Ensuring that research results are implemented is complex. The extended Normalization Process Theory (ENPT) considers the agency that actors have to implement new knowledge in revised practices. In this thesis, I adapted the ENPT to explain constraints met by humanitarian actors when they implement public health responses based on empirical research, in conflict-affected settings. I adapted the ENPT (a-ENPT) in four phases. First, I conducted a scoping review of peer-reviewed papers (n=22). Second, I analysed theoretical approaches cited by authors (n=43) in my initial review. Third, I conducted semi-structed interviews (n=26) with humanitarian practitioners, donors, and academics. Finally, I applied the a-ENPT in two Médecins-Sans-Frontières settings (Pakistan and Guinea) through document review (n=67), social networks analysis (n=2), qualitative interviews (n=20), and a consultative meeting. I present several results. First, most peer-reviewed papers documented mental health interventions (n=15/22) and frontline actors adapted recommendations and negotiated constraints. Second, theoretical approaches (n=43) influenced how frontline actors discussed knowledge and resources distribution. Third, interviewees (n=26) conveyed that frontline actor influenced power imbalances, and that engaging communities meaningfully was challenging. Fourth, field-based research motivated actors and trade-offs were negotiated collectively. Findings from this thesis highlight that frontline actors discuss knowledge, adapt recommendations, negotiate contextual constraints, influence power imbalances, or innovate when they are engaged meaningfully. Using a theoretical approach accounting for their agency is key in conflict-affected settings. My work has several limitations. My scoping review allowed to identify a gap in literature reviewed by peers but does not represent grey literature. My results do not allow for measure of implementation outcomes or cost-effectiveness and are based on data collected within western entities. The a-ENPT should be tested prospectively, within a grassroots organization and in a setting that allows for direct engagement of communities while allowing for improved documentation of outcomes and longer-term economic costs.