Reference: Ann Intern Med. 2021 Jan 2
Low-dose aspirin (LDA) is recommended during pregnancy for prevention of conditions such as preeclampsia and recurrent pregnancy loss, but preconception benefit has yet to be established. A recent paper in the Annals of Internal Medicine applies new statistics to an old dataset first published in 2013 as the EAGeR trial. That trial evaluated the effects of preconception LDA on pregnancy in patients with prior pregnancy loss. The authors of the current study reanalyzed the trial data using a complicated post-hoc per-protocol analysis, stating that the “EAGeR’s primary intention-to-treat (ITT) results demonstrated a 10% increase in live births in the LDA group relative to placebo, but without the precision to warrant changes to clinical recommendations (95% CI 0.98-1.22).” Translation: since results were not statistically significant using the appropriately conservative ITT analysis, the authors attempted to achieve statistical significance using a per-protocol approach.
The original trial randomized 1,227 patients aged 18-40 years with one or two prior pregnancy losses who were actively trying to conceive to either aspirin 81 mg daily or placebo. The intervention was initiated before conception and continued through week 36 of pregnancy if pregnancy occurred. The investigators created simulations with varying adherence and compared them to each other in a per-protocol analysis. The best case scenario of taking LDA over the entire follow-up period predicted a benefit over placebo in live birth (risk ratio [RR] 1.33, 95% CI 1.08-1.64), hCG-detected pregnancy (RR 1.12, 95% CI 1.02-1.23), and pregnancy loss (RR 0.69, 95% CI 0.5-0.95). Perhaps the most interesting findings were that adherence to either placebo or LDA was associated with higher rates of pregnancy and live birth (OR 5.7, 95% CI 4.34-7.52 and OR 1.33, 95% CI 1.07-1.67, respectively) and lower rates of pregnancy loss (OR 0.7, 95% CI 0.51-0.95).
This study demonstrates “adherence effect” bias, wherein an inherent difference in participants who comply with a prescribed intervention can confound the risk of the outcome being measured. Using a per-protocol approach to reanalyze these data introduces additional systematic bias. When participants are selectively excluded from analysis due to deviation from the study protocol, as in per-protocol analyses, the prognostic benefit of randomization is diminished because adherent participants are potentially compared to a mixture of adherent and nonadherent participants. The difference between per-protocol and ITT analyses is not about ideal vs. real-world scenarios, it’s about systematic bias.
Part of the problem with this new study stems from an all-or-nothing approach to statistical significance. The lack of significance in the original trial within the 95% confidence interval does not mean the intervention is not beneficial. It means the benefit is not proven to a degree sufficient for use in clinical practice. Based on the per-protocol analysis, the authors suggest that low-dose aspirin initiated preconception may improve pregnancy outcomes. That may well prove true, but the way to demonstrate this is with additional trials, not using less robust data analysis.
For more information, see the topic First Trimester Pregnancy Loss in DynaMed.
DynaMed EBM Focus Editorial Team
This EBM Focus was written by Katharine DeGeorge, MD, MS, Associate Professor of Family Medicine at the University of Virginia and Clinical Editor at DynaMed. Edited by Alan Ehrlich, MD, Executive Editor at DynaMed and Associate Professor in Family Medicine at the University of Massachusetts Medical School, Carina Brown, MD, Assistant Professor at Cone Health Family Medicine Residency, and Dan Randall, MD, Deputy Editor for Internal Medicine at DynaMed.