September 22, 1999 - FDA Guidance Document on Labeling for Laboratory Tests

The ASM believes this guidance document has the potential to benefit all stakeholders (industry, FDA, end user, laboratory, and health care provider) and provide a rational and uniform labeling standard. ASM wishes to offer the following comments that, in our opinion, will improve the overall wording of the document as well as advance the FDA's objectives.

In the second paragraph of the Introduction, the term "diagnostic state" is used. This term appears to be unique to this document and as such, introduces a new term not frequently used in the laboratory. The term "diagnosis" should be used.

In the Introduction (Laboratory Equivalence), you state that the objective of the guidance is to demonstrate the ability of the new test to agree with the predicate. This "degree" of agreement should be stipulated mathematically by percentage, sensitivity, and specificity. This is further mentioned in the second paragraph under Proposed Labeling where disease prevalence is discussed. The FDA needs to keep in mind that the prevalence of a disease in the population is what governs predictive value. This information should be clearly stated in the labeling.

Further on in the Labeling section, there is a discussion of the candidate device being compared to a predicate. If there is no predicate device some other standard should be considered to address this issue. If the true diagnosis of the patient cannot be used, we recommend that existing standards promulgated by the NCCLS be accepted and cited. Furthermore, if the device is compared to a predicate but not the diagnosis, then the sensitivity and specificity needs to be stated in relation to the predicate device or some other standard must be cited. This latter comparison is often the only criteria on which a laboratory may choose to evaluate and test a patient in their facility.

The ASM hopes that these comments will improve the utility of device labeling. We recognize that it is not the FDA's goal to completely control the wording of device labels. Nevertheless, we do support some degree of uniformity, which in-turn will make it easier for the end user to evaluate and operate a specific device.