Objectives: The workshop will provide an introduction to regression methods for summary ROC analysis, using both fixed effect and random effects (hierarchical) models. At the end of the workshop, participants will be aware of the alternative models, their application and interpretation.

Description: Emphasis will be on the meta-analysis of studies reporting estimates of test sensitivity and specificity. The focus will be on summary ROC methods that account for differences in positivity threshold between studies. The use of these models for investigating sources of heterogeneity through the inclusion of study level covariates, as an alternative to subgroup analysis, will be described and discussed. The application of this approach to comparing the performance of tests will also be discussed. The potential advantages of using hierarchical models will be outlined. Examples, drawn from imaging and laboratory test evaluations, will be used to illustrate and compare alternative (fixed and random effects) models and their interpretation. Software for fitting the fixed effect SROC regression will become available in Revman. Sample programs will be provided for fitting hierarchical models in SAS.

Target Audience: People with a working knowledge of basic statistical methods and some experience with meta-analysis.

Style: Training Workshop