Background: Randomized controlled feeding trials are ideal for studying diet-disease relationships, but interpreting results requires accurate dietary adherence measures. Aim: This study assesses dietary adherence using self-reported measures and estimates from predictive equations in the National Institute of Diabetes and Digestive and Kidney Diseases Body Weight Planner and evaluates differences in adherence across diet treatments and study periods in a feeding trial. Methods: In a randomized, counterbalanced, crossover trial, 12 African American adults with prediabetes or early, untreated type 2 diabetes received either a plant-based diet (PBD) or an isocaloric control diet for eight weeks, followed by a washout period and the alternate diet. Primary measures included self-reported dietary adherence and estimates of actual caloric intake using predictive equations. T-tests evaluated differences by treatment and period. Results: Nine participants completed the study. Study completers (n = 9) consumed a greater caloric excess beyond the kilocalories provided by the study diets during the control treatment (480.6 ± 525.9 kcal/day) versus the PBD treatment (126.0 ± 585.4 kcal/day; p = 0.025) and in the second period (464.8 ± 592.4 kcal/day) versus the first period (141.8 ± 529.5 kcal/day; p = 0.033). Participants also reported omitting more kilocalories from the study diets during the PBD treatment (422.4 ± 289.1 kcal/day) versus the control treatment (276.4 ± 185.0 kcal/day; p = 0.015). Conclusion: Predictive equations showed significant differences in dietary adherence by treatment and period, underscoring the limitations of self-reported intake and highlighting the need for more objective measures. Relevance for patients: Variable adherence to study diets suggests the original study should be viewed as an effectiveness study, not an efficacy study.