EPS6.09 Association of within-individual variability of FEV1 and BMI with mortality in women with cystic fibrosis: preliminary results from the UK Registry

Abstract

Malnutrition and lung function have been associated with mortality in adults with cystic fibrosis (CF), but many longitudinal analyses have focused only on the average trend of FEV1 and BMI over time. Their approach disregards potentially useful information, particularly the fluctuations of the repeated measurements around the mean trajectory (within-individual variability), which could impact on females with CF and their survival gap with males. Valid estimates of the association of the two correlated measures and their within-individual variability with mortality are needed to provide a richer characterisation of disease progression. We propose a novel statistical model to simultaneously analyse the correlation between FEV1 and BMI over time, their within-individual variability and their association with mortality. We examined annual review data from a randomly sampled cohort of women from the UK CF registry (N = 657, aged >18 years, with administrative censoring at the age of 50). A proportional hazards model is used to quantify the associations of FEV1, BMI and their within-individual variabilities with the risk of death, adjusted for year of birth, age at diagnosis and F508 genotype. Women with higher BMI tend to have higher BMI within-individual variability and higher FEV1 on average. Lower risk of death is obtained for women with higher mean FEV1 in dL (HR = 0.829, 95% Bayesian credible interval [0.802, 0.857]) and higher BMI (HR = 0.920 [0.872, 0.973]). Evidence of non-zero within-individual variability is found for both FEV1 and BMI, but only higher FEV1 within-individual variability in dL is associated with a higher risk of death (HR = 1.126 [1.007, 1.260]). Modelling jointly FEV1 and BMI over time in adults brings new insights on the evolution of cystic fibrosis and the quantification of mortality risk. A better characterisation of these associations could improve dynamic risk prediction in clinical settings in the future.