OBJECTIVES: In longitudinal observational studies, patients can meet eligibility criteria for more than one group and can be eligible multiple times for the same group.  We used Epidemiologic Study of Cystic Fibrosis data to explore different inclusion decisions when evaluating dornase alfa treatment. METHODS: The dornase alfa group included patients enrolled ≥2 years before starting consistent dornase alfa therapy.  A lung function test (“index”) separated a 2-year pre-index period from a 2-year post-index period for which intercepts and slopes were independently estimated.  The comparator group included patients not yet reported to have received dornase alfa; their index lung function test was associated with their eighth or subsequent even-numbered birthday.  Comparator patients could contribute more than one set of pre- and post-index periods and could also subsequently be included in the dornase alfa group.  To account for the repeated use of patients, variance components were estimated at the patient level as well as the case level.  Different subsets of the comparator cases were analyzed.  RESULTS: There were 2230 dornase alfa patients; the comparator group included 5970 cases from 3517 patients.  The estimated difference in change in slope was 0.73±.31 (P=0.020).  Subsetting comparators to 4985 cases from 2836 patients not also in the dornase alfa group gave 0.61±0.32 (P=0.058); including each of those patients only the last time eligible gave 0.68±0.36 (P=0.059).  Subsetting to 3662 cases from 2030 patients never on dornase alfa gave 0.32±0.34 (P=0.35).  Patient-level variance components corresponding to difference in slope and difference in intercept were near zero and so were dropped. CONCLUSIONS: In longitudinal observational studies, patients should be included in each group for which they meet eligibility criteria, possibly multiple times (with appropriate covariance structures).  This avoids bias from using future information to decide whether to include a patient and loss of power from limiting cases unnecessarily.