Clinical trial data management is the administration and supervision of tasks associated with the entry, transfer, and preparation of source data and derived items for entry into a clinical trial database, according to the CDISC Glossary definition used by the NCI Division of Cancer Prevention [3]. In practice, sponsors split that broad definition into at least four working layers: what data will be collected and where, which personnel may enter and change those records, how discrepancies are resolved and monitored, and which standardized datasets and metadata must accompany an eventual regulatory submission. FDA guidance treats data handling, integrity, and confidentiality as sponsor obligations, with quality control applied to the relevant stages of data handling so results can be relied upon [1]. The issue is not whether a team manages data, but whether it can show, in advance and afterward, that the workflow was prespecified, documented, and auditable.

Prespecification Distinguishes Data Management From Recordkeeping

The first decision point is whether a sponsor can identify, before enrollment begins, exactly which data elements will be collected and by what method. Under E6(R3), sponsors should prespecify the data and collection method in the protocol; where necessary, additional detail such as a data flow diagram belongs in a protocol-related document, often a data management plan [1]. That requirement pushes operational decisions, such as whether a laboratory value arrives by direct entry, device export, or abstraction from an electronic medical record, into a written specification rather than a later fix.

A data management plan is not a static policy memo. The Society for Clinical Data Management, cited in a peer-reviewed overview, describes the plan as protocol-specific and comprehensive, covering definition, collection, processing, and final archiving or disposal, and the authors note that it is an ongoing iterative document because trial protocols frequently undergo amendments [5]. The CP-CTNet network illustrates the layering: a master plan applies to all studies, while study-specific plans are then developed by the Data Management, Auditing, and Statistical Center and the lead academic organization [3]. A reviewer can therefore ask two separate questions: does the master plan have a version date, and does the study-specific plan name the actual collection locations, roles, and systems for this protocol?

Entry Rights, Correction Rights, and the Blinded Boundary

Once collection is specified, sponsors need an access architecture that separates who may create a record from who may alter one. E6(R3) states that sponsors should not change data entered by investigators or trial participants unless the change is justified, agreed upon in advance by the investigator, and documented; error corrections requested by investigators or participants should also be justified and supported by source records from around the time of original entry [1]. That constraint gives investigators and participants standing in the data flow, rather than treating the database as entirely sponsor-owned.

Blinding adds a second boundary. Procedures should state who was unblinded, when, and for what purpose; who should remain blinded; and what safeguards preserve the blinding [1]. At the site level, the University of South Alabama Clinical Trials Office requires that only staff delegated by the principal investigator on the delegation of authority log may enter data, with documented training on file [8]. CP-CTNet similarly states that the accruing lead academic organization or principal investigator designee is responsible for reviewing and approving entered data, and that procedures for maintaining blinded data integrity must be established per study [3]. Entry rights and approval rights are therefore separate, and sponsors should be able to show both.

Hypothetical Walkthrough: A Device Study With an External Monitor Feed

Consider a suggestion, not a real study, to show how the layers interact. A sponsor runs a multicenter trial of a wearable monitor that exports heart-rate intervals as comma-separated values. The protocol prespecifies the device model, export frequency, and the exact variables to be collected [1]. A data flow diagram in the data management plan shows the path from device export to a trial master file, then to manual entry into an EDC, then to SDTM [1]. The principal investigator at each site delegates two coordinators to enter data, while a third holds read-only access for query responses [8].

A centralized monitor sees an out-of-range pulse value attributed to a participant in a blinded arm. The monitor issues a query without viewing treatment assignment; the site verifies against the source export and requests a correction supported by the original record [1]. The data management plan records that the correction was agreed to in advance and documented [1]. At database lock, the derived heart-rate endpoint is programmed into ADaM with source code submitted alongside the BLA or NDA dataset package [6][7]. This example is a synthetic illustration of how the cited documents connect; it does not describe an actual trial or testing outcome.

Query Resolution and Monitoring Are Time-Bound Workflows

Verification is not an afterthought; it has deadlines. Under CP-CTNet, data entry should be completed in Medidata Rave by the accruing organizations within 14 days [3]. The University of South Alabama SOP instructs sites not to leave fields blank when data are unavailable; instead, sites should indicate the reason, using "UNK" for unknown, "ND" for not done, or "NA" for not applicable [8]. The University of South Alabama SOP is stricter for its own workflows: queries should be resolved no more than five business days from issuance, and all entries should be reviewed against the source document for accuracy [8]. These are operational thresholds at two different institutions, not universal regulatory limits, and sponsors should not substitute one for the other.

Monitoring is the oversight of whether the trial is conducted, recorded, and reported in accordance with the protocol, SOPs, GCP, and applicable requirements, with a monitoring plan describing strategy, methods, and responsibilities [1]. FDA's electronic record guidance adds a risk-based validation focus: sponsors should consider a system's intended use, the importance of the data it holds, and its potential to affect participant rights, safety, or reliability of results [4]. A centralized monitoring approach may review metadata and aggregated discrepancies, but E6(R3) does not prescribe a single monitoring method; it asks that the activities be sufficient to oversee progress and compliance [1].

Regulatory Submission Turns Databases into SDTM and ADaM

The same trial data move through different forms depending on their purpose. FDA's Study Data Technical Conformance Guide recommends that sponsors implement the Study Data Tabulation Model (SDTM) for representing human clinical trial tabulation data prior to the conduct of the study [2]. SDTM gives a standard structure for tabulation datasets, while ADaM datasets support clinical study reports, integrated summaries of safety and efficacy, and other analyses; ADaM may contain imputed or derived data built from SDTM [2]. These are not interchangeable outputs. As the June 2026 guide states, SDTM and ADaM datasets should not share the same dataset label, such as the AE and ADAE pair both being labeled 'Adverse Events' [7].

Submission is also a metadata exercise. Each ADaM dataset should have its contents described in the define.xml file and, as appropriate, the analysis data reviewer's guide [7]. FDA recommends that sponsors discuss key data elements and dataset organization with the review division as early as the pre-IND meeting, because not every data element in a standard is fit for purpose for every trial [6]. Source code should be provided for creating all ADaM datasets, tables, and figures associated with primary and secondary efficacy analyses, in single byte ASCII text, with executable extensions excluded [6]. A data management plan therefore extends into regulatory conformance, not just site operations.