A methodology is formulated to analyze tumor recurrence data when its incidence rate is restricted due to medication. Analytic results are derived to make a probabilistic early warning of tumor recurrence free period of length τ; that is, the chance for a safe period of lengthτ is estimated. The captured data are length biased. Expressions are developed to extract and relate to counterparts of the non-length biased data. Three data sets are considered as illustrations: (1) patients who are given a placebo, (2) patients who are given the medicine pyridoxine and (3) patients who are given the medicine thiotepa.