Evaluating Hazard Rate of Serious Adverse Events in Long-Term Clinical Trials

• Activity Number: 376
• Type: Contributed
• Date/Time: Wednesday, August 6, 2003 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Rafia Bhore*+ and Mohammad F. Huque
• Companies: Food and Drug Administration and Food and Drug Administration
• Keywords: hazard rate ; time-dependent ; serious adverse events ; drug safety ; piecewise exponential ; kernel smoothing

Abstract

In clinical trials, generally hazard rates of drug-induced adverse events vary over time. The most common patterns of hazard rate function are decreasing hazard, where events occur in the early part of exposure; constant hazard, where events occur evenly through the entire period of exposure; and increasing hazard, where the frequency of events increases after some delay in exposure. Crude proportions for estimating incidence of adverse events may be appropriate for events that occur early in treatment or when patients are exposed short-term. However, crude rates often understate the importance of time dependency of events that occur after some delay of weeks, months or even years. Descriptive statistics such as crude rates also ignore important aspects of drug safety data,including varying hazard rates over time, informative censoring and differential follow-up of exposed patients. This presentation will review statistical methods for making inference regarding the characteristics of the hazard rate function. Methods addressed will include piecewise exponential modeling, kernel smoothing, and Cox proportional hazards modeling with time-dependent covariates.