Optimizing the utility of secondary outcomes in randomized controlled trials