Rare events data is commonplace in rodent developmental toxicology studies, where fetal defect rates are often very low in the control or low dose groups. In these studies, the litter serves as a random effect nested within the dose groups, and the occurrence of defects may be positively correlated within the litter. Binary data with random effects are commonly analyzed using generalized linear mixed models (GLMM's); however, it is known that with rare events mixed models may not converge or may provide unreliable results. A sample size adjustment to the Cochran Armitage test developed by Rao and Scott (1994) for clustered binary data provides an alternative to GLMM's, requiring few assumptions regarding the distribution of the response. In this presentation, we provide an example of how contradictory results may arise from the GLMM method when applied to correlated rare events data, and discuss how small changes in the implementation of the Rao-Scott adjustment factor can affect the Type I error rate in the presence of rare events data. The effects of these changes are demonstrated through simulation.