Objective:

We conducted a systematic review of the evidence on the use of low-dose aspirin for the prevention of morbidity and mortality from preeclampsia to support the U.S. Preventive Services Task Force (USPSTF) in updating its previous recommendation. Prior reviews have established that benefits of aspirin prophylaxis are not obtained in populations of healthy or unselected pregnant women not at high risk of preeclampsia. In this review we considered the evidence on benefits and harms of low-dose aspirin for women at elevated risk of developing preeclampsia and consequent maternal and fetal health outcomes. Three key questions (KQs) were systematically reviewed: 1) Is there evidence that aspirin reduces adverse maternal or fetal health outcomes? 2) Is there evidence that aspirin reduces incidence of preeclampsia? and 3) What are the harms of low-dose aspirin use during pregnancy?

Data Sources:

We identified nine existing relevant systematic reviews and performed a search of MEDLINE, the Database of Abstracts of Reviews of Effects, PubMed, and the Cochrane Collaboration Registry of Controlled Trials for studies published from January 2006 through 2013. We supplemented searches by examining bibliographies from previous systematic reviews and retrieved articles, previous USPSTF reviews, and consulting outside experts. We searched Federal agency trial registries for ongoing and/or unpublished trials.

Study Selection:

We conducted dual independent review of 544 abstracts against a priori inclusion and exclusion criteria. The 75 potentially relevant articles identified were then independently evaluated by two reviewers against the same inclusion/exclusion criteria and critically appraised for quality/risk of bias using USPSTF criteria. Discrepancies were resolved in discussion with a third reviewer. A single investigator extracted study characteristics and outcomes for all fair- to good-quality studies into tables and a second reviewer checked accuracy.

Data Analysis:

Evidence for all KQs was qualitatively synthesized. Quantitative synthesis of outcomes where there was sufficient data used random-effects meta-analysis models as the primary analysis. Analyses were stratified by the timing of aspirin administration and dosage, with statistical tests of strata differences conducted. Funnel plots and tests for small-study effects were conducted.

Results:

One large U.S. study (n=2,539), one large international study based in the United Kingdom (n=9,364), and 13 smaller trials were included for evaluation of benefits of aspirin. Additionally, six randomized, controlled trials (RCTs) of women not at increased risk for preeclampsia contributed to the analysis of harms. Five of these studies were prophylaxis RCTs among women with low or average preeclampsia risk: a good-quality multisite study in the United States (n=3,135) and a smaller U.S. study (n=606), a good-quality multisite study in France and Belgium (n=3,294), a good-quality hospital-based study in Barbados (n=3,647), and a fair-quality U.K.-based study (n =122). The sixth study was a good-quality Australia-based RCT of fetal growth restriction treatment (n=51). Two observational studies were also included for the review of harms: a good-quality cohort study following 47,400 women enrolled during pregnancy and a good-quality case-control study based on data from a large prospective cohort study (n=3,129).

Based on pooled results, low-dose aspirin administered after the first trimester of pregnancy to women at elevated risk of preeclampsia reduced the risk of preeclampsia by at least 10 percent (and perhaps 24%), with beneficial effects on perinatal health outcomes; intrauterine growth restriction (IUGR) was reduced 20 percent and preterm birth an estimated14 percent, although the actual effect for these two outcomes may be more modest, given the possible bias due to small-study effects. Consistent with findings of lower rates of preterm birth and IUGR, birth weight averaged 130 g more in infants whose mothers took low-dose aspirin. We did not find evidence of serious harms from aspirin use (i.e., no effect on perinatal mortality), although power was limited for such a rare event. Individual trials were inconsistent, with nonstatistically significant findings in the direction of both modest benefit and modest harm; pooling of perinatal mortality findings suggested a tendency toward a reduced (rather than increased) risk of perinatal mortality (relative risk [RR], 0.92 [95% CI, 0.76 to 1.96]), particularly when analyses were limited to only women at increased risk of preeclampsia (RR, 0.81 [95% CI, 0.65 to 1.01]). Similarly, available evidence on intracranial fetal bleeding suggested no effect with low-dose aspirin (RR, 0.84 [95% CI, 0.61 to 1.16]). Although there was no overall effect of low-dose aspirin on several maternal harms (i.e., postpartum hemorrhage, Cesarean delivery), we could not eliminate the possibility of an increased risk of abruption because of power limitations and heterogeneity of risk for preeclampsia. Pooling limited to trials enrolling higher-risk pregnant women (the target for aspirin intervention) somewhat attenuated the potential for harm from abruption, but results remained heterogeneous. Two observational studies on aspirin use during pregnancy had null findings for the potentially harmful outcomes considered (miscarriage and cryptorchidism).

Limitations:

Very little new evidence has accrued since the completion of a number of large studies conducted in the 1990s. Since then there have been multiple systematic reviews, including one individual-level meta-analysis, and a few smaller trials (n<1,000). The serious health outcomes that are the aim of aspirin prophylaxis are rare and there is insufficient power, even in pooled analyses, to detect effects that could be clinically important.

There is evidence of small-study bias in the evidence we reviewed, based on funnel plots, formal statistical tests, and observation of forest plots sorted by sample size, showing a clear decrease in effect size with increasing sample size. Given that the large studies are from multiple sites, they likely share some of the features of small studies in terms of study operations. Those studies combined in the large multisite trials, however, are necessarily reported in the literature regardless of results, whereas null findings of small independent trials may be less likely to publish null results.

Trial characteristics cannot always be disentangled from study size due to the presence of small-study effects. The ability to draw conclusions related to dosage from the available trial evidence is limited by the fact that the two largest studies used 60 mg of aspirin, although they differed on other important characteristics. Thus, stratification by dosage is potentially confounded; the apparent benefit of a dose greater than 75 mg found in other systematic reviews could be due either to the small sample effect, a true dose effect, or a combination of these factors.

Conclusions:

For women at elevated risk of preeclampsia, prophylaxis with low-dose aspirin (60 to 150 mg) beginning after the first trimester of pregnancy reduced risk of preeclampsia and important adverse perinatal health outcomes. Specifically, modestly reduced risks of preterm birth, IUGR, and possibly perinatal mortality were supported by the evidence. Consistent with lower risk of preterm birth and IUGR, a significant difference in birth weight was also present. Statistical significance was not attained for the estimated 19 percent reduction in risk of perinatal mortality, although power to detect this difference was under 50 percent; there is a risk of incorrectly accepting a null result for perinatal mortality based on currently available data. The effects on perinatal mortality observed in the two largest trials were consistent with a benefit, although more modest.

The pooled results finding reduced risk of preeclampsia with low-dose aspirin supports the causal pathway leading to the observed direct health outcomes. The pooled results may have overestimated the benefit, however, given the evidence of small-study effects and more modest results in the two largest trials. However, given the consistency of the effect size in the large trials and the results of pooled analysis, at least a 10 percent reduction in preeclampsia was supported by the evidence. This reduction in preeclampsia incidence likely underlies the observed perinatal health benefits.

There was limited evidence of harms associated with low-dose aspirin use during pregnancy. A potential increased risk of abruption could not be ruled out, but evidence of harm from other bleeding-related complications, such as postpartum hemorrhage, maternal blood loss, and neonatal intracranial or intraventricular bleeding was not found. The evidence on longer-term outcomes for offspring from in utero aspirin exposure (low-dose) is very limited, but followup data from one large randomized, controlled trial is reassuring.