At Reliance Life Sciences, drug products were released using raw plate-count averages entered into a laboratory information management system (LIMS). The calculation omitted a dilution factor required by the firm’s procedures and the applicable compendial methods. FDA described the omission in its August 2026 warning letter.
The factor already had a place in the written method. It needed a place in the calculation used for release, too. For a laboratory choosing software, that is a much more specific problem than deciding whether a product supports calculations. Someone has to translate the method into the proposed configuration, then establish what another person will be able to inspect when the result is reviewed.
The factor in the method
The raw count and the reported concentration describe different things. A plate count records what was observed in the tested preparation; the dilution factor connects that observation to the quantity of original product being assessed. Averaging the counts does not supply that missing relationship. In its microbial limit test finding, FDA identified the intended result as colony-forming units per gram or milliliter of product and said the required factor had been left out of the LIMS calculation.
A calculation feature leaves several choices open. Consider a proposed configuration in which an analyst enters a dilution factor beside the plate counts. That field could be saved correctly while the calculation continues to use only the counts. A demonstration that the field accepts a value would establish data entry. The consequential step is what the calculation does with it.
The configuration could instead derive the factor from documented sample preparation. That would make a separate manual-entry field unnecessary for the factor. To understand the resulting value, a reviewer would need the preparation record and the calculation that used it. These are different ways of representing the same method dependency. Demanding a particular field before understanding the method could rule out a suitable implementation, while accepting a field without examining its use could approve an incomplete one. In either proposed implementation, the input has to mean what the calculation expects. An observed count and a result already adjusted for dilution would enter the arithmetic at different points. Applying the factor again to an adjusted result would change it a second time.
What an approval can establish
The obligations follow the use of the record. FDA’s Part 11 scope guidance connects applicability to records required under underlying FDA rules and how electronic records are used. Its stated enforcement discretion for specified electronic-record provisions leaves the underlying recordkeeping duties in force. For the drug-release work discussed here, the calculation and its supporting data remain part of the obligation.
For drug CGMP laboratory records, the relationship between an observation and a reported value is explicit in the record requirements. 21 CFR 211.194 includes the test method, the data obtained, calculations with units and conversion factors, and review of the original records by a second person. A result and a signature form only part of that account.
An approval workflow therefore has a substantive question behind it. The reviewer needs to assess how the method was applied to the sample. If the review view offers only the final value, the reviewer has to obtain the missing inputs elsewhere before checking the calculation. A signature can record that approval occurred; the surrounding evidence determines what there was to examine. This affects the work a proposed workflow actually asks the reviewer to perform.
The Reliance inspection also found a more basic break in the evidence. Samples recorded as collected and incubated could not be accounted for, and analysts confirmed to management that they had never been collected. The dilution-factor finding sat within wider laboratory and recordkeeping failures. A reconstruction can follow a calculation faithfully and still rest on a false account of the work.
The questions that shape a LIMS
In its roadmap for the Material Measurement Laboratory, NIST includes downstream data queries in workflow design. These are the questions scientists expect to ask of their data. Defining them helps determine the detail of the structured record and the workflow components required. For structured workflow design, it also describes the engagement needed between research specialists and data science engineers. The scientist’s definition of an experimental attribute has to be expressed in the data the system will capture.
Suppose a scientist needs to identify results produced with a particular method revision. A report may display a method name clearly enough for someone reading it, yet leave the revision inside an attached document. Answering the scientist’s question would then require opening those documents or establishing a structured connection to them. The query exposes a requirement that a report preview alone would leave unresolved. The scientist is searching across results. A release reviewer starts with a selected result and works back through its calculation to the inputs. In the dilution example, that includes recovering how the preparation was accounted for. A record organised for producing individual reports might support that investigation through document review while leaving the cross-result search to manual extraction.
Some of that work may stay in a controlled method document or another laboratory system. The purchasing decision includes whether the proposed links give a reviewer usable access to it. Copying every item into the LIMS could create another version to maintain; a reference that loses its connection to the applicable revision could leave the reviewer looking at the wrong instructions. The choice concerns how the laboratory will preserve that connection over time.
The result after the method changes
For computerized systems used in EU GMP activities, the current Annex 11 links user requirements to documented risk and GMP impact, calls for their traceability through the system lifecycle, and addresses controlled configuration changes. These provisions concern the system as it is used and changed. The Commission’s current listing identifies the January 2011 revision. The later consultation proposal, whose consultation closed in 2025, is a separate proposed text.
A method change gives record reconstruction another dimension. Suppose the laboratory revises a calculation after an earlier batch has been released. Opening that batch against the current calculation could obscure what the analyst and reviewer actually used at the time. Retaining the old reported value helps, but the explanation also depends on the applicable calculation and its inputs remaining identifiable. For a reviewer investigating an earlier release, the original calculation explains the decision made at the time. If the investigation finds an error and leads to a recalculation, that creates another account with its own basis. The reviewer needs to compare them, identify what changed and assess its effect on the release decision. Overwriting the earlier calculation would remove part of the explanation just when the laboratory needs to understand the consequences of its correction.