Clinical data manager role: from CRF design to database lock

Where the job sits in a trial

A trial's data passes through a fixed chain. Sites enter patient data into an electronic data capture (EDC) system; data management checks and cleans it; medical coders map free-text terms to standard dictionaries; statistical programmers convert the locked data into analysis datasets; biostatisticians analyse it. The CDM owns the middle of that chain and is accountable for its quality.

Two documents define the work. The data management plan (DMP) says how data will be collected, cleaned, coded, reconciled and locked for this study. The data validation plan (sometimes called the edit-check specification) lists every automated check. Both are written early and revised under version control, because an auditor may later ask why a check existed or did not.

Standards matter. CRFs are commonly designed following CDISC CDASH, and the collected data is later mapped to SDTM for regulatory submission; see CDISC's standards overview. Adverse events are usually coded with MedDRA and medications with WHODrug, using the dictionary version specified in the plan.

The artefact: rows from an edit-check specification

This is the kind of table a CDM writes and a study builder programs into the EDC. The study, forms and thresholds are hypothetical.

Each row answers five questions: which form, what fires the check, what the site sees, whether the site can save the value anyway (soft) or not (hard), and whether a human reviewer handles it instead of the system. Writing the site message clearly is underrated. A vague message produces a vague answer and then a second query.

After the build, every check is tested with deliberately bad test data before the study goes live. That user acceptance testing record is filed in the trial master file.

A fortnight of query work

Mid-study, a CDM's calendar is dominated by data cleaning:

• Reviewing the open-query listing by site and chasing sites where queries are ageing past the agreed turnaround.
• Running listings that automated checks cannot cover, such as an adverse event of "vomiting" alongside a concomitant medication for nausea started a week earlier.
• Reviewing coding: a site wrote "sugar high" and the auto-coder could not match it; the coder proposes a MedDRA term and the medical reviewer confirms.
• Reconciling serious adverse events between the EDC and the safety database, and lab values between the EDC and the central lab's data transfer.
• Joining the weekly study team call to report data status to the project manager and clinical research associates (CRAs).

Query ageing, lock dates and inspection findings

• Query ageing and volume — how many are open and for how long.
• Timeliness of data entry by sites, which the CDM monitors and escalates.
• Milestones met — first patient in, interim cut-off, last patient last visit to database lock.
• Audit and inspection findings related to data management.
• Reconciliation completeness for external data and safety.

Database lock, in order

1. 1All expected data received and entered; missing pages chased or documented.
2. 2All queries answered and closed.
3. 3Coding complete and approved.
4. 4External data and serious adverse event reconciliation complete.
5. 5Investigator sign-off of casebooks where required.
6. 6Lock checklist signed by data management, biostatistics, clinical and medical leads.
7. 7Database locked and edit access removed. Any later change needs a formally documented reopening of the database.

CRAs, coders, safety and biostatistics

A CDM reports to a data management lead or manager, usually within a study team run by a project manager. Colleagues include CRAs, who visit sites; the study builder or EDC programmer; medical coders; safety physicians; statistical programmers; and biostatisticians.

Not the same as a clinical research associate

CRAs monitor trial conduct at sites: consent, protocol adherence, source data verification. CDMs work centrally on the data. People confuse the two because both deal with sites all day. The statistical programmer, who takes the locked data further, is covered in the statistical programmer guide.

Rave, InForm and discrepancy management in postings

• "EDC experience (Medidata Rave, Oracle InForm, Veeva)" — familiarity with a named platform.
• "Query management, discrepancy management" — core cleaning work, often the entry role.
• "DMP, DVP, edit-check specification" — design responsibility; more senior.
• "MedDRA/WHODrug coding" — a coding-focused role, or coding oversight.
• "GCP, 21 CFR Part 11" — regulatory awareness expected; ask what training is provided.

From query reviewer to study data lead

Month one: complete GCP and SOP training, and shadow query review on a live study. Month two: own query management for a few sites. Month three: write test scripts for a new study's checks. Next roles include senior CDM, data management lead, clinical data scientist, or a move into statistical programming. If the analytical side draws you, the Data Science programme builds the Python, SQL and statistics that make that move realistic.