Health & Place 13 (2007) 238–248 Efficacy calculation in randomized trials: Global or local measures? Michael Emch a,Ã , Mohammad Ali b , Camilo Acosta b , Mohammad Yunus c , David A. Sack c , John D. Clemens b a Department of Geography, Saunders Hall, University of North Carolina, Chapel Hill, NC 27599-3220 USA b International Vaccine Institute, Seoul, Korea c ICDDR, B: Centre for Health and Population Research, Dhaka, Bangladesh Received 13 June 2005; received in revised form 19 December 2005; accepted 6 January 2006 Abstract This study tests whether the effect of a vaccine trial varies in space and why. Analytical z-score maps identify unusually high- and low-efficacy values in a trial area. Relationships between neighborhood efficacy and ecological variables are measured to explain why efficacy varies in space. Efficacy was found to vary regardless of neighborhood size and the variation is related to several ecological determinants. Local efficacy measures can help public health practitioners make better decisions about when and where to vaccinate populations. The concepts offered in this study are pertinent for any health intervention trial, not just vaccines. r 2006 Elsevier Ltd. All rights reserved. Keywords: Intervention trials; Vaccines; Neighborhood determinants; Cholera; Herd immunity; Spatial analysis Introduction Clemens et al. (1996) questioned the utility of conventional vaccine trial methods. In particular, they suggested that public health practitioners cannot use traditional protective efficacy measures to make decisions about whether or not to vaccinate diverse populations. Many trials, including those measuring the efficacy of the Ty21a typhoid fever vaccine, have produced conflicting results in differ- ent settings. The Ty21a vaccine, for example, had an efficacy of 96 percent in Egypt (Wahdan et al., 1982), 77 percent in Chile (Levine et al., 1990), and 53 percent in Indonesia (Simanjuntak et al., 1991). There are various factors that can result in different efficacies including different vaccine formulations, potencies, study designs, case definitions, strains of the wild-type agent, and the ecological (i.e., socio- environmental) circumstances of the trial area. Conventional vaccine trial methods have an under- lying assumption that the effect of the vaccine is the same throughout the trial area. This paper is a case study that tests whether this assumption is true for one vaccine trial. If the vaccine efficacy is spatially heterogeneous in a trial area then public health decisions based on global vaccine efficacy may not be realistic. This study uses three data sets, including a cholera vaccine trial database, a longitudinal ARTICLE IN PRESS www.elsevier.com/locate/healthplace 1353-8292/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.healthplace.2006.01.005 Ã Corresponding author. Tel.: +1 919 962 8901. E-mail address: emch@email.unc.edu (M. Emch).