Randomised controlled trials in nutrition (RCTN) face unique challenges, including the considerable influence of the background diet and the challenge of assuring intervention adherence by participants. The impact of these factors on the outcome of RCTNs has been difficult to quantify, but nutritional biomarkers represent a valuable tool to address these challenges. Using flavanols as a model dietary intervention and a set of recently validated flavanol biomarkers, we here investigated the impact of background diet and adherence on the outcomes of a subcohort of the COcoa Supplement and Multivitamin Outcomes Study (COSMOS, NCT02422745). We found that 20% of participants in the placebo and cocoa-extract intervention arms had a flavanol background intake as high as the intervention, and only 5% did not consume any flavanols. Approximately 33% of participants in the intervention group did not achieve expected biomarker levels from the assigned intervention - more than the 15% estimated with pill-taking questionnaires usually implemented in RCTN. Taking these factors into account resulted in a larger effect size for all observed endpoints (HR (95% CI)) estimated using intention-to-treat vs. per-protocol vs. biomarker-based analyses: total cardiovascular disease (CVD) events 0.83 (0.65; 1.07); 0.79 (0.59; 1.05); 0.65 (0.47; 0.89) - CVD mortality 0.53 (0.29; 0.96); 0.51 (0.23; 1.14); 0.44 (0.20; 0.97) - all-cause mortality 0.81 (0.61; 1.08); 0.69 (0.45; 1.05); 0.54 (0.37; 0.80) -- major CVD events 0.75 (0.55; 1.02); 0.62 (0.43; 0.91); 0.48 (0.31; 0.74). These results highlight the importance of taking background diet and adherence into consideration in RCTN to obtain more reliable estimates of outcomes through nutritional biomarker-based analyses.