Pregnancy follow-up in 2026 trial planning: what ICH E21 adds to an E19 data strategy

TL;DR

Specifying pregnancy follow-up requires more than a “pregnancy: yes/no” field in a trial database. The May 2025 Step 2 draft of ICH E21, as published by FDA, describes study-specific maternal and infant assessments, consent for follow-up, and mechanisms that may need to continue after the parent trial ends. It remains a draft, not a binding 2026 data-collection mandate or a universal visit schedule. [1]

ICH E19 addresses a different question: when selected safety data may be collected less comprehensively in particular late-stage trials. Its general collection list includes pregnancy and lactation exposure and outcomes. It permits consideration of selective collection of some listed data where regional or local rules allow it and the protocol adequately justifies it; preserve both conditions. Nor should E19 be used as a blanket reason to remove pregnancy follow-up. [2]

For sponsors, the practical decision is to define the follow-up obligation before selecting the database structure. Specify what is being followed, for how long and why, whose consent covers the collection, which sources can be used, and who remains responsible after treatment discontinuation or trial closeout. The operational model below is our interpretation of those planning needs, not an architecture required by ICH or FDA.

E19 and E21 answer different questions

E19's selective approach concerns particular late-stage preapproval or postapproval trials where the safety profile is sufficiently characterized to justify it. Gene-therapy and rare/orphan-disease trials fall outside its scope, and local or regional safety-reporting obligations remain applicable. It also distinguishes reduced collection into trial case report forms from the investigator's clinical-care and medical-record responsibilities. [2]

Those boundaries are important when a data-reduction initiative meets a pregnancy. A selective approach to nonserious events still requires separate decisions on pregnancy exposure and outcomes, appropriate clinical monitoring and applicable local reporting duties. The sponsor should review the actual selective-collection rationale, population and protocol provisions rather than apply a global database flag.

E19 section 2.4 lists pregnancy and lactation exposure and outcomes among data that should generally be collected. The following paragraph allows consideration of selective collection of some of the listed data when permitted by local or regional regulations and supported by adequate protocol justification. This is a qualified provision, not a universal exemption and not a statement that every item on the list is absolutely immutable. E19 section 2.7 recommends obtaining prospective regulatory agreement on whether selective safety data collection is justified and on its planned implementation. This is a guidance recommendation, not a universal statutory approval requirement. [2]

E21's draft then helps frame what meaningful pregnancy-related collection can involve. It proposes systematic, study-specific collection considering the product's properties, class risks, timing and extent of gestational exposure, available assessment methods and burden on the participant and child. Outcomes and parameters should have precise definitions and identified sources. [1]

Read together, the documents call for study-specific decisions about pregnancy data collection. They yield a sequence of decisions: whether the selective strategy is appropriate, what pregnancy-related information remains necessary, and how the trial will obtain interpretable information without imposing unjustified burden. Neither source supports a universal promise that a reduced-data strategy is safe for every product or population.

Link the participant, pregnancy and infant follow-up

The pregnant participant, a pregnancy episode and an infant's follow-up represent distinct but related records. One participant can have more than one pregnancy during a long study; one pregnancy can involve more than one fetus or infant; and outcomes can occur at different times. A single subject-level field can describe pregnancy occurrence while still being unable to distinguish these subsequent events.

E21's draft appendix expressly considers outcomes relevant to multiple pregnancies, including loss of one or more fetuses and other multiple-pregnancy-specific circumstances. It separately discusses maternal and gestational outcomes, infant characteristics at birth, and neonatal and later follow-up. [1] These distinctions support a data-model requirement for identifiable relationships, not a requirement for one particular software object hierarchy.

An implementation may use repeating forms, linked records or a separate follow-up system. Each can be acceptable as a design choice if it preserves the necessary relationships, access controls and history. The test is whether reviewers can tell which exposure, pregnancy and infant an observation concerns without copying unnecessary identifiers into every dataset.

Consider a hypothetical multiple pregnancy with different outcomes for the two infants. If the system stores only one “birth outcome” value against the maternal participant, a later update can overwrite or ambiguously combine the outcomes. A model with distinguishable infant records can avoid that ambiguity, but it still needs a defined linkage and a process for corrections. Verify the accuracy of those links and the authorization for collection separately from creating the record types.

The same logic applies to missing information. “Not yet due,” “assessment not performed,” “information not available,” “consent not provided” and “lost to follow-up” describe different situations. Our recommendation is to preserve those distinctions where they affect interpretation, rather than let every empty field become an implied normal outcome. These status categories are an implementation proposal; they should be tailored to the protocol and follow-up process.

Timing of exposure matters more than a generic completed form

E21's draft emphasizes the timing and extent of exposure during gestation and advises caution in interpreting infant adverse-event causality, particularly with small samples or no control arm. It also notes that the underlying maternal disease and other factors can affect pregnancy outcomes. [1]

That makes exposure timing and outcome definition central to interpretability. A pregnancy outcome recorded without the relevant gestational context may be difficult to relate to treatment exposure. A causal assessment of a later adverse outcome requires evidence beyond the temporal link to exposure. The collection plan needs to distinguish observed dates, estimated dates and uncertainty, and preserve the information needed for the protocol's intended assessment.

A useful design discussion asks which exposure intervals matter, which source establishes gestational timing, how a corrected estimate is handled, and whether the outcome is defined consistently across sites. It should also ask which disease or concomitant-treatment context is needed to interpret the outcome. Use those questions to select the maternal information needed for the planned assessment and to identify gaps that could mislead interpretation.

For example, changing an estimated pregnancy date can change the derived gestational timing of an exposure. A system that updates the displayed trimester but loses the previous estimate and reason for correction makes the derivation difficult to review. Our proposed acceptance test is to correct a representative date and verify that the updated interpretation remains traceable to its source and history, with affected downstream records reviewed as appropriate.

The draft notes that general-population reference rates can provide context, while disease-specific registries or observational studies may be more informative. [1] A sponsor should therefore not turn a handful of trial pregnancies into a stable causal rate or compare them casually with a population that has a different underlying risk. A well-structured database improves interpretability; suitable comparators and assessment of confounding remain separate methodological needs.

Appendix 2 is a planning framework, not an identical visit schedule

E21 draft Appendix 2 identifies outcome parameters to consider in light of the disease, product properties, duration of use and therapeutic context. It includes maternal and gestational outcomes, birth characteristics, and neonatal outcomes relevant within the first 28 days. It also recognizes that some developmental conditions may become apparent later. [1]

The neonatal-period reference leaves study-specific reasons for follow-up beyond day 28 intact. Nor does the appendix require every trial to collect every listed assessment at the same frequency. The draft's examples, including APGAR and neurological assessment, require study-specific timing rather than one universal schedule. Specify each visit's scientific and protocol basis; the draft provides no universal day-28, month-12 or month-24 visit requirement.

Section 4.2.8 instead says follow-up duration should be considered case by case, considering factors including half-life, indication, nonclinical data, mechanism of action, timing and duration of exposure, and the time at which outcomes may manifest. It recognizes that follow-up can extend past trial completion and discusses possible mechanisms, including follow-up studies and existing pregnancy registries. [1]

The sponsor therefore needs two explicit choices: the scientific follow-up horizon and the operational mechanism that can sustain it. A protocol might justify a particular set of neonatal assessments and longer follow-up for a product-specific concern. Another may require a different approach. The choice should be documented rather than inferred from whichever visits happen to be available in the standard EDC template.

This matrix is an editorial design aid. It should be adapted to the approved study and local requirements, not adopted as a mandatory universal dataset.

Specify consent and blinding for each follow-up role

The E21 draft's consent section discusses pregnancy-specific information, follow-up of the pregnancy and infant or child, planned duration, additional data sources, release of relevant medical records and confidentiality. It also points to local requirements and circumstances requiring reconsent. Participants who withdraw should understand the importance of pregnancy-outcome follow-up and be encouraged to consent to collection, with their decision respected. [1]

A system design should therefore distinguish the parent trial's consent from the actual permission and conditions for follow-up. The appropriate legal and ethical process is study- and jurisdiction-specific. Verify authorization for both the medical-record source and any registry destination, separately from their technical accessibility.

The proposed operational control is a consent-aware handoff. Before records move to a follow-up process, confirm the applicable permission, the information to be transferred, authorized recipients and continuing responsibilities. Retain enough evidence to explain why the transfer and subsequent collection are allowed. Avoid duplicating direct identifiers or sensitive information merely because a new form is easier to build with a full copy of the original record.

Blinding requires a separate decision. For pregnancies occurring in trials with mandatory contraception, E21 draft section 4.2.11 says the protocol should specify whether and when unblinding is expected; pregnancy should not automatically trigger unblinding. It also discusses protocol-defined options for remaining in the trial with treatment suspended, or in some circumstances continuing treatment after reconsent. [1]

Copying treatment assignment into an infant follow-up record can expose blinded information and therefore requires an explicit access decision. Our implementation recommendation is to provide each role only the information justified by the approved process, with any necessary unblinding handled through the designated pathway. A coordinator may need participant contact details while having no authorized need for randomized assignment.

Treatment continuation, discontinuation and study participation require medical, ethical and protocol decisions. The software should preserve the clinically determined status and follow-up plan, keeping those decisions with the responsible team.

Pregnancy collection and expedited reporting are different decisions

Assess reportability separately from the plan to collect pregnancy exposure and outcome information. The applicable reporting rules and medical assessment still govern. For example, US IND safety reporting under 21 CFR 312.32 distinguishes seriousness, unexpectedness and evidence suggesting a causal relationship; the definition of seriousness includes congenital anomaly or birth defect, among other outcomes. [3]

Connect pregnancy follow-up to the safety-assessment process, so potentially reportable events reach the safety team and a recorded pregnancy receives the appropriate assessment before any SUSAR submission. A protocol-defined notification can be broader than an expedited regulatory-reporting category. Preserve which action was required and why.

A useful test case is a pregnancy exposure with no reported adverse outcome, followed by new information that needs safety assessment. The workflow should retain the original exposure and follow-up history, surface the new information to the appropriate owner, and support the applicable reporting decision. Determine the submission deadline from the applicable reporting category and jurisdiction.

Plan the handoff before the parent trial closes

The parent trial's database lock, treatment completion and pregnancy-follow-up completion are different events. E21's draft recognition of follow-up beyond trial completion makes that separation explicit as a planning issue. [1] A final data cut serves a particular analysis; later outcomes may remain scientifically relevant.

Before closeout, the sponsor should know which follow-up remains outstanding, where it will be performed, who can contact participants or obtain records, and how subsequent safety information will be assessed. If a separate study or registry is selected, confirm that its actual scope and governance meet the intended need. Confirm that the registry accepts the participant, captures the required outcomes and supports the necessary safety-process handoff.

For an EClinCloud EDC or data-management service discussion, bring the approved follow-up requirements, consent boundaries, access model and planned post-closeout process. The relevant demonstration is whether the proposed configuration and handoffs preserve those requirements. Consent decisions, registry arrangements and follow-up completeness each require evidence beyond a platform's availability.

Define the pregnancy-specific question before choosing the breadth of data collection under E19 and planning against the E21 draft. Preserve the information needed to answer it, with justified burden, appropriate consent and a follow-up mechanism that can outlast the original trial where necessary. E21's draft is useful for that design conversation precisely when its recommendations and limitations remain visible.

Sources

1. ICH/FDA. E21 Inclusion of Pregnant and Breastfeeding Women in Clinical Trials. ICH Step 2 draft endorsed May 14, 2025; FDA-hosted draft checked October 11, 2026. Sections 4.2.6–4.2.11, 4.4 and Appendix 2, ICH printed pages 11–18 and 30–32. © ICH. This article paraphrases and analyzes the draft independently; its implementation proposals carry no ICH endorsement and it serves no role as an official translation.

2. FDA/ICH. E19 A Selective Approach to Safety Data Collection in Specific Late-Stage Pre-Approval or Post-Approval Clinical Trials. Final FDA guidance, December 2022. Sections I.C, II.A, II.D and II.G, corresponding to ICH 1.3, 2.1, 2.4 and 2.7: scope, continued care/reporting duties, generally collected data and qualified selective-collection provision, prospective regulatory agreement on justification and implementation.

3. Electronic Code of Federal Regulations. 21 CFR 312.32—IND safety reporting. Paragraphs (a) and (c), accessed October 11, 2026. US reporting definitions and criteria, not a universal international reporting rule.