OBJECTIVES: Cancer treatment models are often based on progression-free survival (PFS) and overall survival (OS) data.  If the model objective requires extrapolating results or exploring “what-if” scenarios, disease progression parameters are calibrated so that the model replicates the PFS and OS data.  Uncertainties in the estimation of the Kaplan-Meier survival curves used as calibration targets, and in the model calibration process itself, are not commonly incorporated into sensitivity analyses.  The objective of this study was to demonstrate methods for incorporating these uncertainties into probabilistic sensitivity analyses (PSA) and to explore their implications. METHODS: We constructed hypothetical PFS and OS survival (with censoring) for two treatments (TxA & TxB) and a corresponding three-state Markov model (Non-progressed (NP), Progressed (P), Dead (D)).  Health states were assigned costs and utilities consistent with advanced cancer.  Three transition probabilities for each treatment (NP->P, NP->D, P->D) were calibrated (using Excel Solver) to simultaneously fit (using mean squared deviation) the PFS/OS curves.  We performed three increasingly comprehensive PSAs using second-order Monte Carlo simulation (SMCS): 1) conventional PSA including only probability distributions of costs and utilities; 2) specifying beta distributions for failure probabilities at each PFS/OS time point, simulating multiple replicates of the PFS /OS data from these distributions, re-estimating and refitting the curves for each replicate, and incorporating the resulting calibrated parameter sets into the SMCS; and 3) incorporating different curve-fitting methods by varying Solver parameters (initial values, constraints, objective function).  Uncertainty in cost-effectiveness results was represented by cost-effectiveness acceptability curves (CEAC). RESULTS: At a threshold of $100,000/QALY, the CEAC for PSA1 showed a 97% probability that TxB is cost-effective versus TxA, corresponding results for PSA2, and PSA3, were 54%, and 58%, respectively. CONCLUSIONS: Failure to consider uncertainties owing to PFS/OS data in oncology models, and to the ensuing calibration procedures, can lead to under-representation of uncertainty in cost-effectiveness results.