5D Library › Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition · 中文

Chapter 5. Two Enzymes in the Liver

The one that clears caffeine

Caffeine is a near-perfect probe for a single enzyme, because almost all of its clearance runs through one route. Carrillo and Benitez catalogue the medicines that change that route and the size of the changes, which range from a modest induction by smoking to a several-fold inhibition by some antidepressants (Carrillo and Benitez 2000). On the other side of the question, Drake and colleagues gave caffeine nought, three and six hours before bed and measured sleep disruption objectively, finding an effect even at six hours (Drake et al. 2013).

Evidence

Published. Carrillo and Benitez review the clinically significant interactions between dietary caffeine and medicines, with the direction and approximate size of each (Carrillo and Benitez 2000). Drake and colleagues measured objective sleep disruption from caffeine taken up to six hours before bedtime (Drake et al. 2013).

Put the half-life multipliers into the exponential and the practical question answers itself.

Calculation

# 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

A hundred milligrams at two in the afternoon leaves 28.72 milligrams at an eleven o’clock bedtime for a person with no modifier, 12.50 for a smoker and 75.79 for someone taking fluvoxamine, a difference of 6.06 times between the extremes from a 7.50-fold difference in half-life. The modelled sleep latency rises from 18.4 minutes in the fastest state to 40.5 in the slowest, a difference of 22.2 minutes from the same cup of coffee, against 24.1 minutes for the person with no modifier at all.

The last table is the one with an instruction in it. For a person with no modifier the latest hour at which a hundred milligrams keeps the bedtime residual under the threshold is 13:00; for a smoker 17:00; on an oral contraceptive 07:00; on ciprofloxacin 02:00. For late pregnancy and for fluvoxamine no hour in the day works at all. A rule of the form “no caffeine after two” is therefore correct for the median person and wrong by six hours in either direction for a sizeable minority, which is the general shape of advice that quotes a clock time without asking what else someone takes.

The one that clears most other things

The second enzyme matters not because one substance uses it but because so many do. Ito and colleagues set out the equation that predicts how much a competitive inhibitor raises exposure to a substrate, given the fraction of that substrate’s clearance the shared enzyme accounts for (Ito et al. 1998). The equation has a property worth more than any single number it produces.

Calculation

# 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

The exposure multiplier is bounded above by one over one minus the cleared fraction, however strong the inhibitor becomes. A substance cleared 0.30 by the enzyme can never exceed 1.429 times its usual exposure; one cleared 0.99 can reach 100.000. At the illustrative piperine ratio of 2.4 the multipliers are 1.269 for a fraction of 0.30 and 2.742 for 0.90, a ratio of 2.16 between them, and those are 0.888 and 0.274 of their respective ceilings.

That bound is the practical content of the chapter. It says the question to ask about an inhibitor is not how strong it is but what else the person takes that depends on a single route for its clearance. A moderate inhibitor beside a substance cleared almost entirely by the inhibited enzyme is more dangerous than a powerful inhibitor beside a substance with three ways out. And since the inhibitor’s own half-life sets how long the effect lasts, a morning dose of a once-daily inhibitor is still present in the evening: a 2.5-hour half-life takes 10.80 hours to fall to a twentieth of its peak.

Note

Where this becomes a real risk and not an arithmetic exercise. Piperine is sold as an absorption enhancer, which is exactly a description of enzyme inhibition, and the enhancement it provides to one supplement it also provides to every medicine cleared by the same route. For a medicine with a narrow margin between a working dose and a harmful one, a multiplier of 2.742 is not a benefit. Chapter 8 counts how many adults take such a medicine. Anyone taking a prescription medicine should treat an absorption enhancer as a drug interaction and ask a pharmacist, not as a supplement that needs no mention.

Hypothesis · not established

The author’s interpretation (not yet tested): because the exposure multiplier is bounded by the inverse of one minus the cleared fraction, the risk from supplement-grade enzyme inhibitors is concentrated almost entirely in people taking at least one medicine whose cleared fraction through the affected enzyme is 0.90 or more, and is negligible for everyone else. What would show it wrong: if reported adverse events from such supplements were distributed across medicine profiles independently of the cleared fraction of the co-administered medicines, the bound would not be governing the risk.

Summary

The caffeine half-life differs by 7.50 times between the fastest and the slowest clearance state, which turns a 14:00 cup into a bedtime residual differing by 6.06 times and a modelled latency differing by 22.2 minutes; for two of the six states no hour of the day is late enough. The exposure multiplier from inhibiting a shared enzyme is bounded by one over one minus the cleared fraction, giving 1.429 at a fraction of 0.30 and 100.000 at 0.99, with 2.742 at 0.90 under the illustrative inhibitor. Chapter 6 puts the whole cupboard on a clock.

References

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).