Decisions that could be made following an analytical instrument qualification procedure might include:
− Is the instrument performing to the manufacturer’s specification? (The fact that an instrument is performing to specification does not necessarily mean that it is suitable for the test method to be performed!)
− Does the instrument’s performance comply with the requirements of the regulatory environment within which it is being used, e.g a Pharmacopoeia?
− Does the instrument’s performance meet the requirements of the specific analytical method to be used?
While the pharmacopoeias and other standards may give generic recommendations for instrument performance, for a specific test method the required performance characteristics may be different and are usually given in the specific test monograph for the analyte. Furthermore they may change if methods are revised or updated, and for this reason “pass/fail” criteria are not given in Starna documentation. For the same reasons claims by instrument manufacturers that a spectrophotometer is “pharmacopoeia compliant” should be treated with caution.
In a regulated environment, the onus of proof is on the user to justify and prove, by the qualification of the system, that the instrument is ‘fit for purpose’ and capable of providing the required accuracy and precision of data.
Essentially, the decision is made by comparing the measured values with the certified values of the Reference(s), to determined the error (bias) for the parameter being tested. If that bias is less than the tolerance allowed for that parameter, the instrument can be judged to be compliant.
It must be remembered that the certified value is not absolute: it is itself the result of a calibration exercise carried out at the Starna Calibration Laboratory. All CRM calibrations are performed on a reference spectrophotometer that has itself been qualified against Primary NIST Standard Reference Materials. The certified value will therefore be subject to any variability of the calibration instrument. This means that there is a small uncertainty surrounding the certified value, the “uncertainty budget” of the CRM. This is stated on the Certificate supplied with each Reference Material. Also included within this budget are other components such as cell variability, drift, etc. The measurement uncertainty statement gives a range within which the measurand’s true value is considered to lie. This range is centred on the measurement value, and while the measurand’s true value is likely to be close to the measurement value, it may lie anywhere within the range. Indeed, there is a small probability that the measurand’s true value may lie outside the range.
To be meaningful, the uncertainty statement must have an associated confidence level. It is necessary to state the probability that the true value lies within the range given. A 95% confidence level means that there is a 95% probability that the true value lies within the stated range, or in other words there are not more than five chances in one hundred that the true value lies outside the range.
The reasons for choosing a 95% confidence level may be summarised as follows:
It is established practice through much of Europe, North America and Asia.
The ISO Guide to Uncertainty of Measurement (GUM) assumes that the combined uncertainty has a distribution that is a close approximation to a normal distribution. It is a widely held view that for most measurement systems, the distribution of the combined uncertainty approximates to a normal distribution out to two standard deviations, but beyond that the approximation is less reliable. A range of two standard deviations approximates to a 95% confidence level .
An approximate 95% confidence interval can be simply obtained by multiplying the combined standard uncertainty by 2, usually stated as ‘k=2’ .
The instrument being tested will also have its own uncertainty budget, e.g. if absorbance is the parameter being qualified, it will be the accuracy of the absorbance scale, as specified by the instrument manufacturer.
If the instrument is performing to specification, the absolute difference between the measured value and the certified value will be less than the sum of the certified uncertainty of the reference and the accuracy of the instrument as specified by the manufacturer. Similarly, it should also be less than the error permitted by the quality regime prevailing in the laboratory or the requirement of the analytical method in use.