Should you stop the antidepressant before surgery?

And why the two-week rule has never been tested — in any operation, by any study design, anywhere in the literature.

Is there a study showing that stopping helps?

This is worth stating plainly because the instruction is given so confidently. Searching for any study — randomised, observational, retrospective, anything — that compared a defined "stop before surgery" group against a "continue" group with a bleeding endpoint returns nothing. A 2025 meta-analysis of ten arthroplasty cohorts covering more than two million patients says so in its own text: no discontinuation-comparison study was identified, and whether stopping is advisable remains uncertain.

What the literature contains instead is dozens of studies comparing surgical patients who take an antidepressant with surgical patients who do not. That comparison carries a problem it cannot solve: depression itself predicts worse surgical outcomes. Any excess found in the exposed group is a mixture of drug effect and illness effect, and separating them requires an analysis most of these studies never ran.

What does the biggest perioperative study actually show?

The anchor study is a retrospective cohort of 530,416 adults across 375 US hospitals, spanning spine, cardiac, vascular, gastrointestinal, gynaecological and orthopaedic surgery. Its headline numbers are the ones usually quoted: in-hospital mortality 1.20, bleeding requiring transfusion 1.09, readmission at 30 days 1.22. All three statistically significant, all three pointing at harm.

Then read the rest of the paper. Among patients taking an SSRI, 41.0% carried a coded depression diagnosis; among those not taking one, 6.2% did. When the authors restricted the comparison to patients with a documented depression diagnosis — comparing depressed patients on a drug with depressed patients off it — the mortality association became 0.86, with a confidence interval from 0.65 to 1.13. It disappeared. The bleeding association did not; it stayed at 1.10.

This is what confounding by indication looks like when a study is honest enough to show it. The mortality signal was the illness. The bleeding signal is probably the drug, and it is a nine-per-cent relative increase on an endpoint that is partly a clinical decision rather than an event.

Does it depend on the operation?

Pooled across seven observational studies of coronary artery bypass grafting, red cell transfusion rose at 1.15 — but reoperation for bleeding was 1.07 with an interval from 0.66 to 1.74, platelet transfusion 0.93, fresh frozen plasma 0.96, and 30-day mortality 1.03. A separate cohort of 132,686 bypass patients found major bleeding at 0.98 and mortality at 0.93, with only transfusion elevated at 1.14.

In hip and knee arthroplasty the picture differs: a meta-analysis of ten cohorts and 2,098,833 patients found transfusion at 1.78, though with heterogeneity of 97%, which is close to uninterpretable. A 2026 propensity-matched analysis of elective colectomy found a bleeding-specific 30-day relative risk of 1.29 and a number needed to harm of 104 — and its authors concluded the data support perioperative awareness rather than routine discontinuation.

Across every setting the same pattern repeats. Transfusion goes up. Reoperation for bleeding and mortality do not. Transfusion is a threshold decision as much as an event, and a surgeon who knows the patient is on an SSRI may reach for it sooner.

What do the societies say?

The Society for Perioperative Assessment and Quality Improvement stated in 2022 that SSRIs, SNRIs, tricyclics and atypical antidepressants should all be taken preoperatively, including on the day of surgery. The American Society of Plastic Surgeons advises that routine discontinuation before surgery in the absence of a careful evaluation should be avoided, reasoning that the risks of stopping in psychologically vulnerable patients likely outweigh any increase in complications.

The European Society of Anaesthesiology and Intensive Care's 2024 preoperative assessment guideline was read in full for the chapter this post draws on. It contains no recommendation on psychiatric medication at all. No antidepressant-specific perioperative statement could be found from the American Society of Anesthesiologists, the American Society of Regional Anesthesia, NICE or the ERAS Society.

So the two-week rule is not a guideline position. It is an inherited habit, and the bodies that have examined it point the other way.

What does stopping cost?

Across 79 studies and 21,002 participants, discontinuation symptoms occurred in 31% after stopping an antidepressant against 17% after stopping placebo. Netting out the placebo effect leaves about 15% with a genuinely drug-attributable syndrome. Severe symptoms occurred in 2.8% against 0.6%. Symptoms typically begin within three days, sooner with short half-life agents.

Two things have not been studied at all: relapse risk over a perioperative window of days to weeks, and whether perioperative cessation affects postoperative delirium, pain scores or opioid requirement. Both are genuine absences rather than negative findings, and both matter for the patient in front of you.

Is there a safer antidepressant to switch to?

A systematic review and meta-analysis examined exactly this recommendation. Patients taking mirtazapine had greater gastrointestinal bleeding risk than patients taking no antidepressant at all, at 1.17 (1.01–1.38). Neither mirtazapine nor bupropion differed from an SSRI. The authors' own conclusion is that it is premature to recommend either agent for patients at bleeding risk.

The transporter-affinity tiers behave no better when tested against outcomes. In a 2026 per-agent meta-analysis, the highest gastrointestinal bleeding estimate belonged to venlafaxine at 1.50 — an intermediate-affinity agent — and the lowest to paroxetine at 1.31, which sits in the top tier. Trazodone, bottom tier, carried an elevated signal in a large primary-care cohort. The tiers predict receptor binding, which is what they were built from. They have not been shown to predict bleeding.

What if they are on an anticoagulant?

In a nested case-control study of 331,305 anticoagulant initiators, adding an SSRI carried an incidence rate ratio of 1.33 for major bleeding. The risk was concentrated early: 1.74 in the first 30 days of combined use, elevated for up to six months, substantially lower after that. Newly combined is the risky state, not long combined.

For calibration, a nationwide cohort of 193,072 anticoagulated patients placed SSRIs and SNRIs at 1.26 for major or clinically relevant bleeding — above low-dose aspirin at 1.14 and NSAIDs at 1.10, below corticosteroids at 1.53 and P2Y12 inhibitors at 1.62. And inside a randomised trial population, adding an SSRI to an anticoagulant produced no significant excess at 1.16 (0.95–1.43).

One agent-specific caution does hold. Fluoxetine and fluvoxamine inhibit CYP2C9, and starting an SSRI raised the risk of an INR of 5 or more by 2.41-fold overall and 3.14-fold for those two agents. That is a reason to avoid those two in a warfarin patient specifically — a pharmacokinetic reason, not a platelet one.

So what do you actually do?

1. List everything acting on haemostasis before touching the antidepressant. NSAID, aspirin, antiplatelet, anticoagulant, corticosteroid. The antidepressant is rarely the largest contributor on that list.
2. Remove the NSAID first. It is the larger multiplier, it usually has an alternative, and stopping it costs nothing psychiatric.
3. If the NSAID has to stay, add acid suppression. Stratified by acid-suppressant use, the antidepressant-plus-NSAID combination ran at 9.1 without and 1.1 with. It is observational, and it is still the best-supported intervention here.
4. Do not stop the antidepressant on bleeding grounds alone. No study supports it, every society with a position advises against it, and discontinuation symptoms occur in about one in seven.
5. Do not switch for bleeding safety. The manoeuvre has been tested and did not work.
6. Watch the first month of any new combination, particularly with an anticoagulant in an older patient.
7. Give the patient the absolute number. Roughly one extra intracerebral bleed per 10,000 people treated for a year; on warfarin, about one extra major haemorrhage per 100 person-years.
8. Write back to the clinic. State that the antidepressant is continuing, that the NSAID has been stopped instead, and why. The surgical team is not being unreasonable — they are following a rule nobody has checked.