Clinical data review rarely moves in a straight line. A team receives a new cut, checks what changed, and decides whether the change is expected, affects a particular subject, or relates to an open query. Starting with a full-dataset comparison can bury that answer under irrelevant rows. A more reliable pattern is to filter to the subjects of interest first, then switch on compare mode. That keeps the review focused and makes each change easier to explain.
Start with the review question and the subject filter
Before opening compare mode, define the population you need to review. The trigger may be a few subjects with unusual visit dates, a specific lab finding, or subjects connected to an open query. Write it down. A short note such as “review subjects 1042, 1043 and 1051 for the Week 12 lab change” gives the team a shared scope and prevents the comparison from becoming a hunt through the whole study.
Subject filters earn their place here. In a study workspace, the team can browse forms, the visit schedule, datasets and documents with subject filters throughout. Filtering first narrows the data to the records that matter and makes the later comparison easier to read, because every row on screen belongs to the subjects under discussion.
Keep the filter consistent across the files you review. If the question concerns a subject’s lab value, look at the lab data, the related CRF form and any analysis dataset rows for that subject. If a document explains the rule behind the data, use page-by-page search for protocols, SAPs and annotated CRFs. Compare the same focused population across the files that can answer the question, not every file in the study.
Treat every upload as a checksummed revision
Compare mode is only trustworthy when the two sides are unambiguous. A clinical trial data platform should keep each data cut as its own version, with every upload recorded as a new revision with its own checksum. Nothing is overwritten, and each person can choose which cut they are looking at. That matters when team members have different cuts open. The checksum gives the cut an identity; the revision history shows that the file on screen is the file intended for review.
State the cut before you compare. “I am reviewing the current ADaM cut against the previous cut for subjects 1042, 1043 and 1051” names the scope, subjects and versions. If someone joins later, they can open the same cut and follow the same filtered path. Earlier cuts may remain available without being deleted, but the review should still identify which revisions are being used so findings are not attributed to the wrong upload.
The checksum also helps with handoffs. A review request can reference a revision rather than a folder name that may have been copied or renamed. That reduces the chance of comparing a new cut with an old file in the wrong place. It also supports the principle that records are retired or superseded rather than removed, so the history of a review remains visible.
Use compare mode after filtering, not before
Once the subject filter is in place and the cut is identified, open the dataset and switch on compare mode. The comparison should show what changed between cuts for the filtered subjects. Reviewing that smaller view is faster and less error-prone than scanning a full-study diff. Focus on the rows that differ, check the corresponding CRF form or lab source, and decide whether the change reflects a correction, a new entry, a coding update, a derived value or a question that still needs an answer.
A practical sequence is:
- Confirm the subject filter and the two cuts.
- Open the dataset and switch on compare mode.
- Read each change against the source form, lab data or document page.
- Decide whether it is expected, needs a query, or needs a note for the analysis team.
The comparison is not only a technical diff. It is a review event. A changed lab value may be explained by a corrected unit; a changed analysis row may trace back to a superseded SAP or a programming update; a changed form may answer an earlier query. The filtered comparison gives the team a place to make that judgment without losing the subject and cut context.
If the change affects a derived dataset, keep the source and derivation in view. The workspace brings CRF data, SDTM and ADaM datasets, define.xml, lab data, protocols, SAPs and TFL outputs together for the study. A reviewer can move between the changed ADaM dataset, the SDTM source and define.xml without leaving the study workspace. Compare mode narrows the question; surrounding files help answer it.
Record issues where the change appears
The review is not complete when the difference is understood. It is complete when the understanding is recorded where the next reviewer can find it. Notes and issues should attach to the exact form, field, row or document page they are about. If a changed lab row raises a question, the issue belongs on that row, not in a separate spreadsheet that must be re-matched later. If a SAP page explains a derived variable, the note belongs on that page.
In-context recording changes the rhythm of review. A reviewer can open compare mode for the filtered subjects, see a difference, open the related CRF form or document page, and raise the issue at the point of evidence. The next person sees the question where the data is, with the cut and subject filter still part of the context. That is more reliable than reconstructing the review from email threads or row numbers that may refer to a different revision.
Two controls deserve a final check. First, respect blinding. Unblinded content needs its own grant; without it, those files, records and search results stay hidden, whatever a person’s role. A filtered review should not become a back door to unblinded information. Second, changes to records, roles and unblinding are written to an append-only audit trail. The issue, note and comparison become part of a review history that should remain inspectable after the meeting.
Used this way, compare mode is a focused review method: filter to the subjects who matter, confirm the checksummed revisions, compare the cuts, and attach the issue to the exact place where the change appears. The team spends less time searching for context and more time deciding what the change means.
Lagos-born, Berlin-educated electrical engineer who blogs about AI fairness, Bundesliga tactics, and jollof-rice chemistry with the same infectious enthusiasm. Felix moonlights as a spoken-word performer and volunteers at a local makerspace teaching kids to solder recycled electronics into art.