Physicochemical Monitoring of Pharmaceutical Water

In addition to the monitoring of physical and microbiological parameters, a monitoring program should also include a regular assessment of measured val­ues and trends. This must be carried out by appropriately trained and qualified personnel.

Scope of physicochemical monitoring

The measuring points must be correctly selected regarding the sensor location in order to prevent malfunctions or incorrect measurements. Care must also be taken to ensure that the measurement ports are free of dead space to avoid possible microbial contamination.

The minimum scope of the measuring points that a physicochemical moni­toring program for a pharmaceutical water system should contain is as follows:

• Temperature after the processing plant
• Conductivity after the processing plant
• Temperature in the return line of the loop
• Conductivity in the return line of the loop
• Pressure and/or flow rate in the return line of the loop (ensuring a minimum flow rate)
• TOC measurement in the return line of the loop

The recording of other parameters, such as the fill level in the storage tank or ozone concentrations, are important for the safe and economical operation of a water treatment plant, but are not mandatory for pharmaceutical operations from a regulatory perspective. Figure 5.G-7 provides an overview of test frequen­cies for the physicochemical parameters that have proven themselves in prac­tice and are recommended by the authors.

Modern technology enables the seamless recording of parameter data. The frequency at which this data should be digitally recorded must be determined in advance by means of a risk analysis. The aim is to be able to evaluate and statis­tically analyse the collected data in a practicable manner for the purpose of data trending.

This consideration must also take into account whether the data should be stored on a server or on an electronic recorder. For new systems, redundant stor­age of data on qualified servers has been established as best practice. The advantage of an electronic recorder is the lower qualification effort, but the extraction and processing of the data for trending is sometimes very time-con­suming. For both systems, however, the requirements of 21CFR Part 11 and Annex 11 of the EU GMP Guide must be met.

Instrument failure

In principle, the impact of sensor and measuring device malfunctions or failures should be evaluated in a risk analysis. This allows actions to be defined in advance as to how to proceed in such a situation. This risk analysis should also include an assessment of which malfunction or failure constitutes a deviation that must be dealt with in accordance with deviation management.

This often leads to redundant measurement concepts. For example, the con­ductivity values and TOC are recorded inline continuously in accordance with the monograph and could theoretically be used to release the water for produc­tion. However, additional samples are often taken periodically. These are then analysed offline in the laboratory and used for release purposes. Theo­retically, the inline data from the loop would be sufficient for release, but then it must be clarified in advance what should happen if the instrument fails or a cali­bration deviation occurs. For example, a redundant conductivity measurement (forward and return flow) or a theoretical backup process with daily offline meas­urements in the event of faults can help here. This backup process then only comes into effect if an inline instrument malfunctions. The measurement down­times during this time must be documented in the system logbook.

The measures resulting from the risk analysis must be described in an SOP regarding each product-relevant quality measurements, e.g. conductivity and TOC.

This water system SOP should at least address the following situations and describe the required procedures:

• System failure: the system must be automatically switched to a safe operating mode,
• Replacement of a sensor or instrument,
• Repair of a sensor or instrument,
• Maintenance and servicing of a sensor or instrument,
• Use of periodic offline measurements,
• Evaluation of data in the event of a fault or failure.

Warning and action limits

The pharmacopoeias specify binding limit values for the quality-relevant param­eters checked as part of the monitoring process. To prevent the specification limits from being reached or even exceeded, appropriate warning limits (also referred to as alert limits) and action limits are set internally.

The internal action limit may be set to be equal to the official specification lim­its. However, it is advisable to set the action limit to be tighter than the specifica­tion limits given in the pharmacopoeia to avoid a system shutdown and to be able to react in good time. If the pharmacopeial specification limit is exceeded or not reached, the system will be blocked. If the action limit does not correspond to this limit value, going outside these limits causes a deviation that must be processed accordingly, but the system is not necessarily blocked.

Warning limits are usually defined in addition to the action limits. These warn­ing limits should be defined in a way that measures can be initiated before the action limit is reached without this resulting in a deviation (however, a deviation must be triggered if the upper or lower alert limit is repeatedly overstepped).

Warning limits are therefore an early warning system that serves to recognize a drift of the corresponding parameter out of the usual range in good time. The warning limit should be close to the working value. In general, the alert limits are subject to regular monitoring and can be adjusted if necessary.

There are no binding specifications for the definition of alert and action lim­its. Trend data is used for existing systems. For new systems, this data must be generated first. Depending on the statistical distribution of the data (ideal case: normal distribution), statistical methods can be applied with more or less suc­cess. If the data is not normally distributed, the percentile method with 95/99% WL/AL is used, for example.

In another common approach, which does not require any special statistical knowledge, the action limit is set at approx. 50–80% of the official limit value (example TOC in Figure 1: specification limit 500­ppb ➜ action limit 250– 400­ppb). This provides sufficient safety margin before an OOS event occurs if the limit is exceeded. In the author's experience, authorities usually require both alert and action limits.

The warning limit is then determined from the difference between the action limit and the working value. It is recommended to add 20 - 50% of the difference to the working value as a buffer range (see conductivity as an example in Figure 1: Difference between action limit and working value = 0.28 μs/cm; buffer = 0.06 - 0.14 μs/cm; warning limit = 0.66 - 0.74 μs/cm).

Figure 1                Definition of warning and action limits

The technical level of instrumentation technology is now so high that it is possible to set very tight limits to ensure the quality of the product. Even the smallest fluctuations often indicate that something is wrong. If the warning limit is exceeded, further actions should be taken, such as checking the last measurements or evaluating trends. If no findings are obtained, the incident is only documented.

Monitoring results must be evaluated at least once a year. The manner and methods used for this evaluation must be defined in an SOP. The SOP must also specify which measures are to be initiated if warning and action limits are exceeded and what consequences are to be drawn if warning and action limits are repeatedly exceeded. This trend evaluation must also include an assessment of whether the warning and action limits still match the parameter measure­ment results and are appropriate. Adjustment of such alert and action limits should be carried out via a change request.