Environmental monitoring for GMP cleanrooms: the risk-based programme, excursions, trends and review

What the module does

• Monitoring plans per cleanroom suite, starting from the EU GMP Annex 1 (2022) action limits for viable monitoring (Table 6) and for particles at rest and in operation (Table 1), to be checked against the site's own; each point with its code, room, grade, methods (active air, settle plate, contact plate, glove print, swab, particles at 0.5 and 5.0 µm), frequency and whether it is a critical location.
• Alert and action levels by grade, method and operational state, each with its basis; a change needs a reason and applies to new results only, so each result keeps the limits it was judged against.
• A risk assessment of every sampling point on five factors (proximity to open product and critical surfaces, activity and personnel, air flow and pressure cascade, difficulty of cleaning and disinfection, history of recoveries) that suggests a frequency for its grade; Grade A is sampled each session whatever its score, and a point sampled less often than its risk suggests needs a written justification.
• The programme record: the contamination control strategy it serves, the rationale, the review period, incubation regimes per viable method, the site's objectionable organisms and a contamination control checklist; it is approved with signatures (the author signs, a reviewer checks, QA approves), locked, and revised under change control as a new version.
• A schedule of every point and method with its last sample, when it is next due and whether it is overdue; the due points go into a sampling round, labels carry plan, point and method in one code, and scanning a label ticks that point and method.
• Sampling rounds registered in the LIMS from the plan, one sample per point and one test per method, with the action level as the test's upper limit and the alert level as its alert limit; a round with a retired point, a method the point does not have or active air without its volume is refused whole.
• Incubation recorded step by step against the plan's regime, in an incubator registered as a LIMS instrument; an early end needs the reason, and the LIMS refuses the plate's count until the last step has ended.
• Particle counter and active air sampler exports imported with their columns, units and time zones reviewed and every excluded row acknowledged; each row becomes a LIMS sample with the file's SHA-256 as its provenance and the file attached as raw data, reviewed and approved in the LIMS like any result.
• Trends over 3, 6, 12 or 24 months: each point and method charted with its limit lines, adverse trends found by rule (consecutive and repeated excursions, excursions in one room, recurring and objectionable organisms, recovery rates above the USP <1116> rate or rising, shifts and runs), statistical process control with Nelson or Western Electric rules, and alert levels suggested from the site's own data by the PDA Technical Report 13 percentile method.

How it fits the platform

Samples and results live in the LIMS. The laboratory receives the plates, swabs and bottles under chain of custody and records, reviews and approves the results under the LIMS's rules; monitoring reads each result back and never takes a typed result. A sampling point is one code in monitoring, the LIMS and the PM Library. Water systems are sampled and tested by point of use in the LIMS's water testing, on the same plans. A sampling round or an instrument import can carry a session reference and a linked MES batch, which appear on the LIMS sample, the monthly log and the exports.

An action excursion raises a deviation investigation with the result written up, linked to the month both ways; the root cause analysis, action items and CAPA actions follow in Deviations and CAPAs, and the result shows the deviation's standing. Any series opens a signal review in Signal detection, and a chart can be saved to Charts and the Q Dashboard. A programme change decided in a periodic review is made on the plan under change control. Q Intelligence, Q Agent, the read-only REST API and the MCP server read the same trend analysis.

Where people decide

The system schedules the points, registers rounds from the plan, judges every result against the plan's limits and finds adverse trends; it evaluates no excursion and opens no investigation by itself. People design the programme and justify each frequency, and its approval is signed by the author, a reviewer and QA. The laboratory records incubation steps and results in the LIMS, where a second person reviews and QA approves. A person evaluates each excursion with its likely cause, product impact and immediate actions, and raises the deviation; the monthly review and the periodic review are signed by people who recorded none of the period's results. Q Intelligence can propose a sampling round or a deviation from an excursion, and each waits for a person's approval; it does not record results, incubation steps or the programme review.