External Data and Reconciliation

Part of Clinical Project Management (PM) Training: Vendor and CRO Oversight.

About 1 minutes. Free, in your language, no account needed.

Data arrives from places the site never touches

Central laboratory results, ePRO, wearables, imaging reads, IRT randomization records and the safety database all produce trial data that no site enters. Each is a transfer, and each transfer can lose or duplicate records silently.

Reconciliation is the control

Counting is the crude version and it is still valuable: the number of samples the site says it sent against the number the laboratory received against the number that reached the database. A discrepancy of three is a question with an answer.

The one that matters most is safety reconciliation: serious adverse events in the clinical database against cases in the pharmacovigilance system. A mismatch means either an event was not reported to safety or a case exists that the clinical data does not know about. Both are serious and both are found only by looking.

Schedule it, do not leave it to close-out

Reconciliation deferred to database lock finds problems at the point they are most expensive and least fixable. A participant who can still be contacted in month four cannot necessarily be contacted in month twenty.

Monthly or quarterly, with a record of what was found and what was done, is the shape that works.

When the numbers will not reconcile

Do not resolve a discrepancy by adjusting a number. Find the record. A sample that left the site and never arrived is a chain-of-custody event with a cause, and the cause is usually something that will happen again.