If specific tests such as sterility tests, pyrogen tests, or special physicochemical tests need to be carried out by another unit or by a specialized external laboratory, the analyst should prepare the request and arrange for the transfer of the required number of units (bottles, vials, tablets) from the sample. Each of these units should bear the correct registration number.

Testing should be started as soon as possible after the preliminary procedures have been completed. If this is not feasible, the reasons should be noted in the worksheet and the sample placed in a special locked storage cabinet.

Detailed guidance on test methods is contained in the general notices and monographs of official pharmacopoeias. The following principles therefore apply only when no pharmacopoeial requirements are available or when ambiguous results are obtained.

Provided the result is unequivocally positive and the analyst is well acquainted with the technique, replicate chemical and physicochemical tests are not, in general, required for identity tests if based upon colour reactions, precipitation tests, infrared spectra, ultraviolet identification, or thin-layer chromatography, nor are they required for purity tests based on the matching of colour or opacity against standards or on thin-layer chromatography. In some laboratories, however, purity tests are routinely run in duplicate as a check against accidental contamination. Assays to assess strength or level of impurity should always be replicated, however, whether they are based on titrimetry, gravimetry, colorimetry, ultraviolet measurements, gas-liquid chromatography, or high performance liquid chromatography. Replicate measurements should also be made of physical properties such as pH values, optical rotations, refractive indices, and melting temperatures. Whenever replicate measurements are made, the results should be recorded as the arithmetic mean of the estimates.

In other cases, the required number of replicate measurements is defined in the description of the method. This applies to physicochemical tests involving gas-liquid chromatography or high performance liquid chromatography and to biological assays whose results require statistical evaluation.

Whenever ambiguous results are obtained, or when the discrepancies between replicate measurements fall outside acceptable limits, at least two further replicate tests should be run, preferably by a different analyst. Any important discrepancies must be investigated. Aberrant results can be rejected only when they are clearly due to error. Otherwise, the mean values obtained by each analyst should be quoted separately to provide clear confirmation that the sample failed the test.

Errors arise not only because of human failings but also as a result of unsuitable or deteriorated reagents and chemical reference substances, inadequate instrumentation, inappropriate methods (particularly methods that are difficult to reproduce), and variations in the laboratory environment. Comparative estimations on standard samples can frequently help to detect such errors, particularly in cases in which the analyst lacks experience in the method he has used.

All values obtained in each test, including blank results, should immediately be entered on the worksheet, and all graphical data, whether obtained from recording instruments or hand-plotted, should be attached.