂到傍晚仍然在场:2.5 小时的半衰期要 10.80 小时才落到峰值的二十分之一。

注

这里从算术习题变成真实风险的地方。 胡椒碱被当作吸收促进剂出售,而这恰恰就是酶抑制的一种描述;它给某一种补剂带来的促进,也同样给每一种由同一通路清除的药物带来。对一种有效剂量与有害剂量之间余地很窄的药物来说,2.742 倍不是好处。第八章数出有多少成年人在服用这类药物。任何在服用处方药的人,都应把吸收促进剂当作一次药物相互作用去问药师,而不是当作一种不必提起的补剂。

假说 · 尚未验证

作者的诠释(尚未验证): 因为暴露倍数被一减清除分数的倒数所封顶,补剂级酶抑制剂的风险几乎全部集中在至少服用一种经受影响酶清除比例达到 0.90 或更高的药物的人身上,对其余所有人则可以忽略。什么会证明它错: 如果这类补剂所报告的不良事件在各种用药谱之间的分布,与同服药物的清除分数无关,那么这个上界就不是在支配风险。

小结

咖啡因半衰期在清除最快与最慢的状态之间相差 7.50 倍,于是 14:00 的一杯在就寝时的残留相差 6.06 倍,建模的潜伏期相差 22.2 分钟;而在六种状态里有两种,一天之中没有哪个时辰够晚。抑制一个共享酶所带来的暴露倍数被一除以一减清除分数封顶,在清除分数 0.30 处为 1.429,在 0.99 处为 100.000,而在示意抑制剂下 0.90 处为 2.742。第六章把整个柜子放到时钟上。

参考文献

Carrillo, Juan A., and Julio Benitez. 2000. “Clinically Significant Pharmacokinetic Interactions between Dietary Caffeine and Medications.” Clinical Pharmacokinetics 39 (2): 127 to 153.
Drake, Christopher, Timothy Roehrs, John Shambroom, and Thomas Roth. 2013. “Caffeine Effects on Sleep Taken 0, 3, or 6 Hours before Going to Bed.” Journal of Clinical Sleep Medicine 9 (11): 1195 to 1200.
Ito, Kiyomi, Tomoyuki Iwatsubo, Satoshi Kanamitsu, Kunihiro Ueda, Hiroshi Suzuki, and Yuichi Sugiyama. 1998. “Prediction of Pharmacokinetic Alterations Caused by Drug-Drug Interactions: Metabolic Interaction in the Liver.” Pharmacological Reviews 50 (3): 387 to 412.

Fan, Y. (2026). Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition: Michaelis-Menten Inhibition Matrices, Chelation Absorption Losses and Circadian Receptor Sensitivity [复合补剂相互作用、时间疗法给药窗口与代谢酶竞争] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090250 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).