Speakers:

• Session chair: Shelly Sundberg
  Bill & Melinda Gates Foundation
  @gatesfoundation

• Ed Frongillo
  University of South Carolina
  @UofSC @DFC_Program

• Jef Leroy
  International Food Policy Research Institute (IFPRI)
  @IFPRI

Summary:

RCTs are increasingly used to evaluate the effectiveness of agriculture, nutrition, and health (ANH) interventions in low- and middle-income countries. ANH trials are fundamentally different from trials to evaluate the efficacy of medical interventions. First, scientist in ANH trials, however, often do not know the exact paths by which the intervention will have an impact. Second, the effect of ANH interventions depends on both biological and behavioral mechanisms. Third, ANH interventions typically include more than one intervention component and are thus expected, by design, to affect multiple outcomes. As a consequence, in ANH trials, we are typically interested in 1) estimating effects on a range of outcomes directly linked to the objectives of the intervention (e.g., agricultural productivity, nutritional status, women’s empowerment) and 2) understanding the mechanisms underlying these effects by documenting effects on outcomes along the impact path. Limiting the number of outcomes measured (as is done in drug trials) thus runs the risk of missing important results. Additionally, in ANH trials there is no simple definition of success based on either having at least one positive effect (among a set of prespecified outcomes) or having a simultaneous positive effect on all (primary) outcomes.

The special characteristics of ANH trials lead to questions and confusion among researchers, funders, and users of ANH evidence about how to correctly conduct ANH trials and interpret their findings. Topics of confusion include, among others, the choice and number of primary and secondary outcomes, the need to control for multiplicity, and the correct use of intent-to-treat and per-protocol analysis.

To ensure that ANH trials are conducted to the highest scientific standards while maximizing the evidence that can be extracted from each trial, new guidelines are being developed. Key aspects of the draft guidelines will be presented in this session followed by Q&A and discussion.