Psychiatric medication in pregnancy and lactation
A working framework for pregnancy and breastfeeding decisions across antidepressants, mood stabilizers, antipsychotics, benzodiazepines, and stimulants, with LactMed pointers.
The framework
Untreated maternal illness isn't a null risk. Depression in pregnancy is associated with preterm birth (relative risk around 1.4), low birth weight, and postpartum depression that impairs bonding and infant development. Untreated bipolar disorder carries about a 50% risk of relapse in pregnancy off medication, with substantially higher risk postpartum. Untreated psychosis has obvious risks to mother and fetus. Anxiety disorders in pregnancy are associated with preterm birth and later child anxiety and cognitive outcomes.
FDA labeling since 2015 is narrative. The letter categories were removed because they compressed nuance into a single letter and clinicians overinterpreted them. Current labels give you a Risk Summary, Clinical Considerations, and Data sections. This is more honest but harder to skim.
Primary sources for real decisions:
- MotherToBaby (mothertobaby.org). OTIS-affiliated. Free fact sheets for patients and clinicians.
- LactMed (bookshelf.nlm.nih.gov). NIH-maintained. The default source for breastfeeding.
- Reprotox and Teris. Subscription-based, more depth.
- MGH Center for Women's Mental Health. Ongoing pregnancy registries (National Pregnancy Registry for Psychiatric Medications).
- Cohen et al bipolar and lithium pregnancy data. The lithium teratogenicity numbers below come from this line of research.
Antidepressants
| Drug | First trimester risk | Later pregnancy risk | Lactation | Notes |
|---|---|---|---|---|
| Sertraline | No consistent teratogenic signal | Small risk of neonatal adaptation syndrome | Lowest infant serum levels of any SSRI; first-line while breastfeeding | Default SSRI in pregnancy for most patients |
| Escitalopram / citalopram | No consistent teratogenic signal | Neonatal adaptation syndrome; possible small PPHN signal | Acceptable, low infant exposure | Fine choices if patient is stable on one already |
| Fluoxetine | No consistent teratogenic signal | Neonatal adaptation, higher infant serum levels due to long half-life | Highest infant serum SSRI levels; some clinicians switch to sertraline for breastfeeding | Long half-life means changes take weeks |
| Paroxetine | Small cardiac malformation signal (VSD, ~1.5 to 2x baseline) | Neonatal adaptation | Fine in breastfeeding, low infant exposure | Avoid in first trimester if possible; if patient is stable, individualize |
| Fluvoxamine | Limited data, no clear signal | Neonatal adaptation | Acceptable | Less commonly used |
| Vortioxetine, vilazodone | Very limited human data | Presumed similar to SSRIs | Very limited data | Insufficient data to recommend as first-line; use if patient responded uniquely |
| Venlafaxine / duloxetine (SNRIs) | No clear teratogenic signal | Neonatal adaptation syndrome, some hypertension data | Acceptable, monitor infant | Duloxetine has less pregnancy data than venlafaxine |
| Bupropion | No consistent teratogenic signal | Limited data on adaptation | Acceptable | Sometimes preferred if smoking cessation is also a goal |
| Mirtazapine | Limited data, mostly reassuring | Sedation in some newborns | Acceptable | Useful when appetite, sleep, or nausea are targets |
| TCAs (nortriptyline, amitriptyline) | No consistent teratogenic signal | Neonatal adaptation, anticholinergic effects | Acceptable, nortriptyline has more data | Older and less used but reasonable when SSRIs fail |
| MAOIs (phenelzine, tranylcypromine) | Very limited data, animal teratogenicity | Hypertensive crisis risk during labor | Limited data, generally avoid | Reserve for refractory cases with specialist involvement |
Neonatal adaptation syndrome (formerly "poor neonatal adaptation") happens in 15 to 30% of infants exposed to SSRIs or SNRIs in the third trimester. Jitteriness, irritability, respiratory issues, feeding difficulty. It's transient (usually resolves in 2 weeks), rarely severe, and doesn't require preemptive tapering in the third trimester (which was tried and abandoned because relapse risk was higher than the benefit).
Persistent pulmonary hypertension of the newborn (PPHN) has been associated with late-pregnancy SSRI use in some studies. Absolute risk is small (baseline around 1 to 2 per 1,000, SSRI-exposed around 3 per 1,000). Not a reason to stop treatment in a symptomatic patient.
Sertraline is the default SSRI in pregnancy and breastfeeding because it has the most data, no teratogenic signal, and the lowest infant serum concentrations during breastfeeding. If a patient is stable on a different SSRI (say, escitalopram), switching just for pregnancy usually isn't necessary.
Mood stabilizers
| Drug | First trimester risk | Later pregnancy risk | Lactation | Notes |
|---|---|---|---|---|
| Lithium | Ebstein anomaly risk 0.05 to 0.1% (baseline 0.005 to 0.01%), about 10 to 20x baseline in absolute terms, still under 1 per 1,000 exposed | Fetal goiter, polyhydramnios; neonatal toxicity if not adjusted around delivery | Traditionally avoided; increasing acceptance with monitoring | Lowest teratogenicity of the mood stabilizers |
| Valproate | Neural tube defects (1 to 2%), craniofacial, cardiac, limb defects; ~10% total major malformation rate | IQ reduction of 8 to 11 points; autism spectrum disorder ~3x baseline | Compatible with breastfeeding but rarely appropriate given pregnancy avoidance | Contraindicated in pregnancy unless no alternative |
| Carbamazepine | Neural tube defects (~1%), craniofacial defects, ~5 to 6% major malformation rate | Vitamin K deficiency at delivery; some IQ signal | Acceptable but not first-line | Less teratogenic than valproate, still not preferred |
| Oxcarbazepine | Some teratogenic signal, less data than carbamazepine | Vitamin K considerations | Limited data | Not clearly better than carbamazepine |
| Lamotrigine | Overall major malformation rate close to baseline (~2 to 3%); oral cleft signal in older data, not confirmed in newer registries | Levels drop significantly (up to 50%) in pregnancy; needs monitoring | Compatible; infant levels can be measurable | Preferred anticonvulsant mood stabilizer in pregnancy |
| Topiramate | Oral cleft signal (~2 to 4x baseline), low birth weight | Ongoing exposure risk | Acceptable but not first-line | Better avoided as mood stabilizer in pregnancy |
Lithium in more detail. The Ebstein anomaly risk was historically overstated. Older estimates suggested 400x baseline, more recent registries and meta-analyses put it at 10 to 20x baseline, so an absolute risk around 0.05 to 0.1%. That's still elevated but far lower than the perception. Fetal echocardiogram at 16 to 20 weeks is standard for first-trimester lithium exposure. In the third trimester, blood volume expansion drops lithium levels, and doses often need to be adjusted upward. Around delivery, doses are typically held or reduced to prevent neonatal toxicity from the volume shift. Levels reset quickly postpartum, and the postpartum period is one of the highest-relapse windows in bipolar disorder, so restarting promptly matters.
Valproate. The teratogenic and neurodevelopmental risks are severe enough that valproate should be avoided in any patient of reproductive potential unless every alternative has failed. Malformation rate is around 10%, IQ reduction is documented, and autism spectrum disorder risk is elevated. If valproate is truly necessary, high-dose folate (4 to 5 mg daily) and detailed anomaly scan are the standard precautions, but the better answer is a different drug.
Lamotrigine's pregnancy pharmacokinetics. Estrogen upregulates UGT1A4, which metabolizes lamotrigine. Levels can drop by 30 to 60% during pregnancy, and dose increases are typically needed. Levels crash back postpartum, so the dose needs to be brought back down within days of delivery to avoid toxicity.
Antipsychotics
| Drug | First trimester risk | Later pregnancy risk | Lactation | Notes |
|---|---|---|---|---|
| Haloperidol | Reassuring data, no clear teratogenic signal | EPS in newborn; withdrawal | Acceptable | Most data of any antipsychotic |
| Risperidone | Reassuring, some hyperprolactinemia data | EPS in newborn | Acceptable | Reasonable choice |
| Olanzapine | Reassuring | Neonatal adaptation, metabolic; gestational diabetes | Acceptable, watch infant sedation | Metabolic effects on mother matter |
| Quetiapine | Reassuring | Neonatal adaptation | Acceptable | Common pragmatic choice |
| Aripiprazole | Limited but reassuring | Limited data | Limited data, presumed acceptable | Growing use |
| Clozapine | Reassuring on malformation, but complicated | Floppy baby syndrome, agranulocytosis surveillance in newborn required | Acceptable, monitor infant CBC | Only agent for treatment-resistant schizophrenia; keep going during pregnancy if needed |
| First-gen high potency (haloperidol, fluphenazine) | Reassuring | EPS in newborn | Acceptable | See above |
| First-gen low potency (chlorpromazine) | Some limb signal in older data, likely not real | Sedation, hypotension in newborn | Acceptable | Rarely used |
| Long-acting injectables | Data limited by design | Prolonged neonatal exposure due to depot half-life | Same as oral parent drug | Consider oral bridging for the last month of pregnancy in some cases |
The general principle for antipsychotics: if a patient with schizophrenia or bipolar with psychotic features gets pregnant, don't stop the antipsychotic. Relapse risk is high, consequences of relapse are severe, and the teratogenic and neurodevelopmental data are reassuring for most agents. Metabolic effects on the mother (gestational diabetes, weight) are worth watching, especially with olanzapine and clozapine.
Clozapine. Special case. Fetal exposure includes agranulocytosis risk to the newborn, so infant ANC monitoring for the first 6 months is standard. Floppy baby syndrome and neonatal adaptation are more common. But clozapine is often the drug that keeps a patient with treatment-resistant schizophrenia stable, and stopping it in pregnancy usually means relapse. The default is to continue with close monitoring.
Benzodiazepines
| Drug | First trimester risk | Later pregnancy risk | Lactation | Notes |
|---|---|---|---|---|
| Lorazepam | Historic cleft palate signal mostly disproven | Floppy baby syndrome, neonatal withdrawal near term | Acceptable in small amounts | Glucuronidated, no active metabolites, preferred if benzo is needed |
| Clonazepam | Similar to above | Similar | Acceptable but longer half-life means more infant accumulation | Common for panic disorder |
| Alprazolam | Similar | Similar, plus significant maternal discontinuation issues | Acceptable but similar accumulation concerns | Short half-life means interdose rebound and worse anxiety |
| Diazepam | Similar to above teratogenic story | Longest infant half-life, most accumulation | Not preferred for breastfeeding | Long half-life a liability across the board |
The cleft palate signal from the 1970s is essentially disproven by newer registries and meta-analyses. It's still commonly cited but shouldn't drive current decisions.
Floppy baby syndrome and neonatal withdrawal are real concerns for third-trimester benzo exposure. Sedation, hypotonia, feeding difficulties, and withdrawal (jitteriness, hyperreflexia) can occur. The magnitude depends on dose and duration.
Benzos in labor. A patient with severe panic disorder or PTSD may need benzo coverage during labor. This is a real conversation to have with obstetrics ahead of time. Lorazepam is the usual choice. The tradeoff is short-term neonatal sedation vs unmanaged severe maternal anxiety, and the answer depends on the patient.
Stimulants and ADHD medications
| Drug | First trimester risk | Later pregnancy risk | Lactation | Notes |
|---|---|---|---|---|
| Methylphenidate | Small association with cardiac malformations in some studies, not confirmed in others | Low birth weight, preterm birth signal | Acceptable, low infant exposure | Most data of any stimulant |
| Amphetamines (Adderall, lisdexamfetamine) | Less data than methylphenidate; historically associated with cardiac defects (largely disproven) | Similar low birth weight, preterm signal | Acceptable | Increasingly used, data improving |
| Atomoxetine | Limited human data | Limited data | Limited data | Nonstimulant alternative but not clearly better |
| Guanfacine, clonidine | Limited data | Neonatal hypotension possible | Acceptable | Sometimes used for anxiety-adjacent ADHD in pregnancy |
| Viloxazine | Very limited human data | Very limited data | Very limited data | Insufficient data to recommend in pregnancy |
| Modafinil / armodafinil | Concerning teratogenic signal (~15% major malformation rate in registries) | Same | Limited data | Avoid in pregnancy |
Stimulants and ADHD in pregnancy is a genuinely evolving area. Older reflexive advice was to stop all stimulants at conception. Newer thinking accepts that severe untreated ADHD affects pregnancy safety (accidents, medication errors, disorganized prenatal care), and that stimulant exposure risks appear smaller than previously feared. Modest increases in preterm birth and low birth weight are consistent findings. Cardiac malformation signals from the 1970s and 1980s have largely not held up.
Decision usually comes down to symptom severity, occupational demands, and the patient's own preference. Many patients choose to taper or discontinue for the first trimester and reassess.
Novel agents
Zuranolone (Zurzuvae). FDA-approved 2023 for postpartum depression. Oral 14-day course. Very limited data on use during pregnancy itself, but designed for the postpartum window. GABA-A modulator. Lactation data limited; label recommends against breastfeeding during use and for one week after last dose.
Brexanolone (Zulresso). IV formulation for PPD, requires 60-hour infusion in monitored setting. Same GABA-A mechanism. Very limited pregnancy data. Rarely used given zuranolone availability.
Ketamine and esketamine. Very limited human pregnancy data. Animal data raise neurodevelopmental concerns. Not recommended in pregnancy absent extraordinary circumstances (treatment-resistant depression with high acute risk). Lactation similar, avoid during breastfeeding sessions or pump-and-dump around dose.
The three decisions that come up most
1. A woman with well-controlled depression on an SSRI wants to conceive.
The instinct to stop the SSRI before conception isn't supported by evidence. Relapse of depression in pregnancy off medication is roughly 50 to 70% in women with prior significant depression. That relapse causes real harm (preterm birth, poor prenatal care, postpartum decompensation). If she's stable on sertraline or escitalopram, the usual advice is to continue at the lowest effective dose. If she's on paroxetine, switching to sertraline preconception is reasonable given the cardiac signal. If she's on fluoxetine and stable, continuing is fine; some clinicians switch to sertraline for breastfeeding reasons.
2. A woman with bipolar disorder on lithium wants to conceive.
Lithium has the lowest teratogenicity of the effective bipolar maintenance drugs, so it's often the right answer to continue. Fetal echocardiogram at 16 to 20 weeks. Level checks every 4 weeks in the second trimester and every 2 weeks in the third as volume expands. Around delivery, hold or reduce the dose to avoid neonatal toxicity from the volume shift. Restart promptly postpartum because the relapse window is severe. Lamotrigine is an alternative if the phenotype is depressive-predominant, though it's less protective against mania. Valproate is off the table. Antipsychotics are alternatives; olanzapine and quetiapine have reasonable pregnancy data.
3. A pregnant woman with panic disorder is asking for a benzo.
The default is SSRI first-line, not benzo. If she's stable on sertraline, continue. If she's new to treatment, start an SSRI. Benzos as needed for acute severe panic are reasonable, ideally low-dose lorazepam and not chronic. The cleft palate signal is essentially gone. Third-trimester exposure produces floppy baby and withdrawal, which is dose-related. If the patient truly needs benzo coverage through delivery, coordinate with obstetrics and neonatology in advance so neonatal monitoring is prepared.
Lactation notes
LactMed rules of thumb:
- Most psychiatric medications are compatible with breastfeeding.
- Infant serum levels matter more than milk levels.
- Watch for excessive sedation, poor feeding, or unusual irritability in infant.
- Preterm and low-birthweight infants have less metabolic capacity, so extra caution.
SSRI-specific lactation notes:
- Sertraline. Infant serum levels are typically undetectable or very low. First-line for breastfeeding.
- Paroxetine. Low infant exposure, fine for breastfeeding despite pregnancy considerations.
- Escitalopram, citalopram. Low infant exposure, acceptable.
- Fluoxetine. Highest infant serum levels of the SSRIs because of long half-life and active metabolite (norfluoxetine). Not a reason to stop, but often the reason clinicians switch to sertraline for breastfeeding.
Lithium and lactation. Historically avoided. Newer data support use with infant serum lithium level and TSH monitoring. Not the default in all practices, but no longer universally prohibited.
Valproate and lactation. Technically compatible (low milk levels), but the underlying pregnancy avoidance usually settles this before lactation is a question.
Antipsychotics and lactation. Most are compatible. Clozapine requires infant CBC monitoring. Watch sedation with olanzapine and quetiapine. Aripiprazole has the least sedating profile in this group.
Benzos and lactation. Occasional low doses of short-half-life benzos (lorazepam, oxazepam) are acceptable. Chronic high-dose exposure isn't recommended.
Stimulants and lactation. Methylphenidate and amphetamines produce low infant serum levels and are generally compatible. Watch infant weight, feeding, and sleep.
Common questions
Which SSRI is safest in pregnancy?
Sertraline is the usual first-line pick. It has the most safety data, no consistent teratogenic signal, and it's the SSRI with the lowest infant serum levels during breastfeeding. Escitalopram and citalopram are also fine. Paroxetine has a small cardiac malformation signal (VSD, roughly 1.5 to 2x baseline) and is usually avoided in first trimester if possible; if a patient is stable on it and switching would risk relapse, that's a judgment call. Fluoxetine works and has a lot of data, but the long half-life makes changes slow and infant serum levels during breastfeeding are the highest of the SSRIs.
Is lithium really contraindicated in pregnancy?
No. That was the old consensus based on the 1970s Register of Lithium Babies data that dramatically overstated Ebstein anomaly risk. Newer registries put the absolute risk of Ebstein anomaly around 0.05 to 0.1%, roughly 10 to 20x baseline but still under 1 per 1,000. Lithium remains the least teratogenic effective maintenance drug for bipolar disorder. Fetal echocardiogram at 16 to 20 weeks is standard for first-trimester exposure. Levels need close monitoring across pregnancy as volume expands and again around delivery. The postpartum relapse window is severe, so continuing (or restarting immediately postpartum) matters.
What about benzos in labor?
Situational. If a patient has severe panic disorder or PTSD and stopping benzos would risk maternal decompensation during a stressful labor, low-dose lorazepam coverage is reasonable. Coordinate with obstetrics and neonatology in advance so the newborn team knows to watch for sedation, floppy baby, or withdrawal. Chronic third-trimester benzo exposure carries a real neonatal risk, but a few doses during labor for a patient who genuinely needs them isn't the same clinical picture.
Can I breastfeed on sertraline?
Yes. Sertraline is the SSRI of choice for breastfeeding. Infant serum concentrations are typically undetectable or very low, and no consistent adverse infant effects have been reported at maternal therapeutic doses. This is one of the easier decisions in perinatal psychiatry.
Are stimulants safe in pregnancy?
The data have gotten better and the risks are smaller than older warnings suggested. Methylphenidate has more data than amphetamines, though amphetamine data are catching up. Both are associated with modest increases in preterm birth and low birth weight. Older signals for cardiac malformations have largely not held up. Decision usually comes down to symptom severity. Some patients taper for first trimester and reassess. Others with severe ADHD and safety concerns (driving, occupational risk) continue. Modafinil and armodafinil are different; those have a real teratogenic signal and should be avoided.
Sources
- MotherToBaby fact sheets (mothertobaby.org) for each drug class.
- LactMed database (bookshelf.nlm.nih.gov) for lactation.
- MGH Center for Women's Mental Health, National Pregnancy Registry for Psychiatric Medications.
- ACOG Committee Opinion 757 (2018) "Screening for perinatal depression" and related opinions on psychiatric medication in pregnancy.
- APA Practice Guideline on the Treatment of Patients with Major Depressive Disorder, perinatal considerations.
- Cohen LS et al, lithium and pregnancy outcomes registry data.
- FDA labels (post-2015 narrative format) via DailyMed.
- Larsen ER et al, Nordic guidelines on psychiatric medication in pregnancy.
- Payne JL et al, perinatal psychopharmacology reviews.
Reviewed against current guidelines as of June 8, 2026. This is not medical advice.
THE KNOWLEDGE PATH
Walk this topic outward.
- GUIDE Psychiatric medication in pregnancy and lactation (current)
- CLASS SSRIs
- MEDICATION Sertraline (Zoloft)
- CONDITION Major Depressive Disorder (on Shrinkopedia)
- CARE Depression care at shrinkMD
The Knowledge Path is a curated walk. Every step is one decision away from the next.
Managing a medication needs a prescriber
Any psychiatric medication has to be started and adjusted by a clinician who can follow you over time. If you don't have a prescriber, our guides section explains the options, including in-person care and telepsychiatry, and how to choose between them.