Nonspecific drug partitioning into microsomal membranes must be considered for in vitro-in vivo correlations. This work evaluated the effect of including lipid partitioning in the analysis of complex TDI kinetics with numerical methods. The covariance between lipid partitioning and multiple inhibitor binding was evaluated. Simulations were performed to test the impact of lipid partitioning on the interpretation of TDI kinetics, and experimental TDI datasets for paroxetine (PAR) and itraconazole (ITZ) were modeled. For most kinetic schemes, modeling lipid partitioning results in statistically better fits. For MM-IL simulations (K I,u = 0.1 mM, k inact = 0.1 minute 21), concurrent modeling of lipid partitioning for an f umic range (0.01, 0.1, and 0.5) resulted in better fits compared with post hoc correction (AICc: 2526 vs. 2496, 2579 vs. 2499, and 2636 vs. 2579, respectively). Similar results were obtained with EII-IL. Lipid partitioning may be misinterpreted as double binding, leading to incorrect parameter estimates. For the MM-IL datasets, when f umic = 0.02, MM-IL, and EII model fits were indistinguishable (dAICc = 3). For less partitioned datasets (f umic = 0.1 or 0.5), the inclusion of partitioning resulted in better models. The inclusion of lipid partitioning can lead to markedly different estimates of K I,u and k inact. A reasonable alternate experimental design is nondilution TDI assays, with post hoc f umic incorporation. The best fit models for PAR (MIC-M-IL) and ITZ (MIC-EII-M-IL and MIC-EII-M-Seq-IL) were consistent with their reported mechanism and kinetics. Overall, experimental f umic values should be concurrently incorporated into TDI models with complex kinetics, when dilution protocols are used.