Impact of using less objective symptoms to define tolerated dose during food challenges: A data-driven approach