Gravar-mail: Common pitfalls in statistical analysis: Intention-to-treat versus per-protocol analysis