Article — "Maternal ADHD Treatment Surges as Diagnosis Reaches Women—But Safety Data Lag Behind": Article — "Maternal ADHD Treatment Surges as Diagnosis Reaches Women—But Safety Data Lag Behind": "Continuing dexamphetamine throughout pregnancy was not associated with an increase in adverse neonatal and maternal health outcomes compared to ceasing."
That sentence, from a Western Australian cohort in which 547 women continued dexamphetamine throughout pregnancy and 297 ceased before the end of the second trimester, quietly inverts the advice most of them were probably given. The women who stopped reportedly had 2.28 times the odds of threatened abortion compared with the women who kept taking it, though the authors reportedly caution this borderline result is uncertain given the small sample size.
Hold that number, because it cuts against the reflex.
The reflex is old and powerful: when in doubt, discontinue. It survives partly because the population it governs used to be small. That is no longer true.
Between 2003 and 2015, the proportion of reproductive-age women in the United States filling an ADHD prescription reportedly rose 344%. In the Quebec Pregnancy/Children Cohort, the share of pregnancies exposed to ADHD medication reportedly climbed 14-fold between 1998 and 2015, from 0.08% to 1.2%.
These are not stable rates nudging upward. They are the sound of a diagnostic gap closing in real time.
For decades, ADHD was read off the bodies of hyperactive boys. Girls who were merely inattentive were missed, then diagnosed late—often in their twenties and thirties, often already contemplating pregnancy, often already on a stimulant that works.
So the wave of newly diagnosed women is not arriving before their reproductive years. It is arriving during them, medication in hand.
And it collides with an evidence base that the lactation literature describes, without embarrassment, as "largely limited to case reports."
That collision produces the framing you see everywhere: diagnosis has outrun safety data, so caution—meaning cessation—is the responsible default. It sounds rigorous. It is mostly wrong.
Diagnosis has outrun safety data, so caution—meaning cessation—is the responsible default. It sounds rigorous. It is mostly wrong.
Start with what actually happens when researchers look at the harms.
A 2025 systematic review in Archives of Women's Mental Health pulled twelve high-quality cohort studies on ADHD medication in pregnancy. Seven found no significant negative effect on mother or offspring. One reportedly found continuing reduced the risk of various bad outcomes. One found stopping raised the risk of threatened abortion.
Three did flag real harms—pre-eclampsia, gastroschisis, omphalocele, transverse limb deficiency—and modafinil stood out as genuinely raising malformation risk, which is why the reviewers say it should be stopped before pregnancy.
But for the stimulants, the reviewers reached a different conclusion: the benefits of continuing, where the drug works for that individual, "are likely to outweigh the risks."
The reason is confounding, and it is not a minor caveat. ADHD itself predicts worse pregnancies.
Women with ADHD reportedly carry higher baseline risks of gestational diabetes, pre-eclampsia, and preterm delivery, and higher rates of postpartum anxiety and depression—before any pill is counted. Many of the harms historically pinned on the medication keep dissolving back into the disorder being treated.
This is why the Western Australian design matters. It did not compare treated women to healthy women; it compared continuers to ceasers and to a matched group who took an ADHD drug before pregnancy but not during.
Against continuers, the truly unexposed did show some advantages—lower odds of pre-eclampsia, hypertension, postpartum haemorrhage, fetal distress, and neonatal special care admission. So the finding is genuinely mixed. The authors themselves conclude that where possible, ceasing before pregnancy may be advisable.
Note the structure of that sentence. Not "cease when pregnant"—cease before, deliberately, as a planned decision.
Because once a woman is pregnant on the medication, the relevant comparison is no longer treated-versus-never-treated. It is continue-versus-stop-now, and on that comparison the stop-now arm carried the 2.28-fold threatened-abortion signal, with no offsetting benefit.
For many women already stabilized on a stimulant that works, continuation is plausibly the lower-risk default. The reflex points the wrong way.
So why do clinicians still freeze?
Because pregnancy is only half the problem, and the second half—breastfeeding—is where the data genuinely thins to almost nothing. You cannot ethically draw repeated blood samples from a nursing infant to measure how much drug it absorbs.
This is exactly where the framing curdles into paralysis. No infant pharmacokinetic study is coming, so "wait for the data" becomes a permanent excuse to say no.
The way out is not more studies. It is a reasoning tool most clinicians underuse: the relative infant dose, or RID.
The RID asks one clean question: what dose does the breastfed infant ingest per day, relative to the mother's daily dose, both standardized per kilogram? A rule of thumb holds that an RID under 10% is likely compatible with breastfeeding.
Its power is that the only biological sample you need is breast milk. The infant never gets a needle.
Work the standard example. A 70 kg mother takes 140 mg/day—2.0 mg/kg/day. You assay her milk across a dosing interval and model the mean concentration; assume it comes to 0.1 mg per 100 mL.
Standardize infant intake at 150 mL/kg/day. A 3 kg infant drinks 450 mL, ingesting 0.45 mg—0.15 mg/kg/day. Divide infant exposure by maternal exposure: 0.15 / 2.0 = 7.5%.
Now the elegant part. Repeat it for a 5 kg infant, or an 8 kg infant. The RID does not move. It stays 7.5%.
That invariance is the whole point. Because everything is expressed per kilogram, one milk assay yields a figure that holds across infant sizes and, under linear pharmacokinetics, across maternal doses too.
A single, ethically feasible study of the mother's milk stands in for the infant blood draws you can never do. This is precisely the substitution the viloxazine lactation study reportedly performed—fifteen lactating women, maximum dose, milk sampled, potential infant exposure estimated. That is what turning thin data into a usable number looks like.
The lesson generalizes. The "safety data lag" is real in raw quantity, but the deeper deficit is not missing numbers. It is the absence of a shared framework for converting the numbers we can get into a defensible clinical choice.
The RID does that for lactation. The continue-versus-cease cohort logic does it for pregnancy. Both already exist. Both are underused because the profession would rather retreat to "we can't be sure" than commit to a default.
And "we can't be sure" is never neutral. It has a direction—toward stopping.
Every time uncertainty defaults to discontinuation, a specific thing is traded away: the maternal function that the medication was restoring. The organization, the impulse control, the capacity to manage the very cognitive and emotional load that pregnancy and a newborn pile on.
The expert guidance is already moving. The 2024 AJOG review reportedly says perinatal ADHD pharmacotherapy largely has reassuring safety data and tells providers to weigh medication risk against the risk of inadequately treated ADHD.
The honest position is not "keep everyone on their meds." Modafinil is a real exception. The unexposed did show some advantages, and planned pre-pregnancy cessation is a legitimate choice for the right patient.
The honest position is that the choice should be reasoned, not reflexive—stated as a default and then adjusted, rather than surrendered to inertia dressed up as caution.
Thin data does not excuse a clinician from deciding. It obligates them to decide well, with the tools that already exist, instead of letting "when in doubt, stop" make the decision for them—and letting a mother's functioning be the thing that quietly pays for it.