SDMS: instrument data, metadata and usable record retrieval

A search returns the expected file, yet the laboratory cannot reconstruct the analysis. Missing relationships, an absent method or the software needed to read the format may be the cause. For an SDMS, successful copying is therefore only one of the checks.

1. Define intended use and the SDMS boundary

A Scientific Data Management System can collect, index and make scientific data from instruments and applications retrievable. Its actual functions depend on configuration. A product name does not demonstrate completeness, adequate retention or compliance.

For each flow, specify who manages the sample, who acquires and processes data, who records activities and where the authoritative record resides. A LIMS may govern samples, a CDS chromatographic acquisition and an ELN experimental work. The SDMS retains or links the objects defined by the architecture. A shared folder does not automatically provide the same controls, but requirements and tests must establish the difference, rather than labels.

The general comparison is covered in LIMS, ELN, CDS and SDMS roles. Here the expected outcome is narrower: retrieving a complete record that remains usable for its purpose.

2. Identify the set that gives data meaning

Start with a representative analysis and trace the result to its origins. List native files, signals, acquisition and processing methods, sample identifiers, units, users, dates, versions and relevant audit trails. Metadata explain the data; relationships identify which components belong to the same activity.

Distinguish originals, verified copies and reports. A verified copy requires evidence of fidelity and completeness for its intended use. A report may represent only part of the record: a chromatogram PDF alone does not demonstrate that the dynamic dataset is available. Keeping the native file without its necessary dependencies may be equally inadequate.

Also document objects retained in the source system and how to retrieve them. A link to an external file does not mean the file has been captured: a changed path can remove an essential part of the record.

3. Control capture, errors and reconciliation

Define when an object is ready for capture and how to identify an incomplete transfer. A file still being written, an excluded folder or a later modification may escape an apparently normal process. Record the source, object, version, transfer status, exception and person responsible for resolution.

Reconciliation compares what the process produced with what the SDMS actually makes available. File counts alone do not establish that the right files are present. In an authorised, separate environment, test interruptions, duplicates, retries and association errors: a retry must neither create an indistinguishable second record nor silently overwrite the earlier one.

The following matrix is a design example to adapt to actual instruments and formats, not a universal regulatory requirement.

4. Test search, access and usable retrieval

Test realistic searches: sample identifier, time range, instrument and method version. Consider corrected identifiers and identical names. The index should distinguish records without retrospectively changing the meaning of the data. Define who may correct metadata and how the history remains visible.

Retrieval must support the intended activities, rather than merely downloading a container. Check permissions, a compatible reader, application versions, configurations and required licences. Confirm that display preserves units, precision, times and relationships. If a format is converted, demonstrate that value and meaning remain suitable for use; a successful export does not establish equivalence.

Retain the identity of the tested set, environment, reader version, expected result, observed result and deviations. The ability to repeat processing is a property to assess for the record concerned, not a blanket obligation to reprocess every retrieved dataset.

5. Simulated case: files present, method missing

A laboratory retrieves a historical sequence: signals and reports are present, but the linked processing method is unavailable. The captured-file count had passed because the collection rule covered only the signals folder.

The process owner records the exception, preserves available sources and assesses which decisions require the record. Together with IT and QA, they search for the authentic method version and investigate the extent of the problem across similar flows. They do not recreate the method from memory or present the PDF as an automatically equivalent substitute.

The correction covers the capture rule, the method–dataset relationship and a new documented test of complete retrieval. Closure must rest on evidence. If a component remains unavailable, that limitation stays explicit in the investigation and impact assessment.

Finally, assign responsibility for retention periods, periodic checks and format or system changes. Before retiring a reader, test the future access route. Operational backup, record retention and migration serve different purposes: the SDMS must fit a defined strategy, rather than replace it with an “archived” checkbox.

6. Sources and scope

Sources checked: 1 October 2026. The EU context here is a GMP laboratory for human medicinal products. The matrix and case are original editorial examples, not validated templates or tests performed on real systems.

• European Commission — Annex 11, January 2011 revision, §§4.3–4.8, 5, 7 and 17: flows, transfers, availability and retrieval.
• European Commission — Chapter 4, January 2011, §§4.1 and 4.10: documentary relationships and retention. Both in operation from 30 June 2011; the 2025 proposals are not treated here as adopted requirements.
• FDA — Data Integrity and Compliance With Drug CGMP, December 2018, Q1 and Q3: metadata, static/dynamic formats and intended use; nonbinding US guidance.