Pulmonary Disease · Chapter 2
The Smoking–CYP1A2 Lever
The interaction with no drug in it. How starting, stopping, or being forced off cigarettes swings clozapine and olanzapine by half in either direction — the forced-quit toxicity trap, the third hit that only pulmonary patients get, and the protocol that prevents both.
Bottom Line Up Front
The 30-second version
- It’s the smoke, not the nicotine. Polycyclic aromatic hydrocarbons in combusted tobacco induce CYP1A2. Nicotine does not. Patches, gum, lozenges, and e-cigarettes leave CYP1A2 completely untouched — a patient on full nicotine replacement who has stopped smoking is kinetically a non-smoker. Cannabis smoke induces it too. Smokeless tobacco is the one genuine uncertainty, and new data have unsettled it.
- The magnitude is roughly half, and it runs both ways. Clozapine levels are about a third lower in smokers than non-smokers in meta-analysis. A widely used correction factor is 1.5× in either direction. Quitting pushes toward toxicity; resuming pushes toward relapse. Neither event involves you writing a prescription.
- There is practically no such thing as a light smoker. Induction looks maximal by 2–3 cigarettes a day in men and 4–5 in women. Cutting from two packs to five a day changes nothing kinetically, and a patient who “barely smokes” will rebound just as hard as a heavy one.
- Act in days; watch for weeks. Enzyme activity starts falling within a day or two of the last cigarette. Case-reported toxicity often surfaces over two to four weeks. Cut the dose early and keep watching long — the two windows are not the same window.
- Pulmonary patients get a third hit nobody counts. The respiratory infection that put them in hospital also raises clozapine levels through inflammatory inhibition of CYP1A2 — on top of the forced quit, on top of the sedative stack. Three mechanisms, one somnolent patient, and the chart blames the pneumonia.
The Mechanism: Why the Patch Doesn’t Help
Tobacco smoke is a combustion product, and combustion products contain polycyclic aromatic hydrocarbons. PAHs activate the aryl hydrocarbon receptor, which drives transcription of CYP1A2 — the enzyme responsible for the majority of clozapine’s metabolism. More enzyme means faster clearance means lower levels. Remove the smoke and the induction unwinds; enzyme activity falls back toward the constitutive baseline, and the same milligram now produces a much larger exposure.
Everything clinically useful follows from one fact about that pathway: nicotine is not the inducer. It is a highly addictive alkaloid that happens to travel in the same smoke. That single dissociation explains the entire set of counterintuitive consequences:
| What the patient is using | Induces CYP1A2? | Consequence |
|---|---|---|
| Cigarettes (or any combusted tobacco) | Yes | Induced. Levels suppressed. This is the baseline you titrated against without knowing it. |
| Cannabis, smoked | Yes | Also combustion, also PAHs. A patient who quits cannabis alone can develop clozapine toxicity. Ask about both. |
| Nicotine patch, gum, lozenge, inhaler | No | Protects the cravings. Does nothing for the level. Prescribing NRT does not prevent the rebound — and it makes the quit more likely to succeed, which makes the rebound more likely to happen. |
| E-cigarettes / vaping | No | Vaporized, not combusted. A patient who “switched to vaping” has quit, pharmacokinetically. They will not describe it that way. |
| Chewing tobacco, snus | Unsettled | No combustion, so the PAH mechanism should not apply, and no human study has shown a smokeless product moving a CYP1A2 substrate level. But a 2026 primary-human-hepatocyte study found smokeless tobacco induced CYP1A2 mRNA 12–26-fold — in vitro, with the in vivo translation unknown. Treat as: do not count a switch to chew as continued induction, and do not count it as a clean quit either. Draw a level. |
The history question that fails
- “Do you use tobacco?” is the wrong question. It returns yes for the snus user (not induced) and yes for the pack-a-day smoker (fully induced), and it misses the cannabis smoker entirely.
- Ask instead: “Are you smoking anything — cigarettes or cannabis — and how many a day?” Combustion and count. That is the pharmacokinetic question.
How Much It Actually Moves
The best evidence is Wagner’s 2020 systematic review and meta-analysis — 23 studies, more than 7,000 subjects, the most comprehensive synthesis available. Its headline: clozapine blood levels run roughly a third lower in smokers than in non-smokers (standardised mean difference −0.39, 95% CI −0.55 to −0.22 — “lower by more than a third” is the authors’ own phrase for it), and on cessation they recommend reducing the dose by about 30% with level monitoring. Two caveats travel with that number and rarely get quoted alongside it: heterogeneity was high (I²=80%), and restricting to the six highest-quality studies shrinks the effect to SMD −0.29. This meta-analysis lowered the field’s estimate — earlier reviews had claimed reductions as large as 2.5-fold.
Haslemo’s naturalistic study supplies the dose-response and a working rule — and it reports two different magnitudes that are routinely conflated, including by secondary sources. In 73 patients, the concentration-to-dose ratio was twice as high in non-smokers as in smokers, for both drugs (p<0.01). Separately, absolute serum concentrations were higher in non-smokers by 67% for olanzapine (p<0.01) and by 50% for clozapine — but that clozapine figure did not reach significance (p=0.058). The authors concluded a 50% lower starting dose in non-smokers is rational. De Leon’s widely cited guidance distills all of this to a correction factor of 1.5 for both drugs in either direction.
| Source | Figure | What it actually is — and how much weight to give it |
|---|---|---|
| Wagner 2020 | Blood levels SMD −0.39 (−0.55 to −0.22) — the authors’ own gloss is “lower by more than a third.” C/D ratios SMD −0.70 (−0.84 to −0.56) | Meta-analysis, 23 studies, 7,125 samples. Anchor The basis for the ~30% cut. Caveats worth carrying: I²=80% for the primary outcome, and restricting to the six high-quality studies attenuates it to SMD −0.29. This meta-analysis revised the estimate down — older reviews claimed swings as large as 2.5-fold. |
| Haslemo 2006 | C/D ratio ~2× higher in non-smokers (both drugs, p<0.01). Absolute levels higher by 67% olanzapine (p<0.01), 50% clozapine (p=0.058, NS). Max induction at 7–12 cig/day; no subjects in the 1–6/day band | Naturalistic, n=73. Strong The dose-response ceiling is the durable contribution. Do not quote the 50/67 figures as C/D ratios — that conflation is widespread and wrong. |
| De Leon 2004 | Correction factor 1.5× | Expert synthesis and a bedside heuristic, explicitly described as a rough approximation. Heuristic Use to anticipate, never to replace a level. |
| Olanzapine magnitude | Ranges from ~23% higher clearance to several-fold C:D differences | Unsettled The direction is certain; the multiplier is not. Anyone quoting one precise olanzapine number is over-reading their source. Chapter 5 adjudicates it. |
Why two numbers, and which one predicts your patient
Notice that in both studies the dose-corrected gap is bigger than the absolute-level gap: Haslemo found a ~2× C/D difference but only a 50–67% difference in raw levels; Wagner found C/D SMD −0.70 against blood-level SMD −0.39. That is not a contradiction — it is the fingerprint of clinical practice. Smokers are already on higher doses, because someone titrated them up against a suppressed level. Cross-sectional raw levels therefore understate the enzyme effect; the C/D ratio isolates it. Flanagan’s audit shows both halves at once and settles it: non-smokers were prescribed only 75–80% of the smokers’ doses, yet still attained higher median clozapine and norclozapine concentrations. Lower dose, higher level — that is the enzyme, visible in 231,707 samples.
Which matters at the bedside, because when your patient quits, the dose is fixed. The variable that moves is clearance — so the C/D gap, not the raw-level gap, is what predicts the rise. That is a reason to treat the ~30% cut as a floor rather than an answer, and it is exactly why Wagner’s recommendation says “informed by steady-state trough levels.” Cut early, then measure.
Three cigarettes is a fully induced man
- Practically, there is no such thing as a light smoker on clozapine. Two or three a day — the social smoker, the patient who bums one at break, the one who swears they “barely smoke” — is kinetically indistinguishable from a two-pack-a-day patient, and will fall the same distance when they stop.
- The threshold is sex-dependent, and that is new. Men appear maximally induced by 2–3 a day; women by around 4–5. Women already attain higher dose-adjusted clozapine concentrations than men, so the two effects compound rather than cancel. Nothing else in this course carries a sex-specific number.
- “Cutting down” is not a kinetic intervention. A patient who goes from 40 a day to 5 has done something genuinely good for their lungs and nothing whatever for their clozapine level. Do not adjust the dose for it, and do not let them believe they have half-quit.
- Flanagan also proposes that the optimum predose target may differ by smoking status — roughly 0.35–0.45 mg/L in smokers versus 0.50–0.60 mg/L in non-smokers. One TDM audit’s proposal, not an established range.
The dose-response finding deserves its own line, because it is the one that changes what you do on Monday. Haslemo put the ceiling at roughly 7–12 cigarettes daily, with no significant difference between the 7–12, 13–19 and ≥20 per day subgroups — but that study enrolled nobody smoking 1–6 a day, so it could never say where induction begins. Flanagan’s 2023 audit answered the question, and the answer is lower than anyone was teaching. Across 231,707 TDM samples, with cigarette counts available for 12,842 male and 3,948 female samples — including 574 male and 253 female samples from people smoking 1–9 a day, precisely the band Haslemo lacked — the effect of smoking on dose and on plasma clozapine and norclozapine appeared maximal after 2–3 cigarettes daily in men, “perhaps fewer,” and near-maximal after some 4–5 a day in women.
Ground the numbers in what a level means. Response in treatment-resistant schizophrenia is generally associated with a trough at or above roughly 350 ng/mL; adverse effects become more likely above roughly 600 ng/mL; and seizure risk rises above roughly 1,000 ng/mL. Treat those as gradients, not gates — there is wide inter- and intra-individual variation, seizures are reported at lower levels, and there is no level that is categorically safe. A 1.5× swing on a patient sitting at 700 ng/mL lands them past 1,000. That arithmetic is the chapter.
Three Clocks, Not One
Almost every error in managing this interaction is a timing error, and it comes from collapsing three different clocks into one. They run at different speeds, and the literature genuinely disagrees about the middle one.
| Clock | Speed | Evidence — and what it means for you |
|---|---|---|
| 1. Enzyme activity how fast induction unwinds |
Days | Faber and Fuhr phenotyped 12 heavy smokers (≥20/day) through a sudden quit. Caffeine clearance fell by 36.1% (95% CI 30.9–42.2), and the apparent half-life of the CYP1A2 activity decrease was 38.6 hours (27.4–54.4). This is direct measurement of the enzyme — via a caffeine probe, not clozapine — and it is fast. Their conclusion: doses of narrow-therapeutic-range CYP1A2 substrates should be decreased immediately on cessation. |
| 2. The drug level time to a new steady state |
~5–7 days | Follows from the enzyme clock plus clozapine’s own half-life. By the end of the first week, the level has largely arrived at its non-smoking value. Your patient is somewhere new before anyone has looked. |
| 3. Clinical toxicity when it announces itself |
2–4 weeks | The case-report literature, and de Leon’s synthesis of it, repeatedly describes toxicity emerging over two to four weeks — and infers from that a slower de-induction than Faber and Fuhr measured. The two are in genuine tension. |
Resolving the tension honestly
Direct phenotyping says de-induction is a matter of days. Case reports say toxicity shows up at two to four weeks. Both can be true: the enzyme recovers quickly, and the clinical consequence of the resulting exposure — accumulating sedation, a seizure, an aspiration event — declares itself later, is noticed later, and gets attributed later. A case report is dated from the symptom, not from the kinetics.
You do not need to resolve it, because both readings give the same instruction: act in days, watch for weeks. What you must not do is take the two-to-four-week figure as permission to wait two to four weeks. That is the single most common way this goes wrong — a clinician who knows the interaction, plans to act on it, and schedules the dose review for after the level has already peaked.
Faber and Fuhr turned their finding into the most useful rule of thumb in this chapter: a stepwise daily dose reduction of approximately 10% until the fourth day after cessation, accompanied by therapeutic drug monitoring. It lands you near the 30% total reduction that Wagner’s meta-analysis independently recommends — two different methods converging on the same answer, which is about as much confidence as this literature offers.
The Forced-Quit Trap
Here is what makes this a pulmonary chapter rather than a general-psychiatry footnote. The three ways a patient stops smoking are not equally visible, and the most dangerous one is the one nobody documents.
| How they stopped | Do you find out? | The risk |
|---|---|---|
| Planned quit attempt | Usually They tell you, or the cessation agent appears on the list | Manageable. This is the version the guidelines are written for — and the rarest. |
| Illness-related too sick to smoke | Rarely Nobody codes “stopped smoking because breathless” | Silent. And it coincides with the illness that is also raising the level — see below. |
| Forced by policy admission to a smoke-free hospital or facility, incarceration | Almost never It is an administrative event, not a clinical one | The trap. No order is written. No note records it. The medication reconciliation captures every drug and misses the one variable that just changed. |
The published cases are not subtle. Bondolfi and colleagues reported two patients genotyped for CYP1A2: in the first, stopping smoking produced severe sedation and fatigue within two weeks alongside an approximately three-fold rise in plasma clozapine; in the second, a level of 3,004 ng/mL was measured days after a 16-day stay in a general hospital where smoking was prohibited. Both carried the CYP1A2*1F AA genotype associated with high inducibility — and in the second case a concurrent switch from omeprazole to pantoprazole may have contributed, which is exactly the kind of honest caveat that makes the case more instructive, not less. Zullino and colleagues made the same point for cannabis: tobacco and cannabis cessation can both precipitate clozapine or olanzapine intoxication.
Now the mirror image, which is the half almost nobody connects. Discharge is also a dosing event. The patient goes home, resumes smoking, CYP1A2 re-induces, and the level falls back by a third or more — often from a dose you had just carefully reduced, which makes the fall larger than it would have been. Three to four weeks later they relapse. That gets written up as non-adherence or loss of response. It is a pharmacokinetic event with a lighter, and it was entirely predictable on the day you signed the discharge summary.
The Third Hit: Infection Raises the Level Too
This is the part that belongs to this course and to no other, and it is the reason a pulmonary admission is uniquely dangerous for a patient on clozapine.
Systemic inflammation releases cytokines that inhibit CYP1A2. Clark and colleagues systematically reviewed the published experience: 40 cases across 23 publications of elevated clozapine levels associated with infection — and the infections were commonly respiratory in origin. The typical adverse event was sedation. De Leon and Diaz made the same argument from a case of serious respiratory infection years earlier.
So count what actually lands on a patient with schizophrenia admitted with pneumonia or a COPD exacerbation:
| Input | Direction | Mechanism |
|---|---|---|
| 1. Forced smoking cessation | Level ↑ ~1/3, over days | Loss of PAH-driven CYP1A2 induction. |
| 2. The infection itself | Level ↑ magnitude unpredictable | Inflammatory cytokines inhibit CYP1A2. Independent of the first, and additive to it. |
| 3. The sedative stack | Reserve ↓ | Lorazepam for the anxiety, a Z-drug for the steroid insomnia, an opioid for the air hunger — Chapter 1’s Axis 1, arriving in the same 72 hours. |
| The result: a somnolent, hypersalivating, unsteady patient on day 3 to 5. The chart says “delirium, pneumonia not improving.” Three mechanisms summed. None named. The clozapine dose was never changed — which is precisely the problem. | ||
The two axes may not be independent — and if so, the admitted smoker is the worst case
Chapter 1 taught the axes as separate mechanisms that happen to co-fire. A 2025 study suggests they interact. Smith and colleagues measured CRP against dose-adjusted clozapine in 126 patients and found the correlation held in non-smokers only (R=0.492, p<0.001) and not in smokers (R=0.191, p=0.166). Among non-smokers, dose-adjusted clozapine rose 48% at CRP 5–50 and 204% at CRP above 50, versus CRP under 5. Among smokers, no significant difference across CRP bands at all.
The natural reading — and this is inference, not the authors’ claim — is that PAH induction is strong enough to swamp cytokine inhibition, so inflammation cannot move a level that smoking is already holding down. If that is right, then Axis 2 does not merely add to the third hit; it gates it. Your admitted smoker stops smoking on day 0, de-induces across days 1–4, and thereby converts into exactly the phenotype in which inflammation drives levels up threefold — while the pneumonia that admitted them supplies the CRP. The forced quit doesn’t just raise the level; it unlocks the second mechanism. Treat this as a hypothesis worth holding, not a finding to quote: one cross-sectional study, 126 patients, and the mechanism is my inference from it.
And the loop closes on itself, which is worth saying out loud: clozapine is itself associated with higher rates of pneumonia and worse outcomes from it than other second-generation antipsychotics. Sedation, hypersalivation, and impaired swallow feed aspiration; aspiration feeds pneumonia; pneumonia raises the level; the higher level deepens the sedation. Clozapine raises the risk of the illness that raises clozapine.
The evidence grade just moved — and so did the advice
- Clark’s 2018 review noted that no cohort, case-control, or cross-sectional data existed on this phenomenon. That was true in 2018 and is no longer true. Espnes and colleagues measured CRP and alpha-1-acid glycoprotein across 1,106 routine clozapine TDM samples and found a clear association with dose-adjusted clozapine concentration: a rise in CRP from 5 to 60 mg/L would be expected to more than double it. This is no longer a pile of case reports.
- But the mechanism is not only enzyme inhibition, and that changes what the level means. Clozapine binds heavily to alpha-1-acid glycoprotein — an acute-phase protein that rises in inflammation. So part of the measured increase is bound drug. Total concentration climbs; whether free drug climbs proportionally is unresolved. Espnes’s own conclusion is pointed: during infection the clozapine concentration may be unreliable, and CRP plus clinical signs should guide you rather than the number alone.
- Which lands somewhere more useful than “check a level.” Check a CRP alongside it, and treat the pair as the reading. A high clozapine level with a CRP of 90 in a comfortable patient is a different object from the same level with a normal CRP and a drowsy one. Patients do get toxic during infection — Clark’s cases were mostly sedation, and they were real — so this is not a reason to dismiss a high level. It is a reason not to reflexively slash the dose off a number that partly reflects protein binding.
Caffeine, Cannabis, and the Rest of the Substrate List
Two more levers move in the same window, and one of them moves because the patient quit smoking.
Caffeine is the hidden accomplice. It is both a CYP1A2 substrate and a competitive inhibitor. Smokers clear caffeine faster, so they drink more of it — and when they stop smoking, caffeine clearance drops with everything else. Two things follow. First, their caffeine levels rise on an unchanged coffee habit, producing jitteriness, tremor, insomnia, and anxiety that look exactly like nicotine withdrawal and get treated as such. Second, the higher caffeine load inhibits CYP1A2 further, pushing clozapine up again on top of the de-induction. De Leon proposed a caffeine correction factor of 0.6 for patients who begin consuming significant quantities. Controlled volunteer studies show modest effects at ordinary doses — single-digit-to-twenty-percent range — so the clean summary is: at usual intake, minor; at heavy intake, and stacked on a forced quit, not minor. Advising a quitting patient to cut caffeine roughly in half is a free intervention that solves two problems at once.
Cannabis is the same mechanism as tobacco: combustion, PAHs, induction. It is not in most drug-interaction checkers, patients do not volunteer it, and a cannabis quit produces the same rebound as a tobacco quit.
Which drugs actually care:
| Tier | Agents |
|---|---|
| Act on it narrow window, real consequences |
Clozapine — the whole reason this chapter exists: heavily CYP1A2-dependent, narrow therapeutic range, dose-related seizures and sedation, and a level you can actually order. Olanzapine — same lever, wider window, so the consequence is more often loss of efficacy on resumption than toxicity on cessation. |
| Know about it real but wider margin |
Fluvoxamine (a CYP1A2 substrate and a potent inhibitor of it — which is why it sets up the theophylline trap in Chapter 4, and why adding it to clozapine is its own event). Duloxetine and mirtazapine. Theophylline, which is not a psychotropic but sits on your pulmonary patient’s list and has a narrower window than any of them. |
| Largely exempt not CYP1A2 substrates |
Quetiapine, risperidone, aripiprazole and the SSRIs other than fluvoxamine. When the smoking lever is the dominant problem and the antipsychotic is negotiable, moving off the lever is often the cleanest fix available — but remember from Chapter 1 that escaping Axis 2 is not escaping Axis 1. Quetiapine is kinetically clean and pharmacodynamically sedating. |
The Protocol
Everything above collapses into a short operational sequence. The hard part is not the pharmacology — it is noticing that the event happened.
| When | On cessation (level rising) | On resumption (level falling) |
|---|---|---|
| Before at every visit |
Record combustion and count — cigarettes and cannabis, number per day — in the medication section, not the social history. Tell the patient to notify you before any change. This one habit prevents most of what follows. | |
| Day 0 | Draw a baseline level if you can. Begin the reduction now — not at the two-week review. Faber and Fuhr’s rule: approximately 10% of the daily dose per day, through day 4, landing near a 30% total cut. | Do not pre-emptively raise. Confirm the resumption is real and sustained first — patients over-report and under-report in both directions. |
| Days 1–7 | Complete the stepwise reduction. Watch for sedation, hypersalivation, unsteadiness, myoclonus. Level at the end of the first week, when the new steady state has largely arrived. | Watch for symptom re-emergence. Level rather than assume. |
| Weeks 2–4 | The vigilance window. This is when case-reported toxicity surfaces. Re-level. Do not interpret an uneventful week one as the all-clear. | The relapse window. A patient who deteriorates three to four weeks post-discharge gets a level before they get a diagnosis of non-adherence. |
| Also | Halve the caffeine on cessation. Ask about cannabis separately. If a respiratory infection is in the picture, lower your threshold for a level further — the third hit is additive and it is not in any interaction checker. | |
One caution about the arithmetic, and it matters: the 1.5× factor and the 30% cut are starting positions, not answers. Inter-individual variability is wide, CYP1A2 inducibility is partly genotype-dependent, and the same patient varies over time. The factor tells you which way to move and roughly how far. The level tells you where you landed. On this axis, unlike almost everything else in this course, you have a real measurement — use it.
The Action Ladder
Stable outpatient, smoking status unchanged
Document combustion and count as a medication variable. Counsel the patient to tell you before any change — and tell them explicitly that nicotine replacement does not cover this. Baseline level if they are on clozapine and you have never drawn one.
Planned quit attempt
The best-case scenario, and still an active intervention. Start the stepwise ~10%/day reduction from day 0 through day 4. Level at the end of week 1 and again at weeks 2–4. Halve the caffeine. Do not let a successful quit — which is a genuine clinical win — become an ICU admission.
Admission to a smoke-free facility
Treat day 0 as a medication event and write it down as one, because nobody else will. Begin the reduction immediately rather than waiting for sedation to announce it. If the admission is for a respiratory infection, expect the third hit on top and level early.
Discharge
The reduction you made in hospital is now wrong. If the patient will resume smoking — and most do — plan the re-titration, level at two to four weeks, and put the plan in the discharge summary in words the outpatient prescriber will actually see. This handoff is where the relapses are made.
Clinical Pearls
Pearls
- The patch is not the cigarette. PAHs induce; nicotine does not. Full nicotine replacement leaves the patient kinetically a non-smoker. This is the fact most clinicians get backwards, and it is the whole chapter in one line.
- Three cigarettes a day is a fully induced man; four to five, a fully induced woman. There is practically no such thing as a light smoker on clozapine. “I only smoke a couple” is not reassurance — it is a fully induced enzyme that will rebound exactly as hard as a pack-a-day habit.
- Act in days, watch for weeks. The enzyme clock runs in days — a measured CYP1A2 activity half-life of about 39 hours. The toxicity clock runs in weeks. Never use the second to justify delaying on the first.
- Ask about combustion, not tobacco. “Do you use tobacco?” catches the snus user who is not induced and misses the cannabis smoker who is. Ask what they smoke, and how many.
- Vaping is quitting. A patient who switched to e-cigarettes has stopped inducing CYP1A2 and will never describe themselves as having quit. Neither will the patient who switched to chew.
- A 1.5× swing on 700 ng/mL is above 1,000. The magnitude only matters relative to where the patient already sits. Know the level before the smoking status changes, not after.
- Coffee is part of this. Quitting smoking raises caffeine levels, which mimics nicotine withdrawal and inhibits CYP1A2, pushing clozapine higher still. Halve the caffeine at the quit. Free intervention, two problems.
- Pneumonia raises the level, and clozapine raises the pneumonia. Inflammatory cytokines inhibit CYP1A2; sedation and hypersalivation feed aspiration. A closed loop, and the only course in this series where you meet both halves of it.
- Discharge is a dosing event. Relapse three to four weeks after discharge in a resumed smoker is a kinetic event until a level proves otherwise — not non-adherence.
Red Flags — Stop and Reassess
Draw a level and act
- Any patient on clozapine or olanzapine admitted to a smoke-free facility — regardless of the admitting diagnosis, and regardless of whether nicotine replacement was ordered. Day 0 is the action point, not day 14.
- New sedation, hypersalivation, unsteadiness, myoclonus, or slurred speech in a clozapine patient whose smoking has changed — myoclonus and speech difficulty in particular can precede a seizure.
- A clozapine patient with a respiratory infection who becomes drowsy — assume the level is rising from inflammation as well as any change in smoking, and confirm rather than attribute it to the illness.
- A patient who reports switching to vaping, chewing tobacco, or snus — they have stopped combusting without saying the word “quit,” and the level is already moving. For vaping the direction is clear. For smokeless tobacco it is not, which is a reason to measure rather than predict.
- A cannabis user who has stopped, whether or not they smoke tobacco — same mechanism, same rebound, and it will not appear in any interaction checker.
- Deterioration two to four weeks after discharge in a smoker — level before you conclude non-adherence or loss of response.
Patient Counseling Script
Plain-language script
“There’s something about this medication that surprises almost everyone, and I need you to remember it. Cigarette smoke changes how fast your body clears this drug — not the nicotine, the smoke itself. So if you stop smoking, this medication builds up in your blood even though you’re taking exactly the same dose, and it can build up enough to make you very drowsy or even cause a seizure. Now, I want you to quit — it’s the single best thing you can do for yourself. I just need to know before you do, so I can lower the dose and check a blood level as you go. And this is important: the patch, the gum, and vaping don’t protect you from this. Only the smoke matters. If you switch to vaping, that counts as quitting for this purpose — tell me. Same with cannabis, if you smoke it. It works the same way. It also runs in reverse: if you start smoking again after a hospital stay, the medicine may stop working and you could get unwell again, so tell me that too. And when you do quit, cut your coffee roughly in half — quitting makes caffeine hit harder, and people mistake that for withdrawal.”
EMR / Documentation Template
References
- Wagner E, McMahon L, Falkai P, Hasan A, Siskind D. Impact of smoking behaviour on clozapine blood levels — a systematic review and meta-analysis. Acta Psychiatr Scand. 2020;142(6):456–466. (23 studies, >7,000 subjects; clozapine levels approximately one-third lower in smokers, with individual studies ranging roughly 20–50%; ~30% dose reduction plus monitoring recommended on cessation.)
- Haslemo T, Eikeseth PH, Tanum L, Molden E, Refsum H. The effect of variable cigarette consumption on the interaction with clozapine and olanzapine. Eur J Clin Pharmacol. 2006;62(12):1049–1053. (n=73. The concentration-to-dose ratio was twice as high in non-smokers for both drugs, p<0.01, and did not differ significantly between the 7–12, 13–19 and ≥20 cigarette/day subgroups. Separately, absolute serum concentrations were higher in non-smokers by 67% for olanzapine (p<0.01) and 50% for clozapine (p=0.058, not significant) — these are not C/D ratios, a conflation common in secondary sources. 7–12 cigarettes daily probably sufficient for maximum induction; 50% lower starting dose in non-smokers. No subjects were enrolled in the 1–6/day band.)
- Faber MS, Fuhr U. Time response of cytochrome P450 1A2 activity on cessation of heavy smoking. Clin Pharmacol Ther. 2004;76(2):178–184. (12 heavy smokers, caffeine phenotyping; clearance fell 36.1% [30.9–42.2]; apparent half-life of CYP1A2 activity decrease 38.6 hours [27.4–54.4]; proposes stepwise ~10% daily dose reduction through day 4 with therapeutic drug monitoring.)
- de Leon J. Atypical antipsychotic dosing: the effect of smoking and caffeine. Psychiatr Serv. 2004;55(5):491–493. (Smoking correction factor of 1.5 for clozapine and olanzapine, explicitly a rough approximation; caffeine correction factor of 0.6; induction effects described as emerging and resolving over two to four weeks; inductive effects also expected from cannabis smoking.)
- Bondolfi G, Morel F, Crettol S, Rachid F, Baumann P, Eap CB. Increased clozapine plasma concentrations and side effects induced by smoking cessation in 2 CYP1A2 genotyped patients. Ther Drug Monit. 2005;27(4):539–543. (Approximately three-fold rise with severe sedation within two weeks; a level of 3,004 ng/mL following a 16-day smoke-free hospital stay, with a concurrent omeprazole-to-pantoprazole switch noted as a possible contributor; both patients CYP1A2*1F AA.)
- Zullino DF, Delessert D, Eap CB, Preisig M, Baumann P. Tobacco and cannabis smoking cessation can lead to intoxication with clozapine or olanzapine. Int Clin Psychopharmacol. 2002;17(3):141–143.
- Clark SR, Warren NS, Kim G, et al. Elevated clozapine levels associated with infection: a systematic review. Schizophr Res. 2018;192:50–56. (40 cases across 23 publications; infections commonly respiratory in origin; sedation the typical adverse event. Case reports and small series only — no cohort, case-control, or cross-sectional data exist on this phenomenon.)
- Flanagan RJ, Hunter S, Obee SJ. Effect of cigarette smoking on clozapine dose and on plasma clozapine and N-desmethylclozapine (norclozapine) concentrations in clinical practice. J Clin Psychopharmacol. 2023;43(6):514–519. (231,707 TDM samples 1996–2017; cigarette counts for 12,842 male and 3,948 female samples, of which 574 and 253 respectively reported 1–9/day. Effect on dose and on plasma clozapine/norclozapine appeared maximal after 2–3 cigarettes daily in males, “perhaps fewer,” and near-maximal after ~4–5/day in females. Non-smokers were prescribed 75–80% of smokers’ doses yet attained higher median concentrations. Proposes different predose targets by smoking status: 0.35–0.45 mg/L smokers vs 0.50–0.60 mg/L non-smokers.)
- Espnes KA, Spigset O, Bratt GE, et al. Association between levels of the acute phase proteins alpha-1-acid glycoprotein and C-reactive protein and serum concentrations of clozapine: a study of 1106 therapeutic drug monitoring samples. Basic Clin Pharmacol Toxicol. 2026;138(3):e70208. (Clear association between higher AGP/CRP and higher dose-adjusted clozapine; a CRP rise from 5 to 60 mg/L would be expected to more than double it. Authors recommend monitoring CRP rather than AGP, and caution that the clozapine concentration may be unreliable during inflammation — CRP and clinical signs should guide dosing.)
- Smith RL, et al. Correlation between clozapine and CRP levels in relation to smoking status. Acta Psychiatr Scand. 2025. (n=126, 47% smokers. Dose-adjusted clozapine correlated with CRP in non-smokers only, R=0.492 p<0.001, vs smokers R=0.191 p=0.166. In non-smokers, dose-adjusted clozapine +48% at CRP 5–50 and +204% at CRP>50 vs CRP<5; no significant difference across CRP bands in smokers.)
- de Leon J, Diaz FJ. Serious respiratory infections can increase clozapine levels and contribute to side effects: a case report. Prog Neuropsychopharmacol Biol Psychiatry. 2003;27(6):1059–1063.
- Schoretsanitis G, Kane JM, Ruan CJ, Spina E, Hiemke C, de Leon J. A comprehensive review of the clinical utility of and a combined analysis of the clozapine/norclozapine ratio in therapeutic drug monitoring for adult patients. Expert Rev Clin Pharmacol. 2019;12(7):603–621.
- Differential induction of hepatic CYP enzymes by tobacco products and e-cigarettes: recommendation to rethink smoking status in clinical pharmacology. Arch Toxicol. 2026. doi:10.1007/s00204-026-04428-6. (Primary human hepatocytes exposed to 6 cigarette brands, a heated tobacco product, 6 cigar brands, 2 smokeless tobacco brands and 3 e-cigarette brands: smokeless tobacco induced CYP1A2 mRNA 12–26-fold; e-cigarettes induced CYP3A4 and CYP2B6 far more than CYP1A1/1A2. In vitro mRNA induction only — no in vivo human pharmacokinetic data yet link a smokeless product to a change in a CYP1A2 substrate level, and the clinical translation is unknown.)
Last reviewed July 2026. Part of the Psychiatry Education Forum Academy; for clinician education — it supports, and does not replace, individual clinical judgment and current local protocols.
That was the axis with a number. The other one is harder.
Axis 2 gives you a level to check and a protocol to run. Axis 1 gives you neither — just a stack of individually reasonable sedatives and a patient with no reserve. The member chapters start where the measurement runs out: the additive-burden math, and then the drug-by-drug calls.
Educational use only. Refer to the sources cited above and current prescribing information for clinical decisions. Psychiatry Education Forum and authors assume no liability for use of this material.
Responses