Low-dose aspirin reduces preterm birth in high-risk pregnancy, and two anxieties dominate its use: that starting late loses the benefit, and that imperfect adherence wastes it. Both anxieties drive clinicians to abandon aspirin in women who present after the ideal window or who admit missing doses. This post hoc analysis of the ASPIRIN trial tested whether either concern is justified.
The trial randomised 11,943 women at 6 to 13 weeks of gestation across hospitals in low-resource settings in Africa, Asia and Latin America, with outcomes available for 11,908. Women received prepackaged two-week allotments and adherence was assessed by pill count every two weeks. Median gestation at initiation was 10.1 weeks, and 85.5% of women had adherence above 90%.
Neither factor modified the effect. For each additional week of gestational age at initiation, the treatment risk ratio was unchanged for preterm delivery (0.97, 95% CI 0.93 to 1.02), for preterm delivery before 34 weeks (0.98, 0.89 to 1.07) and for perinatal mortality (1.04, 0.96 to 1.12). For each 5% increase in adherence, the treatment risk ratio was 1.01 (0.99 to 1.04) for preterm delivery, 0.99 (0.95 to 1.03) before 34 weeks, and 1.02 (0.98 to 1.06) for perinatal mortality. There was no effect modification by region.
So a woman booking at 13 weeks gets the same benefit as one booking at 6, and a woman taking most of her tablets gets the same benefit as one taking all of them. In practice this removes two reasons clinicians give for not starting aspirin, and it should be said to the patient too — a woman told that missing doses ruins the treatment is more likely to give up entirely.
- Start low-dose aspirin at any point between 6 and 13 weeks; later initiation within that window did not reduce benefit
- Do not withhold or stop aspirin because adherence has been imperfect
- Tell the patient that missing an occasional dose does not undo the treatment — that framing supports adherence rather than undermining it
- The finding held across Africa, Asia and Latin America, so it applies in low-resource settings
- This is post hoc: it tests effect modification within a positive trial, not whether aspirin works
The statistics, in plain English
This analysis tests effect modification, which is a different and harder question than whether a treatment works. All six confidence intervals here are narrow and centred close to 1.0, meaning the treatment effect really did not vary with timing or adherence — this is a well-powered null rather than an absence of evidence. One caveat about the adherence finding: 85.5% of women exceeded 90% adherence, so the analysis compares good adherence with very good adherence, and says nothing about a woman who takes a quarter of her tablets. Being post hoc, it generates confidence rather than proof.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for obstetrics & gynaecology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free