Many interventions in public health act in settings where individuals are connected to one another and the intervention assigned to randomly selected individuals may spill over to their network members. The direct effect measures the intervention effects among those directly treated, while the spillover effect measures the effect among those in the network of those directly treated, also denoted interference. Here, we develop methods for study design in this setting aimed at estimating the direct and spillover effects. We investigate improved study designs with the goal of leveraging the most influential participants to increase the power of the study to estimate the effects. In particular, we develop an ego-network-based randomized design in which index participants are sampled from the population and randomly assigned to treatment while data is also collected on their untreated network members. We connect the potential outcomes framework to two clustered regression modeling approaches. We develop sample size formulas for detecting the direct effect and the spillover effect, accounting for how these vary as a function of index participant covariates, with pre-specified power.