The annual product quality review, drafted the morning after the period closes — every number cited
Drafted the morning after the period closes, every number cited.
See one case, screen by screen ↓Software that builds each product's yearly review — the US annual product review under 21 CFR 211.180(e) and the EU product quality review — from the records of the year. Annual Product Quality Review has seven agents collect the year, make a trend call on every attribute with its test, reconcile every count and draft the 14-chapter report, for the product quality lead to decide and sign.
How to choose AI solutions for life sciences →
Updated
Every licensed product needs its yearly review under 21 CFR 211.180(e) and EU GMP Chapter 1: every batch made and its disposition, release and in-process results, OOS and OOT investigations, deviations and CAPA, changes, variations, stability, complaints, returns and recalls, last year's actions, qualification status and the quality agreements. The product quality lead pulls that year out of ERP, MES, LIMS and the QMS into spreadsheets, one export at a time, before a single attribute is trended.
So the trending, the part inspectors read, is done last and thinly. A dissolution shift after a second cellulose supplier, a forming depth that drops after a sealing temperature change, a stability pull that sits above its prediction interval: these surface a year late. And with 20 to 40 reviews a year at a mid-size manufacturer, the 90-day clock is missed: last year 15 of 24 reviews were approved inside it.
Estimated split for one product's review, by hand and with the solution, from the 122 author hours by hand and the 41-hour target.
Seven agents collect the year, call a trend on every attribute and write the review; the product quality lead decides and signs.
PQR-2026-014 — Corventa 20 mg film-coated tablets, Plant 2 — Shannon, period Aug 1, 2025 – Jul 31, 2026, due Oct 29. 186 batches, 41 deviations, 4 OOS, 23 changes and 58 complaints. Here is what happens, screen by screen, in the working solution.
Priya authors 17 of the 24 reviews. The review year shows all 24 product reviews of the 2026 cycle on a dial — 8 in progress, 16 approved — with the clocks still running. Talvexa oral solution is 9 days late, waiting on the QP. Corventa 20 mg has 22 days left, 9 of 14 chapters ready, 3 findings to decide and 1 count to reconcile.
“3 findings · owner Priya Raman · 3 findings to decide · 1 count to reconcile”
On Oct 1 at 06:02 the collectors pulled the closed period's 186 batches from ERP, MES, LIMS and the QMS. By 06:24 the analyst had drawn 24 small multiples — release testing, in-process controls, process and packaging, events and stability — each with its limit, centre line, capability and a trend call. Four need a decision, two are improving, eighteen are stable.
“186 batches loaded from ERP, MES, LIMS and QMS · pulled Oct 1, 06:02 · trend calls by the Trend and capability analyst, Oct 1, 06:24”
Batch mean fell from 94.1% to 89.6%: a Welch t-test on 91 against 95 batches gives Δ −4.5 points, t = −12.6, p < 0.001, and Nelson rule 2 finds 23 batches in a row below the earlier centre line. Cpk is 1.08 against the 1.33 target, down from 2.61. The change point sits on CC-26-0412, the second cellulose supplier — and one click on CV26-1017 opens its batch record: the first batch with Halvern cellulose, dissolution 91.8%.
“CV25-1188 · OOS-25-114 — rejected, left out of the SD”
Priya switches the year to Corventa 20 mg made at Bracken Pharma Services: 72 batches, original cellulose only. Dissolution is stable — before against after Feb 9, Δ −0.2 points, p = 0.71, Cpk 1.41. Six in-process panels wait for data, because Bracken has not yet sent 2 of 6 exports; the call is made the day the data arrives.
“Bracken buys microcrystalline cellulose only from the original supplier (Avenor Fibre) — the contrast that points F1 at the new source.”
Forming depth on blister line 2 dropped from 4.62 to 4.41 mm after the sealing temperature was lowered from 195 to 185 °C under CC-25-0388 on Mar 30: Δ −0.21 mm, p < 0.001, Cpk 1.00 against 2.11 before. Still inside 4.20–5.00 mm — but the same batches carry five broken-tablet complaints from April to June, a cluster against a rule of 3 in 90 days.
Common factor: “Blister line 2, after CC-25-0388 (5 of 5)”.
On the 2025 annual stability batch CV25-1045 at 25 °C / 60% RH, impurity B reached 0.31% at 18 months, above the 0.29% upper bound of the 95% prediction interval. Sam Patel's OOT-26-019 found no analytical cause. The regression projects 0.42% at 36 months, upper bound 0.47%, inside the 0.5% limit.
Recommended: keep on watch, add a 21-month pull, review again at 24 months. No change to shelf-life.
“Open F1 in chapter 4” takes her into the draft review: the critical quality attribute table with each Cpk, and the finding with its four citations. The analyst recommends a CAPA — add particle size (d50) to the incoming test for cellulose, test dissolution on every batch for 6 months, and ask MSAT to assess a supplier-specific granulation water amount. She raises it as recommended.
Her choices: “Raise CAPA as recommended” · “Watch only” · “No action — give a reason”.
The complaint system counts 58 complaints for the period; the QMS product link counts 57. The checker found why: CMP-26-0311, a discolouration complaint from May 2026, is linked to Corventa 40 mg in the QMS, but the batch on the complaint photo, CV26-1058, is a 20 mg batch. Priya relinks it — 58 in both systems.
A count that differs between two systems blocks sign-off until a person resolves it.
With all three findings decided — F2 a CAPA for a forming-depth check at line start-up, F3 on watch — and the counts reconciled, the signature opens. The electronic signature records who signs, the version and its meaning, and sends the review on to Dr. Omar Haddad, then Dana Okafor.
“Author — I prepared this review; its content and conclusions are accurate.”
The QP signs that the review supports continued certification of Corventa 20 mg batches; the Head of Quality approves the review, its conclusions and its actions. PQR-2026-014 is locked as version 1.0 and filed on Oct 7, 22 days before its due date, with the review history drawn from the audit trail.
“Approved — PQR-2026-014 is locked and filed, 22 days before its due date”
The same review, seen by the five people who carry it — what their review season looked like, and what it looks like now.
Seven agents collect the year, call a trend on every attribute and write the review; the product quality lead decides and signs.
Pulls production orders, usage decisions, yields and batch-record exceptions for the review period from ERP and MES — one record per batch with status, yield and exceptions.
Collects deviations, OOS/OOT investigations, CAPA, change controls, complaints, returns and recalls linked to the product, keeping the original classification and the links between events.
Pulls release results for every critical quality attribute and every stability pull from LIMS, with condition and time point.
Lists the variations filed, approved or refused in the period, open commitments with their due dates, and the technical and quality agreements for each site.
Makes a trend call on every attribute of the year — 24 per product — with the test, its numbers and the cited batches: Cpk, run rules, stability regression to shelf-life and complaint clusters. Raises each adverse trend as a finding with a recommended action.
Counts every number the review uses in two systems — ERP against QMS disposition, complaint system against the QMS link, MES against LIMS lots — and opens a reconciliation item with the records that explain any difference.
Fills the 14-chapter template with plain statements, charts and a citation behind every number, and drafts the conclusions on whether the process remains in a state of control.
Decides each finding and signs. The agents propose; a named person decides.
Ask about any product year, trend call or number — every answer cites the record it came from.
Which attributes moved this year on Corventa 20 mg?
Of 24 attributes trended across 186 batches, four need a decision: dissolution at 30 min (shift — Welch t, p < 0.001, Cpk 1.08), forming depth on line 2 (shift — Δ −0.21 mm, p < 0.001), complaints (cluster — 5 broken tablets in 90 days) and impurity B on stability (out of trend at 18 months). Two improved: granule moisture after the probe, and tablet weight after press 4 was requalified.
Why did dissolution drop after February?
The drop lines up with the second cellulose supplier. Batch-mean dissolution fell from 94.1% (91 batches, Aug–Jan, Cpk 2.61) to 89.6% (95 batches, Feb–Jul, Cpk 1.08). The first lower batches are the first made with Halvern cellulose under CC-26-0412, whose particle size is larger — d50 128–141 µm against 104–112 µm. Every batch still passed except CV25-1188, which failed for a different reason.
Which other products use the new cellulose supplier?
Two products use Halvern cellulose. Corventa 40 mg shows the same shift — dissolution Cpk 1.17 since February, raised as F-020-1 in its draft review — and Corventa 20 mg is finding F1, Cpk 1.08. The contract-made Corventa 20 mg at Bracken uses only the original supplier, and its dissolution Cpk is 1.41 — no shift. F-020-1 is now linked to F1, so one CAPA can cover both strengths.
Add a rule: flag complaint clusters of 3 or more in 60 days
Added for your approval as a trending rule, applied to every product from the next run. With it, the broken-tablet cluster on Corventa 20 mg would have been flagged on May 21, six weeks before the period closed.
13 screens from the working solution, on its sample data. Pick one to see it large.
How reviews are built, trended and signed are settings. Changes take effect on the next agent run and are recorded in the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Template | Combined US annual review + EU quality review (14 chapters) | Combined (14 chapters) · EU product quality review only (13 chapters) · US annual product review only (9 chapters) |
| Review period | Rolling 12 months per product, ending on its anniversary date | Rolling 12 months per product · Calendar year for all products |
| Capability target (Cpk) | 1.33 — below this a critical quality attribute is raised as a finding (SOP-QA-118 §7.3) | 1.00 · 1.33 · 1.67 |
| Run rules | 8 in a row on one side of the mean · 6 in a row rising or falling | On or off, each; describe a new rule in plain words |
| Stability out of trend | A pull outside the 95% prediction interval of earlier pulls | On or off |
| Count reconciliation | Every number counted in two systems; any difference blocks sign-off | On or off |
| Draft built | Next morning at 06:00 after the period closes | Next morning at 06:00 · Within 3 days · On request only |
| Due date | 90 days after period end (SOP-QA-110 §5.2) | 90 days · 60 days after period end |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample product data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (Sparta Systems / Honeywell product quality review white paper — 1–3 months per manual review cycle). Regulations named as context: 21 CFR 211.180(e), EU GMP Chapter 1 (product quality review), ICH Q10. People, companies, products, sites and record numbers named on this page are fictional — characters and sample data in the working solution.
Software that builds each product's yearly review — the US annual product review under 21 CFR 211.180(e) and the EU product quality review — from the records of the year. Annual Product Quality Review has seven agents collect the year, make a trend call on every attribute with its test, reconcile every count and draft the 14-chapter report, for the product quality lead to decide and sign.
Yes. The default template combines the US annual review and the EU quality review in 14 chapters, each mapped to its reference — 211.180(e) and EU GMP 1.10 items (i) to (xii). An EU-only (13 chapters) or US-only (9 chapters) template is a setting.
The trend and capability analyst draws every attribute across the year's batches and makes a call — stable, shift, out of trend, cluster or improving — with the test that made it: a Welch t-test for a shift, Mann–Kendall for a trend, an F-test for variance, Nelson run rules, a c-chart for event rates and a stability regression with a 95% prediction interval. Each call cites the batches behind it.
Every number in the review links to the records it was counted from, and the reconciliation checker counts it in two systems where it can. On the sample product the complaint system said 58 and the QMS said 57; the difference was one complaint linked to the wrong strength, and sign-off waited until the product quality lead resolved it.
No. Findings are proposals with a recommended action. The product quality lead decides each one — raise a CAPA, watch, or no action with a reason — and can agree with or challenge any trend call, with the reason kept.
Yes. Conclusions stay draft until the author signs. The review is signed in order by the author, the Qualified Person and the Head of Quality, each with the meaning of the signature, and a signed review is locked.
Yes. The Cpk target, run rules, stability out-of-trend rule, count reconciliation, template, review period, when the draft is built, the due date and the signers are settings. New rules can be described in plain words, and every change is on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your APR/PQR template, trending SOP and read-only access to your source systems — and it goes live once every quality gate has passed. We will walk you through it on a past year of one of your own products first.
Any batch-record question answered in seconds, cited to the page and step.
1-min demoValidation Document AutomationProtocols drafted in minutes, every value traced to its source.
1-min demoManufacturing Batch AnalyticsBatch analysis that took three months now takes hours, with the query behind every answer.
1-min demoWe'll run Annual Product Quality Review on a past year of one of your own products.