Aubert M
CNEVA, Laboratoires d'Etudes sur la Rage et la Pathologie des Animaux Sauvages, Malzéville, France.
Dev Biol Stand. 1992;79:113-20.
The variability of a control test is sometimes difficult to assess. In fact this variability should be considered as a common basis of any biological test. The example of the NIH or PER test is given for veterinary rabies vaccines. The same acceptability threshold is accepted by everybody, both producers and control laboratory. But biological variability should induce the producers to determine a "working acceptability threshold" higher than the conventional one. This safety margin should be determined by the producer to counterbalance the variability of the test both in the producer and control laboratories so that most batches submitted to the official control are accepted. On the other hand, control laboratories should fulfil two "fair" requirements: reveal their flow charts and their margin of error to allow the producer to make the needed adjustments; reduce as far as possible the uncertainty of control tests by using the most accurate statistical interpretation to allow the producers to establish a "working acceptability threshold" not too much higher than the official one.
对照试验的变异性有时难以评估。事实上，这种变异性应被视为任何生物学试验的共同基础。文中给出了用于兽用狂犬病疫苗的NIH或PER试验的例子。生产者和对照实验室都接受相同的可接受性阈值。但生物学变异性应促使生产者确定一个高于传统阈值的“工作可接受性阈值”。生产者应确定此安全边际，以平衡生产者实验室和对照实验室中试验的变异性，从而使提交官方检验的大多数批次能够被接受。另一方面，对照实验室应满足两个“公平”要求：公布其流程图和误差范围，以便生产者进行必要调整；通过使用最准确的统计解释尽可能降低对照试验的不确定性，以便生产者确定一个不会比官方阈值高太多的“工作可接受性阈值”。
