HPTN054 is a cluster randomized trial designed to compare two approaches to providing single dose nevirapine to HIV-seropositive mothers and their infants to prevent mother-to-child transmission of HIV in resource limited settings. A number of challenging issues arose during the design of this trial. Most importantly, the need to achieve high participation rates among pregnant, HIV-seropositive women in selected prenatal care clinics led us to develop a method of collecting anonymous and unlinked information on a key surrogate endpoint instead of pursuing linked and identified information on a clinical endpoint. In addition, since group counseling is the standard model for prenatal care in sub-Saharan Africa, the prenatal care clinic serves as the unit of randomization. However, constraints on the number of suitable clinics and other logistical difficulties necessitated a unique type of hybrid parallel/stepped wedge cluster randomized design in which some clinics cross over between the two treatment modalities and some do not. We describe the design for the HPTN054 trial with an emphasis on the logistic and statistical features that allowed us to address these issues. We also provide some general statistical results that are useful for computing power in parallel, crossover, stepped wedge or mixed designs of cluster randomized trials.


Clinical Trials | Design of Experiments and Sample Surveys