Design: Prospective multicentre observational cohort
Population: 947 women with epilepsy aged 18–45 in China; 451 pregnancy samples
Primary outcome: Change in antiseizure medication concentration-to-dose ratio by trimester
Effect: Lamotrigine −28.8%, −54.3%, −63.2% by trimester; nadir −65.8% at 32 weeks

This prospective multicentre cohort in China followed 947 women with epilepsy aged 18–45 and measured steady-state trough levels, expressed as concentration-to-dose ratios, in 1,638 samples — 451 during pregnancy.

Lamotrigine fell the most: by 28.8% in the first trimester, 54.3% in the second and 63.2% in the third, reaching a nadir of −65.8% at 32 weeks. Levetiracetam fell by up to 40.1%, bottoming at 24 weeks, and the active metabolite of oxcarbazepine by up to 44.3%. Lacosamide fell modestly in the second trimester. Other antiseizure medicines shifted baseline levels but did not change the gestational pattern, and higher body weight was linked to lower levels.

The variation between individuals was larger than any measured factor explained. That is the argument for measuring, not predicting: a pre-pregnancy baseline level and regular checks through pregnancy, with dose adjustment, and a plan to reduce the dose after delivery.

Measure a baseline antiseizure drug level before conception or at the first antenatal visit.
Recheck lamotrigine levels at least each trimester, and more often in the second and third.
Increase the dose when levels fall below the woman's own seizure-free baseline, not a population range.
Monitor levetiracetam and oxcarbazepine too — both fell by a third or more.
Plan a dose reduction schedule for after delivery, when levels rise quickly and toxicity can follow.

Why it matters

It quantifies how far and how early levels drop, turning 'consider monitoring' into a schedule.

The statistics, in plain English

The concentration-to-dose ratio shows how much drug is in the blood per milligram taken; a 60% fall means the same dose gives less than half the level. The β coefficients are on a log scale; the percentages are the easier numbers to use. This was observational, so it describes what happens, not whether monitoring prevents seizures — but the size of the fall makes the case.