Validation of a predictive survival model in Italian patients with cystic fibrosis

BACKGROUND: In 2001 Liou published a 5-year survival model using CFF Registry data. AIMS: To evaluate its validity in predicting survival in Italian CF patients. METHODS: In a retrospective study on 945 patients, the 9 variables selected by Liou were analyzed, vital status on December 2008 recorded and observed and expected deaths compared. To develop a new model, patients were randomly divided into a derivation (n=475) and a validation sample (n=470). RESULTS: A significant difference was found between observed and expected deaths based on Liou's model (62 vs 94), with a 34% reduction in mortality (p<0.05). A new model (based on FEV1, Staphylococcus aureus and Burkholderia cepacia complex infection, number of pulmonary exacerbations/year) was generated, that correctly predicted survival in the validation sample (31 observed vs 29 expected deaths, p=0.660). CONCLUSIONS: The Liou model did not adequately predict 5-year survival in our CF population that, compared to the one in which it was originally tested, could benefit from 10 years of improvement in treatments and practice patterns. A new generated model, based on only four variables, was more accurate in predicting 5-year survival in Italian CF patients.