OBJECTIVES:   To evaluate lung function decline before and after the initiation of dornase alfa (DA) through a multivariable mixed-effects piecewise linear model using data from the Epidemiologic Study of Cystic Fibrosis (ESCF).  METHODS:   Patients aged 8-38 years enrolled in ESCF for 2 or more years prior to initial treatment with DA were selected if they remained on treatment for at least 2 years.  A comparator group included cystic fibrosis patients not yet reported to have received DA.  FEV1 percent predicted (pp) was analyzed before and after an index measurement within 30 days of either DA initiation (DA group) or an encounter within 1 year following the 8th or subsequent even-numbered birthday (comparator group).  For each patient, we fit a regression line to FEV1 pp separately for the pre-index and post-index periods (both 2 years in duration) using a mixed-effects piecewise linear model adjusted for age, gender, pulmonary exacerbations, respiratory therapies, and nutritional supplements.  Patients were categorized by age group or by age-adjusted deciles of the index FEV1 pp.  RESULTS:   The DA group (n=2,230) had a lower FEV1 pp at index and a more rapid decline during the pre-index period.  There was an acute improvement in FEV1 pp (change in intercept) associated with the initiation of DA therapy.  Furthermore, the mean rate of FEV1 pp decline was more attenuated for the DA group than for the comparator group (n=5,970) across age groups and deciles.  CONCLUSIONS:   The use of DA for a 2-year period is associated with both an acute improvement in FEV1 pp (previously shown in clinical trials) and a reduction in the rate of FEV1 pp decline (shown for the first time).  These results demonstrate the value of using mixed-effects piecewise linear models in observational studies to evaluate the effect of instituting a therapy on both the slope and intercept of a continuous outcome.