Contextual equipoise: a novel concept to inform ethical implications for implementation research using randomised controlled trials in low- and middle-income countries

The call for universal health coverage requires the urgent implementation and scale-up of interventions that are known to be effective, in resource-poor settings. Achieving this objective requires high-quality implementation research (IR) that evaluates the complex phenomenon of the influence of context on the ability to effectively deliver evidence-based practice. Nevertheless, IR for global health is failing to apply a robust, theoretically driven approach, leading to ethical concerns associated with research that is not methodologically sound. Inappropriate methods are often used in IR to address and report on context. This may result in a lack in understanding of how to effectively adapt the intervention to the new setting and a lack of clarity in conceptualising whether there is sufficient evidence to generalise findings from previous IR to a new setting, or if a randomised controlled trial (RCT) is needed. Some of the ethical issues arising from this shortcoming include poor-quality research that may needlessly expose vulnerable participants to research that has not been adapted to suit local needs and priorities, and the inappropriate use of RCTs that denies participants in the control arm access to treatment that is effective within the local context. To address these concerns, we propose a complementary approach to clinical equipoise for IR, known as contextual equipoise. We discuss challenges in the evaluation of context and also with assessing the certainty of evidence to justify an RCT. Finally, we describe methods that can be applied to improve the evaluation and reporting of context and to help understand if contextual equipoise can be justified or if significant adaptations are required. We hope our analysis offers helpful insight to better understand and ensure that the ethical principle of beneficence is upheld in the real-world contexts of IR in low-resource settings