To the Editor: We appreciated the letter by Neuhoff and Tucker, which corroborated the methodology adopted in our recent study, “Physiologically based modeling of pravastatin transportermediated hepatobiliary disposition and drug-drug interactions” (1), while providing further comments on the scope of ‘bottomup’ prediction of transporter-mediated drug-drug interactions (DDIs). We agree that there is a growing knowledge in the in vitro-in vivo extrapolation (IVIVE) of transporter data, which could facilitate a comprehensive “bottom-up” approach in the future. However, there are considerable knowledge gaps at this stage which made us adopt a ‘middle-out” approach, wherein, a whole-body physiologically based pharmacokinetic (PBPK) model was developed based majorly on preclinical data and later refined using clinical pharmacokinetic data. The apparent discrepancy in the active transport kinetics could be qualitatively justified by the differences in the protein expression levels. For example, sandwich cultured human hepatocyte (SCHH) system showed up-regulated MRP2, but down-regulated OATP transporters, based on the protein quantification generated in our laboratory (2, 3). Although a ~5-fold higher MRP2 expression in SCHH could justify the scaling factor estimated for canalicular efflux of pravastatin, such direct translation is not apparent for sinusoidal uptake transporters (OATPs relative expression factor ~2-7 (2) versus scaling factor ~31(1)). In fact, based on our previous studies (4), the scaling factors derived using similar approach seem to be compound-specific, which suggests the need to understand differences in both transporter expressions, as well as transporter functional activity of the in vitro experimental systems compared to that in vivo. Collectively, protein quantification data are undoubtedly essential for model building and provide confidence in the translation of in vitro parameters, however, the discrepancies in IVIVE cannot be explained by only considering transporter expression levels. While further understanding in this area is warranted, we strongly believe that refining the mechanistic model based on clinical pharmacokinetic data provides the confidence necessary to make quantitative DDI predictions. The estimated in vivo Ki values for cyclosporine and gemfibrozil are more potent compared to mean in vitro values. Interestingly, this trend is consistent with the CYPmediated DDIs, where relatively large differences were noted between the in vitro and in vivo Ki values for lipophilic inhibitors (5). Nevertheless, we note that the estimated in vivo Ki values also predicted DDIs for other OATP1B1 substrate drugs (6). We agree that the discrepancies between in vitro and in vivo Ki demand for careful considerations, with an emphasis on elucidating the mechanism of inhibition (competitive versus possible time-dependent inhibition). However, lack of data on interaction parameters such as Kdeg (degradation rate constant) for drug transporters precludes incorporation of time-dependent inhibition mechanism in transporter DDI predictions. Finally, the other major challenge in prospective predictions of the transporter-mediated DDIs is sparse availability of the clinical interaction data to validate the mechanistic models. As noted by Neuhoff and Tucker, the majority of interaction studies of various statins with cyclosporine were done in a organ-transplant patient population, who may not only have altered physiology but may also exhibit downregulation of drug transporters and metabolic enzymes. While the influence of elevated cytokines on transporter activity is not completely understood in vivo, decreasing the relative expression factor in the model for organ-transplant M. V. S. Varma (*) : Y. Lai : B. Feng : J. Litchfield : T. C. Goosen Pharmacokinetics, Dynamics and Metabolism, Pfizer Inc. Groton, Connecticut, USA e-mail: manthena.v.varma@pfizer.com