Noninferiority of Diagnostic Tests

• Activity Number: 25
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Lakshmi Vishnuvajjala*+
• Companies: FDA/CDRH Division of Biostatistics
• Keywords: Non-inferiority ; Tests ; Diagnostic ; Performance

Abstract

When comparing two tests if sensitivity is higher and specificity is lower or sensitivity is lower and specificity is higher, the overall performance depends not only on the magnitude of difference, but also the consequences of false positives and false negatives, which in turn can depend on how and for what purpose the test is used, and any inherent risk the test may pose; for example, imaging can expose the patient to radiation. When we talk about a test (diagnostic device) being non-inferior (really saying not much inferior), it can involve either sensitivity or specificity, or both being lower, but not by much. Sometimes tests are compared using either positive and negative predictive values, or likelihood ratios of positive and negative tests, which give rise to similar issues. Sometimes a test which is non-inferior with a margin that is clinically acceptable may be preferred for various reasons: Less risk to patient; less invasive; less expensive; results available quickly and possibly other reasons.