Clinical study teams rarely work from a single, static dataset. A data cut moves from an EDC export to SDTM, then to ADaM, and then into tables, figures and listings. A lab transfer may arrive after a monitoring visit. A define.xml file may be updated to match a derivation. A protocol or SAP may be revised. In the middle of this activity, one habit causes a surprising amount of trouble: treating the newest file as the only file. When the previous version is overwritten or renamed, the team loses the ability to answer a basic question with confidence: what exactly changed between the last cut and this one? A clinical data management software workspace such as Limora keeps each study in its own workspace and treats each data cut as its own version, so review can focus on change rather than on reconstructing what was lost.
Checksummed revisions give every cut a fixed identity
The first defense against overwriting is simple to state and easy to skip when files are moving quickly: every upload should become a new revision, not a replacement. Limora describes its data cuts as being kept in order with checksums. Each upload is a new revision with its own checksum, nothing is overwritten, and each person chooses which cut they are looking at. That matters because study team members often discuss a file in shorthand. One person says “the ADaM cut,” another thinks of the SDTM transfer, and a third is looking at a lab refresh that arrived later. A checksum gives the file a fixed identity. It does not explain which rows changed, but it removes ambiguity about which exact revision is under discussion. It also supports a practical habit: when a question is raised about a dataset, the team can point to the specific revision rather than to a folder name or a timestamp. Records are retired or superseded rather than removed, and changes are written to an append-only audit trail. That combination gives reviewers a stable chain of custody while keeping the current cut easy to find.
Compare mode turns a new cut into a focused review list
A checksum tells you that two files are different. Compare mode tells you where the difference sits. In Limora, a reviewer can open any dataset, filter to the subjects they care about, and switch on compare mode to see what changed between cuts. That sounds like a small convenience, but it changes the shape of a review meeting. Instead of opening two exports side by side and scrolling through subjects, the team can start with the subjects who have open queries, recent lab transfers, or a known derivation update. A data manager can look at the records tied to a site visit. A biostatistician can focus on the ADaM dataset after a derivation change. The review becomes finite: these subjects, this dataset, these revisions, these differences. The prior cut remains available, so when a difference is found, no one has to ask whether the old value was lost. The team can compare against the exact revision they reviewed before and then attach a note or issue to the exact form, field, row or document page it concerns. That is the difference between a change log and a working review: the finding stays next to the data it affects.
A practical routine for the next data cut
Teams do not need a complicated process to get value from checksummed revisions and compare mode. They need a repeatable routine that keeps the old cut visible and makes the new cut reviewable.
- Register the new cut as a revision. Confirm that the prior cut is still available and that the new file has its own checksum. Do not rely on a filename to prove identity.
- Choose the baseline deliberately. Before comparing, decide which earlier cut the team is reviewing against. Different questions may call for different baselines, so state the choice rather than assuming it.
- Filter before comparing. Use subject filters to narrow the review to the records that matter for the meeting: open issues, recent transfers, affected sites, or a dataset that changed after a derivation update.
- Review differences in context. When a change is expected, note the reason. When it is unexpected, raise an issue at the exact record or page. Keep the discussion attached to the data, not in a separate spreadsheet.
- Respect access and blinding. People see only the studies they belong to, with viewer, collaborator or owner roles. Unblinded content needs its own grant; without it, those files, records and search results stay hidden.
The routine also protects the team from a common failure mode: reviewing only the final export. If a change is made after the previous cut has been overwritten, the team can confirm the current value but not the history behind it. With revisions kept in order, the history remains available. The append-only audit trail covers changes to membership, roles, unblinding and records, while each file revision is kept. That is not about slowing every cut down. It is about making the cut reviewable when a question arises weeks later.
When a change is expected, record why
Not every difference between cuts is a defect. New subjects are enrolled, queries are closed, lab data arrives, and analysis datasets are rebuilt after a derivation update. The problem is not that changes happen. The problem is proving which changes were expected and which were not. A study workspace helps teams keep that distinction visible. Protocols, SAPs and annotated CRFs are searchable page by page next to the data they describe, so a reviewer can check a definition without leaving the study context. Notes and issues attach to the exact form, field, row or document page, so a comment about a changed lab value does not drift away from the record. Each data cut remains its own version, and each person chooses which cut they are looking at. When a biostatistician asks why a subject’s baseline differs from the previous cut, the team can compare the revisions, inspect the source record, and record the answer in context. The old file is still there. The new file is not a replacement for it. That is the practical value of checksummed revisions and compare mode: the team can move forward to the current cut without losing the evidence of what changed along the way.

