Two-Phase Study Design: Weighting Considerations

• Activity Number: 687
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shoshana Daniel*+
• Companies: Covance
• Keywords: Epidemiologic Methods ; Two-phase Design ; Weighting Schemes ; Prevalence Estimation ; Confidence Intervals ; Sensitivity and Specificity

Abstract

In disease prevalence studies, cases can be identified by applying the diagnostic procedure to all subjects in the sample. However, when ascertainment of disease status is difficult and/or costly, a two-phase design can be employed as an alternative. In the first phase, all subjects are screened using a simple and inexpensive screening procedure (e.g., questionnaire, simple laboratory test). In the second phase, sub-samples of those who screened positive and negative undergo a confirmatory procedure to verify the true disease status. Statistical methods are available to extrapolate prevalence estimates for the subjects in the first phase based on the prevalence estimates obtained in the second phase (e.g., McNamee, 2002; Morvan et. al., 2008). We extend the literature prevalence and variance results to allow for weighting of non-diagnostic disease related criteria in the estimates. In so doing, we obtain conditional prevalence estimates for the population when the covariate distribution differs between the population at large and the screener.