Best (but oft-forgotten) practices: intention-to-treat, treatment adherence, and missing participant outcome data in the nutrition literature.

Among clinical trials of adequate size, randomization balances both known and unknown prognostic factors between trial arms, thus allowing an unbiased comparison of intervention and control. To preserve this benefit, all randomly assigned participants should be followed to study termination and analyzed in the arm to which they were allocated. There are 2 potential limitations in study implementation: 1) patients are nonadherent and continue with follow-up visits, or 2) patients are lost to follow-up and their outcome data are missing. Herein, we address these issues with an emphasis on binary outcomes, and discuss how authors of randomized trials should address issues of both noncompliance and missing data.