OBJECTIVE: Continuous enrolment requirement becoming a “rule” rather than “exception” in pharmaeconomical studies. However, continuous enrolment requirement contains potential bias especially patients with incomplete data structurally different from the ones with complete data. Moreover, for studies involving rare diseases the requirement significantly reduces sample size and creates power issues. In this paper, we outline the methodology which can be used to remove continuous requirement. METHODS: This is a two stage method. First, we estimate probability of being continuously enrolled using every observation in our data set. This estimation can be done either with parametric methods (such as logit, probit) or non-parametric methods (such as Kaplan-Meier). Then according to these probabilities we create a weight which is inverse of the estimated probabilities at first stage. Using these weights and only observations which are continuously enrolled we estimate weighted least square models at the second stage. Therefore second stage regressions use probabilities containing information from everybody at the first stage. We showed that this eliminate possible selection bias due to continuous enrolment. RESULTS: We used medicare claim files for application. A total of 773 patients with incident cases of lung, prostate, colon and breast cancer were analyzed for two years of cost. Continuous enrolment requirement would decrease the sample size to 541 and would create potential selection bias. We estimated the results with and without continuous enrolment requirement and showed that the results are statistically different from each other (p=0.000). CONCLUSIONS: Continuous enrolment requirement should not be applied blindly. Data sets created based on this requirement yields consistent results if there is no systematic differences between complete and incomplete observations.