A prognostic model for the prediction of survival in cystic fibrosis

Abstract

Background: The treatment for endstage cystic fibrosis is, where appropriate, double-lung, heart-lung or, occasionally, heart-lung-liver transplantation. Optimising the timing of transplantation depends upon an accurate prediction of survival, but while current criteria give some guidance to this, they are not based upon statistically derived prognostic models.

Methods: Data collected prospectively on 403 patients with cystic fibrosis, recruited between 1969 and 1987 (cohort A), were analysed by log rank and univariate Cox regression analysis to determine variables that accurately predict survival. The significant variables were then subject to time dependent multivariate Cox regression analysis to generate a prognostic model. The model was validated, within the study population, using split sample testing, and was subsequently validated in a further cohort of patients recruited since October 1988 (cohort B).

Results: One hundred and eighty eight (50.4%) of the study cohort died within the study period. Percentage predicted forced expiratory volume in one second (FEV1), percentage predicted forced vital capacity (FVC), short stature, high white cell count (WBC), and chronic liver disease (as evidenced by the presence of hepatomegaly) were negatively correlated with survival. These variables, when combined into a prognostic index, accurately predicted one year survival in the study population and in the cohort recruited since 1988.

Conclusion: This prognostic index may prove valuable in predicting prognosis in other cohorts with cystic fibrosis and thereby improve the timing of transplantation.