API batches: separate production completion and quality release
An active pharmaceutical ingredient batch can be physically complete while its records, tests or investigation remain under review. Completed manufacturing quantity therefore does not by itself establish released inventory available for distribution. Research needs batch identities and quality dispositions, rather than a general order-backlog roll-forward.
The deliverable is a batch-state ledger that separates manufacturing completion, sampling, laboratory work, deviations, quality review, release and rejection. Its purpose is operating-evidence reconciliation. It does not approve a batch, interpret a test result medically or recommend manufacturing or retesting procedures.
Locate the guidance and its scope
For this US-focused research example, use the publicly available FDA Q7A Good Manufacturing Practice Guidance for Active Pharmaceutical Ingredients, dated August 2001, as read on October 25, 2026. Its introduction states the guidance’s nonbinding status; Scope 1.3 identifies its API coverage and limits. It is not a universal finished-drug batch-release rule.
Responsibilities of the Quality Unit(s), section 2.2, and Batch Production Record Review, section 6.7, identify review and release responsibilities. Section 6.6 concerns laboratory records. These passages support distinguishing completion from quality disposition; they do not impose the invented turnaround times below or establish a company’s actual compliance.
Keep the applicable product, manufacturing site and material stage beside each batch. An intermediate remaining within the manufacturer’s control, a distributed API and a finished drug are not interchangeable categories. Do not extend this worksheet to a different product or jurisdiction without reading the relevant sources and the actual company’s procedures.
Rebuild a fictional completed cohort
Assume twelve API batches, each with one thousand kilograms physically present at a reporting cutoff. Manufacturing has completed for all twelve. Eight batches have documented release, three are awaiting deviation investigation and review, and one has a documented rejected disposition. No batch has yet shipped or been destroyed in this teaching snapshot.
| Disposition | Batches | Kilograms | Research treatment |
|---|---|---|---|
| Released | 8 | 8,000 | Passed the defined quality-release gate |
| Pending investigation/review | 3 | 3,000 | Not counted as released |
| Rejected, physically retained | 1 | 1,000 | Present but excluded from released stock |
| Total physical cohort | 12 | 12,000 | All recorded quantities reconciled |
The physical total closes at twelve thousand kilograms, while released stock is eight thousand. Calling all twelve thousand saleable because production is complete erases the unresolved and rejected states. Conversely, calling pending batches failed would turn an incomplete decision into a conclusion unsupported by the disposition record.
Released does not establish an actual sale. Customer qualification, order specifications, destination requirements, acceptance and shipment timing can remain separate gates. In this example, eight thousand is release-eligible inventory under the stated batch evidence, not proof that a buyer will take all of it or that revenue has been recognised.
Follow state transitions without new production
At the next cutoff, suppose two of the three pending batches receive documented release and the remaining pending batch enters an authorised rework status. The original rejected batch remains physically retained and rejected. Assume no quantity loss, shipment or other new material.
The ledger now has ten released batches, one pending in rework and one rejected, still twelve batches and twelve thousand kilograms. Released stock has increased by two thousand to ten thousand. This is a state transition of already manufactured material, not two thousand kilograms of additional manufacturing output.
If rework splits or combines material, preserve parent-child batch identifiers and a mass bridge. A new administrative batch number does not create new kilograms. The example does not assume that any particular rework is allowed or successful; the evidence must establish the actual decision and final disposition.
Suppose three released batches are subsequently shipped. With all other assumptions unchanged, released on-hand stock falls to seven thousand kilograms, while cumulative released volume for the cohort remains ten thousand. Use different columns for cumulative release, available on hand and cumulative shipments. Adding those totals together would count the same material more than once.
Rebuild time at the correct stage
For every batch record manufacturing completion, sample receipt, test completion, investigation milestones, review and disposition dates where disclosed. Elapsed completion-to-release time includes different activities; a long interval does not reveal which activity was constrained without those timestamps.
Assume the eight initially released batches have completion-to-release intervals of two, three, three, four, four, five, seven and twelve days. Mean completed interval is five days and median four. Three still-pending batches have ages of ten, fifteen and twenty days at the same cutoff. Those ages are not completed release times and should not disappear from the picture.
Reporting only the mean of the eight released batches selects completed cases. Show the pending count and age distribution beside it. Do not average twenty days pending with a completed interval as though both measure the same event. Later closure permits the batch’s full elapsed interval to replace its age in the completed-cohort analysis.
Test a different mechanism: retesting and investigation
Suppose a batch receives another laboratory test while its deviation investigation remains open. An additional test record alone does not establish release, a failed product or resolution of the investigation. Preserve the original result, the investigation record and the authorised final disposition as separate evidence.
The worksheet must not recommend repeating tests until a favourable result appears or infer clinical safety from a laboratory summary. Its task is to trace what the quality unit documented and which records support the state transition. If the actual record is unavailable, leave release status unknown rather than invent a passed test or decision.
A larger backlog can also arise from more completed batches entering review, not solely slower laboratory work. Compare arrivals, dispositions and pending ages for the same product and site. A public segment inventory balance without batch states cannot independently identify a quality-release queue.
Use the batch verification sheet
| Input | Obtain | Check |
|---|---|---|
| Identity and mass | Batch register and location records | Stable IDs and parent-child links |
| Manufacturing | Completion records | Physical completion distinct from release |
| Laboratory | Sample, test and review timestamps | Results not substituted for final disposition |
| Deviations | Investigation and resolution records | Open matters not silently removed |
| Quality decision | Dated release/rejection approval | Scope, material and authorised status |
| Inventory and shipping | Quantity movements | No double counting of rework or cumulative release |
When systems conflict, match batch number, product, site, unit and cutoff before interpreting status. An enterprise inventory label such as finished goods may be an accounting classification, not the quality system’s release state. Seek the actual reconciliation rather than let one software label overrule the other.
If public disclosure lacks batch records, report only the narrower claim supported by the issuer. A hypothetical queue can show what evidence would matter, but must not be presented as a reconstruction of confidential quality files. Missing review dates do not prove misconduct, nor does a fast average establish compliant handling of every batch.
Update after documented dispositions, quantity adjustments, shipment, destruction, rework decisions or procedure changes. Preserve the previous state and decision date so a later release does not retroactively appear available at an earlier reporting cutoff. The conclusion should state which kilograms passed which gate, what remains pending and what the evidence cannot establish about sale or product quality.
Close a rework mass branch
Return to the second snapshot before any shipment: ten thousand kilograms released, one thousand pending rework and one thousand rejected. Assume later authorised records show 980 kilograms from the pending parent batch released under a linked child identifier and twenty kilograms recorded as process loss. This is an invented documented outcome, not a recommendation or an assumption that such work is generally permitted.
The new reconciliation is 10,980 released + 1,000 rejected + 20 recorded loss = 12,000 original kilograms. The 980 is not new manufacturing added to the original twelve thousand. Keep the parent batch, child batch, disposition and loss record together. If the records do not provide the quantity conversion or approval, leave the reworked amount pending instead of creating a saleable child batch from a status label.
FAQ
Is physically complete API automatically released inventory?
No. The relevant quality records and documented disposition are separate evidence.
Does a long pending age prove the batch failed?
No. It identifies an unresolved interval, not the final quality decision or a medical outcome.