Aspirin cut recurrent preterm birth, but only with prior placental vascular damage

Low-dose aspirin is already standard advice for women at high risk of pre-eclampsia, where it works by improving how blood vessels in the placenta develop. Whether it also prevents recurrent preterm birth has been less clear: the main APRIL trial, which tested aspirin against placebo in Dutch women who had already had a spontaneous preterm birth, did not find a convincing overall benefit [s2]. A new secondary analysis published on 23 October in BJOG asks a sharper question — does aspirin help the specific women whose previous placenta showed signs of vascular damage? [s1]

The idea being tested

Maternal vascular malperfusion (MVM) is a pattern pathologists can read in a delivered placenta: it reflects poor blood flow from mother to placenta, the same vascular problem aspirin is thought to counter [s1]. The hypothesis is intuitive. If aspirin works through blood vessels, it should help women whose earlier pregnancy carried a vascular fingerprint, and do little for women whose preterm birth had other causes [s1].

The researchers went back to the APRIL study, which ran across 34 centres in the Netherlands and compared low-dose aspirin with placebo in women with a history of spontaneous preterm birth [s1]. They retrieved the pathology reports from each woman's previous preterm delivery and classified them for MVM using the standard Amsterdam criteria [s1].

What they found

Placental histology from the earlier birth was available for 41.3% of participants — 160 of 387 — and of those, MVM was present in 27.5%, or 44 women [s1]. That is the first caution: the key subgroup is small.

Within it, the contrast was stark. Among women with MVM, recurrent preterm birth occurred in just 4.2% of those taking aspirin — 1 of 24 women — compared with 45.0% on placebo, 9 of 20 [s1]. That works out to a relative risk of 0.09 (95% CI, 0.01 to 0.43; p = 0.02): on its face, a roughly 90% reduction [s1].

Among women without MVM, aspirin showed no benefit and, if anything, a numerically higher rate: preterm birth in 29.3% on aspirin (17 of 58) versus 17.2% on placebo (10 of 58), a relative risk of 1.70 that did not reach statistical significance (95% CI, 0.87 to 3.56; p = 0.13) [s1]. The authors conclude that in the vascular-mediated subgroup, aspirin in a later pregnancy was associated with fewer recurrent preterm births [s1].

Why restraint is warranted

The effect size is dramatic, but it rests on a handful of events. A single preterm birth in the aspirin group versus nine in the placebo group drives the entire result, and with numbers that small, chance and the play of a few cases can swing a relative risk enormously [s1]. The confidence interval runs from 0.01 to 0.43 — wide enough that the true benefit could be far more modest than the point estimate [s1].

This is also a post-hoc subgroup analysis: the women were not randomised by MVM status, histology was only available for four in ten participants, and the original trial did not find a clear overall effect [s1][s2]. Subgroup findings this striking, from trials that were neutral overall, have a long history of failing to hold up when tested directly. The reason is statistical as much as biological. Split any trial finely enough and some subgroup will, by chance alone, show a large apparent effect; the more impressive it looks, the more caution it deserves, especially when the parent trial found nothing overall.

There is also a question of why the women without vascular damage did numerically worse on aspirin, even if the difference was not significant [s1]. Taken at face value it would be an odd and unwelcome signal, but with 58 women in each of those groups and a wide confidence interval, it is most likely noise rather than evidence of harm. It does, however, underline how unstable estimates become once a modest trial is carved into pieces.

There is a deeper practical limit, too. MVM can only be read after a placenta is delivered, so the marker identifies which previous pregnancy had vascular trouble — it cannot be measured in the current pregnancy before deciding to start aspirin [s1]. Using it to guide treatment would mean acting on last time's pathology, which may or may not repeat.

The honest reading is that this analysis generates a hypothesis rather than settling one. It offers a biologically coherent reason why aspirin might help some women and not others, and a candidate way to tell them apart. What it does not provide is the kind of evidence — a trial that enrolls women by placental pathology and randomises them prospectively — that would justify changing practice. Until that exists, the result is a lead worth chasing, not a rule worth following.