Reviewing and Endorsing Reported Data

Part of Principal Investigator (PI) Training: The Data You Sign For.

About 4 minutes. Free, in your language, no account needed.

Why this matters

At some point in every trial an electronic system will present you with a list of forms and a button that applies your signature to all of them. What that button means is that you have reviewed the data reported from your site and you are standing behind it.

It is the most under-considered signature in clinical research, and it is the one that connects your name to every number the sponsor analyses.

What the guideline asks

Three sub-sections combine.

§2.12.3 — access and review. The investigator should be given access to, and is responsible for reviewing, data relevant to the trial. This explicitly includes data from external sources that affect participant eligibility, treatment or safety — central laboratory results, imaging reads, adjudication outcomes, data from participant-held devices — unless access is restricted to protect blinding.

That sub-section closes a real gap. Data generated outside your site still describes your participant, and you cannot manage someone's safety on data you never see. If a source of external data is not reaching you, that is a start-up conversation with the sponsor.

§2.12.5 — endorsement. The investigator should review and endorse the data reported to the sponsor at milestones agreed with the sponsor. Two things follow: the milestones are agreed rather than imposed, and endorsement is an act of review rather than an administrative sign-off.

§2.12.6 — accuracy. Reports should be accurate, legible and complete, and consistent with the source records. Corrections are traceable, do not obscure the original entry, and are explained where necessary.

What a real review looks like

Reviewing every field of every form is neither expected nor useful. Reviewing by risk is both. A practical order of attention:

  1. Eligibility data — the criteria that required judgment, and their source.
  2. The primary endpoint — every value, every participant. This is the data the trial exists to produce and a transcription error here is the most costly error your site can make.
  3. Safety data — every adverse event: term, dates, severity, seriousness, causality, outcome. Check that your assessments were transcribed as you made them.
  4. Concomitant medications — particularly anything prohibited, and anything that started around the time of an event.
  5. Protocol deviations — that each one you know about appears, and that nothing appears you do not recognize.
  6. Everything else — sampled, not exhaustive.

Then check the two consistencies that matter: reported data against source, and forms against each other. An adverse event with an onset date before the participant's first dose, a concomitant medication for a condition that appears nowhere in the medical history, a visit date outside the window with no deviation logged — these are the discrepancies that show up in a cross-form read and never in a field-by-field one.

Batch signing, and why it is a problem

The pattern is familiar. Sixty forms accumulate; a system prompts; a signature is applied to all of them in one action late on a Friday.

Nothing prohibits signing many forms at once, and at a busy site it is often the only workable rhythm. The problem is signing forms you have not looked at. The remedy is not heroism, it is scheduling: make data review a recurring appointment rather than an interruption, and keep the batches small enough that each one is genuinely reviewable.

If a batch is too large to review, that is information — about staffing, about the milestone schedule, or about how far behind the site has fallen. It is worth saying so to the sponsor rather than signing.

Queries are data too

Data queries raised by the sponsor's data management team come to your site continuously, and they are usually answered by a coordinator. Two categories should not be:

Those responses are yours. A query that asks you to reconsider a causality assessment is a question to a physician, and the answer may reasonably be that your assessment stands.

Key takeaways