Background: Drug resistance is promoted by intratumour heterogeneity and reversible plasticity, while many adaptive-therapy models divide tumours into only sensitive and resistant compartments. 
Methods: A non-local phenotype-structured advection-diffusion-reaction model was coupled to drug exposure, toxicity and feedback treatment. Positivity, boundedness, moment dynamics and an instantaneous fitness optimum were derived. A conservative positivity-preserving finite-volume method compared continuous, fixed intermittent, burden-adaptive and phenotype-aware therapy. 
Results: In synthetic experiments, progression occurred at 618.5, 668.5, 742.75 and 762.5 days, respectively. Phenotype-aware feedback prolonged control by 23.3% relative to continuous treatment while using 15.3% fewer treatment-equivalent days. Across 160 Latin hypercube parameter sets, median progression times were 1.67, 1.72, 1.94 and 2.03 years. Grid refinement showed convergent burden and mean-phenotype estimates without negative densities. 
Conclusion: Phenotype information can improve adaptive decisions when plasticity and resistance costs materially affect future response. The framework is methodological and requires tumour-specific calibration before clinical use.