The first blog in this series introduced us to continuous environmental monitoring in the cleanroom, and how GMP standards have changed over the years. Now let’s dive into the configuration of alert and action levels and potential reactions, and their reporting requirements.
What Changed from the Old GMP Standard?
Alert and action limit requirements are carried over from the 2008 version of the EU Commission Annex 1 and clarified in the new 2020 draft. They do not mention what number determines periodic events, nor do they define what number constitutes an actionable threshold. That is instead documented with the Risk Assessment and Contamination Control Strategy (CCS). For reference, see sections 9.8, 9.22, and 9.23 of the document. From reading this content, you should see that monitoring procedures should define the approach to trending. Trends can include, but are not limited to:
- Increasing numbers of action or alert level breaches
- Consecutive breaches of alert levels
- Frequent or regular isolated breaches of action limits that may have a common cause
How Do I Set My Particle Alert and Action Limits with Trending?
Alert and action limits can be applied to reflect a particle trending condition referred to as an N:M, where the number of events (N) is within a population of events (M). For example, an N:M of 2:6 would refer to a rolling set of numbers where any 2 samples out of a total of any 6 samples would require an action, with the initial trigger setting the clock to review the next 6 samples for a second triggering event. If none are present, the particle counter is reset until the next activation event.
Using this application, the system alerts are converted to non-actionable events, where the trigger is only used for data review. To ensure the system does not trend adversely over longer periods, the non-actionable events allow for comparison to similar time periods (e.g., weeks, months, quarters) or events (e.g., filling, batch, lot). This ensures the limits prescribed in the system are reliable triggers of adverse conditions, and reflects a loss of control over the environment as it pertains to product quality.
The particle levels are set with each of the following:
- Alert – If any value goes above the ‘normal’ operating condition values when normalized to cubic meters (x35), an alert event is recorded. If alert levels are exceeded, operating procedures should prescribe assessment and follow up, which includes consideration of an investigation and/or corrective actions to avoid any further deterioration of the environment. This is essentially a non-actionable event.
Data is reviewed periodically for adverse trending and can be correlated against the room production events to ensure that where potential contamination may occur, the future revisions can reflect the required changes.
- Alarm – The frequency (N:M) is set such that a single location must be in ALERT for a determined frequency before triggering an ALARM event, beacon, siren, audit trail etc. When these events occur, operating procedures should prescribe a root cause investigation, an assessment of the potential impact to product and requirements for corrective and preventive actions. This is an out of control condition.
It follows then that a determination of the level to be used to catalyze needs should be determined.
The PDA Technical Report 13 recommends using the following:
- Cutoff Value – A 95% or 99% control limit based on data previously gathered from a single site or collection of similar sites.
- Normal Distribution – The mean and standard deviation of a set of data, where high counts exist. Where a predominance of zeros prevail, skewed data sets may occur and non-parametric rules apply.
- Non-parametric Tolerance – The statistical determination that a room meets compliance with at least 95% confidence.
What About Alert and Action Limit Reports?
System reports should be considered as part of the CCS. As we noted in the previous blog, a cleanroom monitored using a portable particle counter produces data that needs to be ‘absolute’ and shown in all reports. However, the transition to a continuous environmental monitoring system now continuously demonstrates adherence to control parameters (defined by alert and action levels), and only instances where there is a loss of control requires investigation. Overall trend reports and audit trails show when the system was out of control and any remedial actions taken as part of the SOP.
System reports should focus on:
- The report boundary, what sensors are included, what time frame the report covers and any additional batch information that uniquely identifies this period.
- The audit trail, alert and action threshold excursions, remedial actions, user log in, start end time data, system changes, etc.
- The trend report as a way to show data as a easy to read trend plot (graph) image, demonstrating no loss of data.
- Where any out of tolerance has occurred and an action level exceeded. This period can be covered by an exception report to demonstrate that prior to the event, all was running typically and post-event normal conditions were restored.
Where to Go From Here
Check out Particle Measuring Systems’ advisory services to get started on your Contamination Control Strategy!
Learn about Particle Measuring Systems’ complete environmental monitoring and data management solution.
A data review is required per sample location, or group of sample locations which exhibit similar properties. The practice of normalization is more to put the data into an SI unit mentioned in regulations than understanding process controls. For example, pressure can be used as either inches of water gauge (H2O), pascals (Pa) or millibar (mBar), but each has a different number value due to scaling. Data per minute gives a balance between measuring at a short enough interval for response rate, while providing enough sample volume (assuming a 28.3 LPM flowrate) and therefore associated potential counts and statistical confidence in the data.
We can also consider the extremes of each time period. Long sample periods (35.3 minutes) will give great statistical confidence in the data, but poor response to control conditions and very short sample periods (5 second intervals) will give great dynamic responses. Low confidence data is where the few counts yielded requires multiple records to gain statistical insights and leads back to low responses. As such, the raw data from the instrument can be used, given as counts per minute, or counts per cubic foot. However, industry has found it more accommodating to reflect the standard in counts per cubic meter.
Regardless of what units you use, the number of particles which trigger the event and the frequency with which an event is escalated to actionable is on your to-do list. Check out some of our products with built-in alert and action level configuration below, and download the full paper on this topic: