Phase I study build is the configuration phase that sets the structural conditions for everything that follows: query volume, data completeness, CDISC mapping complexity, and ultimately the study closeout timeline. Most Phase I program delays that appear to originate in data management, site monitoring, or regulatory submission can be traced back to a configuration decision made during study build — or a decision that wasn't made and defaulted to a system behavior that didn't match the study's needs.
The build window for a typical Phase I first-in-human oncology program is compressed. Protocol finalization, IRB submission, site activation, and study build often run in parallel, with the expectation that the EDC and IRT will be ready for enrollment when the first site is activated. In this environment, the study build decisions that require the most lead time — DMP alignment, visit schedule configuration, and IRT integration specification — are precisely the ones most likely to be deferred.
This guide covers the configuration decisions that have the highest downstream consequence for Phase I programs, with specific attention to FIH oncology designs running 3+3 dose escalation or model-based dose escalation approaches such as BOIN or mTPI.
DMP Alignment Before EDC Build Begins
The data management plan governs the study build — not the other way around. EDC configuration that begins before the DMP is finalized creates build-time decisions that may conflict with DMP requirements once the DMP is written, requiring eCRF amendments to reconcile. Those amendments, executed on a live study, consume change management time and introduce version control complexity in the audit trail.
The DMP for a Phase I FIH program specifies the eCRF structure decisions that the EDC build must implement: which forms are required versus conditional (a tumor assessment form that only applies to subjects with measurable disease at baseline), which visit windows trigger data entry reminders, how dose modification records are captured (whether dose reductions are entered as new EX records or as modifications to the existing exposure record), and how the dose-limiting toxicity assessment form is structured relative to the evaluation window specified in the protocol.
Practically, this means the CDM team and the EDC builder should have a DMP alignment session before the eCRF library is populated. That session needs to produce decisions on at minimum: the visit schedule and window definitions, the conditional form and field logic, the SAE eCRF form structure relative to the safety database, and the cohort tracking approach for the dose escalation design.
Visit Schedule Configuration for Dose Escalation Designs
Phase I dose escalation studies — whether using 3+3, BOIN, or mTPI-2 designs — have visit schedules that don't follow the regular interval pattern of Phase II programs. Dose escalation visit schedules typically include a dose-limiting toxicity (DLT) observation period per cohort (commonly 21 or 28 days), a cohort evaluation visit that determines escalation or de-escalation, and then a continuation phase for subjects who proceed to the next cycle.
The EDC visit schedule must reflect the actual protocol-defined visit structure, not a simplified approximation. A common build shortcut is to configure Phase I visits as a generic repeating cycle structure (Cycle 1 Day 1, Cycle 1 Day 8, Cycle 1 Day 15, Cycle 2 Day 1, and so on) without capturing the DLT observation period as a discrete scheduled visit. That shortcut creates a data management problem: DLT assessments that occur outside the standard cycle structure don't have a designated visit slot, and coordinators enter them against the nearest scheduled visit or create unplanned visits, generating edit check queries and complicating the DLT analysis dataset structure.
The correct approach is to configure the DLT observation period as a distinct visit type with its own eCRF forms — DLT assessment form, vital signs, ECOG performance status, and any scheduled labs. When the EDC visit schedule mirrors the protocol visit schedule exactly, data entry follows naturally and the downstream CDISC mapping to the SDTM TA (Trial Arms), TE (Trial Elements), and TV (Trial Visits) domains is straightforward.
Cohort Tracking and the Dose Escalation Decision Workflow
BOIN and mTPI-based designs make real-time escalation decisions based on the observed DLT rates in the current cohort. The escalation decision requires that the clinical team has a current, accurate view of which subjects are in the DLT observation window, which have completed DLT assessment, and what the observed DLT count is for the current dose level.
EDC configurations that require the CDM team to generate ad-hoc queries for this information slow the escalation decision workflow at the moment when the team needs it fastest. The study build should include a cohort status view — a summary display of active dose level, subjects in DLT window, DLT assessments completed, and DLT count — generated from the eCRF data in real time. Building this view requires that the dose level assignment field in the eCRF is structured as a queryable variable (not a free-text field), that DLT assessment form completion is tracked by the visit completion flag, and that the DLT classification field uses controlled terminology that can be aggregated programmatically.
We're not saying every Phase I EDC needs a full escalation decision support dashboard — for a 3+3 design with three to six subjects per cohort, a simple subject-level data extract reviewed by the medical monitor is adequate. The point is that the data elements needed for the escalation decision should be explicitly configured as structured, queryable fields at study build, not inferred from free-text adverse event narratives.
IRT Integration: Specification Before System Configuration
For Phase I studies using an IRT or IWRS system for drug dispensation, the integration specification between the EDC and the IRT should be completed before either system is built. Retrofitting the integration — mapping an IRT subject identifier scheme to an EDC subject number scheme after both systems are live — is the most common source of IRT-related delays in Phase I study execution.
The integration specification needs to establish at minimum: how the USUBJID in the EDC corresponds to the subject identifier in the IRT, whether randomization (if applicable) is triggered by the EDC through an API call to the IRT or manually by the site in the IRT after an EDC confirmation, how kit dispensation records flow from the IRT to the eCRF EX domain, and what the reconciliation procedure is when IRT kit records and eCRF dispensation records disagree.
For FIH programs using a sentinel dosing design — where the first subject in a cohort is observed for a defined period before additional subjects are dosed — the IRT configuration must support this sequencing. The IRT should be configured to hold subsequent randomizations at a dose level until the sentinel observation period is cleared, which requires a flag or status field in the IRT that the DM team or medical monitor controls. IRT systems that don't support this logic natively require a manual workaround protocol that must be documented in the DMP and communicated to all sites before enrollment begins.
eCRF Form Design for Phase I Safety Data
Adverse event data quality in a Phase I study determines the quality of the safety package, and safety package quality in a Phase I determines the IND annual report, the FDA's review of the dose escalation, and the regulatory basis for Phase II. Getting the AE eCRF form right matters more in Phase I than in almost any other study phase.
The AE form should capture: verbatim term as reported by the subject or investigator, onset and resolution dates (with partial date handling explicitly configured per the DMP), CTCAE grade (with version number specified — CTCAE 5.0 is now standard for most oncology Phase I programs), relationship to study drug (using the controlled list specified in the protocol), seriousness criteria (with each criterion as a separate checkbox field, not a single "SAE yes/no" field), and action taken with study drug.
The common build error is to configure CTCAE grade as a free-text field or a generic 1-5 dropdown without linking it explicitly to the CTCAE 5.0 grading criteria. When investigators use CTCAE 5.0 at one site and CTCAE 4.03 at another — because the form doesn't specify — the safety dataset requires a coding adjudication pass before analysis that adds a week to the closeout process.
User Acceptance Testing Before First Subject In
User acceptance testing (UAT) for a Phase I EDC is the last gate before enrollment opens. UAT verifies that the eCRF forms and edit checks behave correctly for a representative set of subject data scenarios — including the edge cases that are specific to the dose escalation design.
A UAT script for a BOIN-design Phase I program should include at minimum: a subject who completes DLT window without a DLT, a subject who experiences a DLT in cycle 1 day 8 (triggering the DLT assessment form and the cohort decision workflow), a subject with a dose modification in cycle 2 (testing EX form update logic), and a subject who withdraws during DLT observation (testing DS form completion and the data lock condition for a pending DLT assessment).
UAT failures caught before first subject in cost an afternoon. UAT failures discovered after enrollment — when sites have been entering data for three weeks with incorrect form logic — cost amendment cycles, site retraining, and retrospective data correction work that must be documented in the deviation log. The asymmetry is decisive: protect the UAT window, even when protocol finalization is running late and the pressure to open enrollment is high.
The study build effort for a Phase I FIH program is measured in days of configuration work, reviewed against a protocol that took months to write. The quality of that configuration work determines the data collection, monitoring, and closeout quality for a study that may run 18 months or longer. It is not the place to cut time.