Docket Management
Docket: 03D-0044 - Statistical Guidance on Reporting Results From Studies Evaluating Diagnostic Test.
Comment Number: EC -2

Accepted - Volume 1

Comment Record
Commentor Mr. Qing Chen Date/Time 2003-04-10 17:52:17
Organization Diagnostic Products Corporation
Category Company

Comments for FDA General
Questions
1. General Comments Diagnostic Products Corporation Comments on the Draft Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests; Draft Guidance for Industry and FDA Reviewers 1) (Top of page 6) We believe qualitative assays should be reviewed and presented as qualitative assays even if it is derived from an underlying quantitative assay. ROC is applicable to a qualitative assay. However, the range of results and histograms of results are not and, therefore, should not be required in the labeling presentation for a qualitative assay. It is obvious that this was the basis for FDA holding some of our qualitative assays to the standards of quantitative assays. (Not the least burdensome!) In these cases, FDA reviewers told us to add a caution to warn the users not to interpret the magnitude of the results (usually expressed as a ratio) while at the same time they also told us to show quantitative statistics, even though the assay was a qualitative assay. 2) (Page 11) The term agreement of new test with imperfect standard-positive is extremely cumbersome. What is the rational for introducing this term in the Guidance? We suggest replacing it with positive agreement, or relative sensitivity? Both were terms favored by FDA reviewers in the past. We do believe that these statistics are helpful to the users. 3) (Page 12, Table 5A) There is no need to have a table to explain agreement doesn't mean correct, which we believe is very clear in the context where the predicate device is an imperfect standard. 4) (Page 14, Table 7) Although a good presentation, truth (i.e., resolver) is usually not readily available. If it was, a manufacturer would have used it from the beginning. This part of the guidance doesn't provide practical help to the users. What does FDA recommend in the absence of a resolver? 5) (Pages 14-16) The document addresses repeat testing of discrepant results and recommends that only the original results be presented. We agree that presenting the revised numbers from the repeat testing along with the original numbers is not statistically sound. Is there any value for repeating the discrepant results from the FDA point of view? If there is, the document should provide guidance on how to use them. 6) It is appropriate to have a guidance document on this issue. However, the document has not provided guidance for the treatment, presentation, and statistical calculation of tests that include equivocal (i.e., indeterminate) results.




EC -2