The Tofts tracer kinetic models are often used to analyze dynamic contrast-enhanced MRI data. They are derived from a general two-compartment exchange model (2CXM) but assume negligible plasma mean transit time. The 2CXM estimates tissue plasma perfusion and capillary permeability-surface area; the Tofts models estimate the transfer constant K(trans), which reflects a combination of these two parameters. The aims of this study were to compare the 2CXM and Tofts models and report microvascular parameters in patients with cervical cancer. Thirty patients were scanned pretreatment using a dynamic contrast-enhanced MRI protocol with a 3 sec temporal resolution and a total scan duration of 4 min. Whole-tumor parameters were estimated with both models. The 2CXM provided superior fits to the data for all patients (all 30 P values <0.005), and significantly different parameter estimates were obtained (P <0.01). K(trans) (mean = 0.35 +/- 0.26 min(-1)) did not equal absolute values of tissue plasma perfusion (mean = 0.65 +/- 0.56 mL/mL/min) or permeability-surface area (mean = 0.14 +/- 0.09 mL/mL/min) but correlated strongly with tissue plasma perfusion (r = 0.944; P = 0.01). Average plasma mean transit time, calculated with the 2CXM, was 22 +/- 16 sec, suggesting the assumption of negligible plasma mean transit time is not appropriate in this dataset and the 2CXM is better suited for its analysis than the Tofts models. The results demonstrate the importance of selecting an appropriate tracer kinetic model in dynamic contrast-enhanced MRI.