174 – Biomarkers and Endpoint Validation
Comparison of Methods to Generate Reference Limits
Bipasa Biswas
CDRH, FDA
Nairita Ghosal
University of Illinois at Chicago
Clinical laboratory tests often require a reference interval for quantitative tests and the construction of such intervals are common for laboratory tests. Reference Intervals are used to determine unusual or extreme measurements in laboratory medicine. The reference interval is the central interval bounded by the reference limits. In vivo diagnostic devices in ophthalmology and neurology often compare an individual patient's medical device output/test result against a database of output/test results from subjects deemed to be in good health, for clinical management of the individual patient. The device output/test results from the subjects in good health constitute a reference database (also commonly known as normative database). Reference database for In Vivo diagnostic devices is composed of measurements of multiple anatomical or physiological features from healthy individuals. 1st, 2nd, 2.5th, 5th, 95th, 97.5th, 98th or 99th percentiles are usually reported from the reference database. This presentation compares three common methods - Nonparametric, Harrell-Davis and the Robust method to generate reference limits.