Tapering Psychotropics in Pregnancy: Risk-Benefit Assessment Frameworks
Pregnancy forces a binary that rarely has a clean answer: continue a psychotropic with known or uncertain fetal exposure, or taper it and accept the risk of maternal relapse. The reflexive instinct to discontinue "to be safe" is frequently the higher-risk decision, because untreated maternal psychiatric illness carries its own measurable obstetric and neonatal harms. This article sets out a structured risk-benefit framework for the prescriber managing antidepressants, mood stabilizers, antipsychotics, and benzodiazepines across the perinatal period.
The central error: treating discontinuation as the null hypothesis
The most common clinical mistake in perinatal psychiatry is framing medication continuation as the action requiring justification while treating discontinuation as the safe default. It is not. Abrupt or poorly timed cessation of psychotropics in a patient with recurrent major depression, bipolar disorder, or a psychotic disorder produces relapse rates that are both high and consequential.
The evidence is strongest in bipolar disorder. In the prospective study by Viguera and colleagues (2007, American Journal of Psychiatry), women with bipolar disorder who discontinued mood stabilizers around conception had a recurrence rate of approximately 85.5% during pregnancy, compared with 37% in those who continued. Median time to first recurrence after discontinuation was less than 2 weeks for abrupt cessation. For major depressive disorder the relapse signal is weaker but real, and it concentrates in women with recurrent illness and recent symptomatic episodes.
Untreated antenatal depression and anxiety are not benign. They are associated with preterm birth, low birth weight, reduced fetal growth, impaired maternal-infant attachment, and elevated risk of postpartum depression and suicide — the latter a leading cause of maternal death in the year after delivery per UK MBRRACE data and corresponding US maternal mortality reviews. The risk-benefit calculation must therefore weigh teratogen exposure against the documented harms of the underlying disorder, not against a hypothetical risk-free alternative.
A four-axis assessment framework
Before deciding to taper, continue, or switch, the prescriber should explicitly characterize four variables. Each shifts the threshold for continuation.
| Axis | Continue medication when… | Consider tapering when… | ||---| | Illness severity and recurrence | Recurrent MDD (≥3 episodes), bipolar I, psychotic disorder, prior severe postpartum episode | Single past episode, long remission, mild illness, situational presentation | | Time since last episode | Symptomatic within 6–12 months, or relapsed on prior taper | Sustained euthymia >12–24 months | | Drug teratogenic profile | SSRI (excluding paroxetine), lamotrigine, most SGAs | Valproate, carbamazepine, paroxetine, high-dose benzodiazepines | | Gestational timing at presentation | Already past organogenesis (>10–12 weeks); exposure has occurred | Pre-conception planning window |
A patient with bipolar I disorder maintained on lamotrigine who presents at 9 weeks' gestation sits firmly in the "continue" column on three of four axes. A patient with a single remote depressive episode, euthymic for 3 years on sertraline, planning conception, is a reasonable candidate for a pre-pregnancy taper trial. Most patients fall between these poles, and the framework's value is in making the trade-offs explicit rather than reflexive.
If tapering is chosen: do it before conception, not during
The single most important principle of perinatal tapering is timing. If discontinuation is appropriate, attempt it before conception — ideally completing the taper and observing a stability window of at least 3 months before attempting pregnancy. Tapering during the first trimester is the worst of both worlds: the fetus has already been exposed during organogenesis, and withdrawal-driven instability compounds the early-pregnancy period when relapse risk is highest.
Use a hyperbolic taper. The pharmacological rationale developed by Horowitz and Taylor (2019, Lancet Psychiatry) applies with equal force in pregnancy: because receptor occupancy is a hyperbolic, not linear, function of dose, reductions should be proportional to the current dose and the final steps should be the smallest. For an SSRI or SNRI, reduce by roughly 5–10% of the current dose every 2–4 weeks once below the minimally effective dose, monitoring for both withdrawal and re-emergent illness. The Maudsley Deprescribing Guidelines provide drug-specific hyperbolic regimens that translate directly to the preconception setting.
If a patient is already pregnant and stable on medication, abrupt discontinuation is rarely justified by the exposure that has already occurred. The exception is an established human teratogen where ongoing exposure changes outcomes (valproate, discussed below).
Drug-class specifics
SSRIs and SNRIs
For most SSRIs the absolute teratogenic risk is low. Large cohort and registry analyses, including FDA labeling and NICE antenatal mental health guidance (CG192), do not support a clinically significant increase in major congenital malformations for sertraline, citalopram, escitalopram, or fluoxetine. Paroxetine is the exception: it carries a small but replicated signal for cardiac septal defects (roughly 1.5–2% absolute risk vs. ~1% baseline) and is the one SSRI for which a pre-pregnancy switch or taper is specifically favored.
Two third-trimester considerations recur in clinic. Poor neonatal adaptation syndrome — jitteriness, respiratory distress, feeding difficulty — occurs in up to 30% of exposed neonates, is self-limiting over days, and is not a reason to taper before delivery; doing so trades a transient, manageable neonatal syndrome for a peripartum maternal relapse at the highest-risk moment. The persistent pulmonary hypertension of the newborn (PPHN) signal is small in absolute terms (estimated 1–3 per 1,000 vs. ~1.2 per 1,000 baseline). Neither finding justifies routine late-pregnancy discontinuation.
Sertraline is generally the preferred agent for initiation or switching in pregnancy and lactation given its exposure data and low milk transfer.
Mood stabilizers
This is where teratogenicity is dose-dependent and class-defining.
| Agent | Major malformation risk | Clinical position in pregnancy |
|---|---|---|
| Valproate | 10–11% overall; dose-dependent neural tube defects; mean IQ reduction ~7–10 points; ~30–40% neurodevelopmental disorder risk | Contraindicated in pregnancy and in any person of childbearing potential without a pregnancy-prevention program (MHRA, FDA) |
| Carbamazepine | ~2.6–5.5%; neural tube defects | Avoid; switch preconception where feasible |
| Lamotrigine | ~2–3%, near background at doses <300 mg/day | Preferred mood stabilizer in pregnancy; monitor levels (clearance rises markedly) |
| Lithium | Ebstein anomaly ~0.6–1% (RR elevated but absolute risk low) | Acceptable when needed; level-driven dosing, fetal echocardiography |
Valproate is the one psychotropic where the framework tilts hardest toward avoidance: per the MHRA pregnancy-prevention program and FDA labeling, it should not be used in pregnancy for bipolar disorder, and exposure already incurred warrants specialist counseling and high-dose folate. For a patient with bipolar disorder stable on lamotrigine, the dominant management issue is not teratogenicity but pharmacokinetics: lamotrigine clearance increases substantially across gestation (estrogen-induced glucuronidation), levels can fall by 50% or more, and breakthrough episodes follow if the dose is not tracked and uptitrated against serum levels — then promptly reduced postpartum to avoid toxicity.
Lithium is not the absolute contraindication it was once considered. The Ebstein anomaly risk, while a real relative increase, is small in absolute terms. Where lithium is the only agent that has achieved stability, continuation with serum-level monitoring (checked more frequently as volume of distribution and GFR shift), a switch toward divided dosing, fetal echocardiography at 16–20 weeks, and careful peripartum management of rapidly changing renal clearance is generally preferable to switching to a less effective agent.
Antipsychotics
Second-generation antipsychotics as a class do not show a strong major-malformation signal in current registry data, and abrupt discontinuation in a patient with schizophrenia or bipolar I risks relapse with poor adherence to antenatal care. Olanzapine and quetiapine have the largest reassuring datasets. The principal caveats are metabolic: SGAs increase gestational diabetes risk, mandating early glucose tolerance screening, and neonatal extrapyramidal/withdrawal signs can occur. Continuation is the default for psychotic and bipolar I disorders; the decision is which agent, not whether.
Benzodiazepines and z-drugs
Chronic benzodiazepine use is the class most appropriately targeted for tapering before or during pregnancy, but not abruptly. The first-trimester oral cleft signal is small and inconsistent; the more robust concerns are late-pregnancy "floppy infant syndrome" and neonatal withdrawal. The clinical aim is to reach the lowest effective dose or discontinue using a slow, Ashton-Manual-style taper (typically via conversion to a long-half-life agent such as diazepam, then reductions of ~5–10% of the diazepam-equivalent dose every 1–2 weeks, slowing near the end). Abrupt benzodiazepine withdrawal carries seizure risk and is never appropriate in pregnancy. Where an anxiety disorder is driving use, transition the patient onto an SSRI and taper the benzodiazepine underneath it.
Shared decision-making and documentation
Perinatal medication decisions are preference-sensitive and should be documented as such. The prescriber should record: the specific malformation and neurodevelopmental risks discussed (in absolute terms, not relative), the risks of untreated illness, the patient's stated values, and the agreed monitoring plan. NICE CG192 frames this as a structured conversation rather than a unilateral medical decision, and good documentation is also the prescriber's best medicolegal protection.
Quote absolute risks, not relative ones. "Sertraline does not meaningfully raise the roughly 1-in-100 background risk of major malformation" is interpretable; "a 1.2-fold relative risk" is not, and it reliably frightens patients into unsupervised discontinuation.
Clinical pearls
- Default to continuation in recurrent or severe illness. Bipolar I, recurrent MDD, psychotic disorders, and any history of severe postpartum episodes weight strongly toward maintaining an effective regimen. Relapse is not a soft outcome.
- If you taper, do it preconception with a hyperbolic schedule and a stability window of at least 3 months before attempting pregnancy. First-trimester tapering combines completed exposure with peak relapse risk.
- Valproate is the line in the sand. Avoid in any person of childbearing potential without a pregnancy-prevention program; it is the one psychotropic where the data mandate near-categorical avoidance.
- Track lamotrigine and lithium levels across gestation. Lamotrigine clearance rises sharply (uptitrate, then cut postpartum); lithium pharmacokinetics shift with GFR and volume of distribution. Dose to serum levels, not to a fixed milligram value.
- Do not stop SSRIs before delivery to prevent poor neonatal adaptation syndrome. It is transient and self-limiting; pre-delivery discontinuation trades it for a peripartum maternal relapse.
- Never discontinue benzodiazepines abruptly in pregnancy. Convert to a long-half-life agent and taper slowly; cover the underlying anxiety disorder with an SSRI first.
- Document absolute risks and the patient's values. Preference-sensitive decisions require a recorded shared-decision conversation, not a unilateral order.
For more clinician resources on safe deprescribing and tapering, visit tapermeds.com.