Analyzing Hypoglycemia Events eith Negative Binomial Regression Model

• Activity Number: 522
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Junxiang Luo and Yongming Qu*+
• Companies: Eli Lilly and Company and Eli Lilly and Company
• Keywords: hypoglycemia ; negative binomial ; overdispersion

Abstract

Negative binomial (NB) regression is a standard model in analyzing hypoglycemia events in diabetes clinical trials. It could potentially increase estimation efficiency of NB regression when adjusting for baseline covariates. However, adjusting for covariates raises concerns on model misspecification, in which NB regression is not robust due to its strong assumption restriction. McCulagh and Nelder (Generalized Linear Models. Chapman & Hall 1989) suggested to correct standard error of maximum likelihood estimates through introducing overdispersion estimated by the Deviance or Pearson Chi-square. We proposed to use sandwich estimation to calculate the covariance matrix of the parameter estimates together with Pearson overdispersion correction (NBSP). This research compared several commonly used NB model options with our proposed NBSP. Simulations and real data analyses showed NBSP is most robust to model misspecification, and the estimation efficiency will be improved by adjusting for baseline hypoglycemia.