A Bayesian Approach to Benefit-Risk Assessment for Ophthalmic Devices

• Activity Number: 676 - Analysis and Reporting: Benefit-Risk and Robust Models
• Type: Contributed
• Date/Time: Thursday, August 2, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Chul Ahn*
• Companies: FDA-CDRH
• Keywords: Benefit-Risk Measure; Visual Acuity; Power Prior; Dirichlet Distribution; Multinomial Distribution; Posterior Mean

Abstract

Developing quantitative methods for benefit-risk evaluation is indispensable to assess medical products. In this paper, we present several measures that can simultaneously evaluate benefit and risk in medical devices. As an example, we consider ophthalmic devices with a main objective of demonstrating improvement in uncorrected near visual acuity where losing uncorrected distance visual acuity may be considered as risk. The visual acuities from subjects can fall into four mutually exclusive quadrants because of four possible combinations of (1) benefit with no-risk (2) benefit with risk (3) no-benefit with no-risk and (4) no-benefit with risk. We assume that the number of individuals in each of four quadrants follow a multinomial distribution with multinomial cell probabilities following Dirichlet distribution. We use a power prior through the likelihood function to discount the information from previous visits, and derive the posterior distributions of the cell-probabilities at multiple visits. We then derive the estimates of the posterior means and credible intervals for the benefit-risk measures we propose.