Pregnant patients are nearly invisible in clinical trial data. One registry field could change that.

The federal clinical-trial registry lists more than 600,000 studies and draws over three million visitors a month, but it has no structured way to tell a pregnant patient whether a study will enroll her. That missing field is the lever an international guideline is trying to pull.

ICH E21, the International Council for Harmonisation's draft pregnancy-inclusion guideline, would move drug development from presuming pregnant and lactating people should be excluded from trials to recommending they be included by default. The guideline is in active draft. If finalised, it would push trial sponsors and ethics committees to plan for pregnant participants from the start, rather than bolt inclusion on later.

ClinicalTrials.gov, the federal registry run by the National Library of Medicine, is the surface most patients, clinicians, and systematic reviewers actually query. Its structured data elements include age, sex, diagnosis, study phase, and ZIP code. Pregnancy and lactation status are not among them. The information sometimes lives in the free-text "Eligibility Criteria" box, alongside unrelated filters like body mass index, language proficiency, and recent travel. A pregnant patient running that search cannot filter for pregnancy status. She has to read each eligibility paragraph, and the same opacity shows up in the registry's bulk download that downstream systematic reviews query.

In a STAT News opinion piece, a group of researchers and ethicists argue for the smallest meaningful fix: add a single pregnancy and lactation flag to the participation eligibility module, with an enrollment count when the answer is yes. The module already exists; the field is the addition. The same authors note that a checkbox would make pregnancy-relevant trials discoverable to patients and would let evidence-synthesis groups like Cochrane see the data a guideline like E21 would require sponsors to generate.

The mechanism matters because most drugs prescribed in pregnancy were never formally tested in pregnant people. Dosing and safety guidance is extrapolated from non-pregnant adults, postpartum physiology, and case series. The thinness of the evidence base predates any single guideline. A registry field does not manufacture new data, but it decides whether the data sponsors do generate shows up where reviewers and prescribers can use it. Even when sponsors enroll pregnant participants, that fact generally does not surface in the registry in a way a search can return.

Federal officials across the current administration have criticised the lack of drug safety evidence in pregnancy, including around antidepressants and acetaminophen, even as the underlying scientific conclusions in those debates remain disputed. Public attention also grew through Emily Oster's writing on pregnancy evidence and the early controversy over excluding pregnant people from COVID-19 vaccine trials. The bioethics framing has shifted accordingly: pregnant people are increasingly described as a "scientifically complex" group protected through research, rather than a "vulnerable" one protected from it the way children are. ICH E21 is the operational form of that reframe.

The argument is narrow on purpose. A registry field is not a clinical trial, and no field can replace a sponsor's decision to enroll pregnant participants. The eligibility module is the smallest piece of data infrastructure with the largest downstream effect, because every systematic review, drug label update, and prescribing decision that touches pregnancy eventually runs a query against ClinicalTrials.gov. Without the field, the data does not show up.

Two parallel timelines have to align. The International Council for Harmonisation has to finalise E21. The National Library of Medicine has to add a pregnancy and lactation field to the ClinicalTrials.gov data element definitions in step, with the controlled vocabulary sponsors need to fill it in consistently. Neither timeline is formally tied to the other, which is the smallest meaningful delay, and the smallest meaningful ask.