OBJECTIVES: The ability to model cancer recurrence could assist in the optimization of surveillance strategies. However, capturing the dynamics of cancer recurrence in order to simulate follow-up surveillance after initial extirpative surgery presents a significant methodological challenge. The difficulty of modeling recurrence patterns is that relevant experimental and observational data is collected in the context of heterogeneous protocols for follow-up.  Using the example of colorectal cancer, we propose a method of controlling for choice of follow-up regimen in order to infer the value of key natural history parameters.  Once these values are inferred, any hypothetical follow-up regimen can be superimposed upon the natural history model to project clinical and/or economic outcomes. METHODS: The subset of stage I-III colon cancer patients who will experience recurrence face a constant rate rd of transition from undetectable to theoretically detectable recurrence during a given interval.  These same patients face a constant rate ru of transition from resectable (i.e. potentially curable) to unresectable metastatic disease with a minimum interval xdu between the point of detectability and the point of unresectability.  A third constant rate parameter rs will determine when, on average, individuals develop recurrence-related symptoms prompting them to seek medical advice before the next scheduled evaluation.  The mean point of symptom development will follow the point at which a recurrence becomes detectable by a span of at least xds.  However, a normally distributed error term Eds will mean that, for a given simulated patient, symptoms may initiate before or after the patient reaches unresectability.  RESULTS: A best-fitting set of these natural history parameters can be selected by calibrating to targets of time-to-detection of recurrence, time-to-death, and proportion of patients who present with recurrence-related symptoms prior to scheduled assessments.  CONCLUSIONS: The data sources for these targets can be existing experimental, observational, or registry data where follow-up schedule and compliance levels are known.