Estimation of lung function decline from the longitudinal COSYCONET cohort: does drop-out information matter?

<bold>Background:</bold> Health economic decision models for COPD are based on stage transition probabilities which depend on lung function decline. Longitudinal spirometry data is available from large cohort studies, but hampered by high intraindividual variation over short-term follow-up and high drop-out rates over longer follow-up periods. <bold>Aim and objectives:</bold> To estimate average lung function decline in COPD patients over at least 1.5 years and to test if participants with complete follow-up differ from drop-outs classified into different categories. <bold>Methods:</bold> In the German COSYCONET cohort, patients with COPD were followed for up to 9 years with lung function measurements every 1.5 year. For all patients terminating before the last study visit, reasons for drop-out were categorized. Annual decline in forced expiratory volume in 1 second (FEV1) as percent predicted (“%p”) was the primary outcome. Linear mixed models with FEV1%p were used to compare participants who remained in the study with those in various drop-out groups. <bold>Results:</bold> 2065 patients with at least 18 months follow-up were included in the analysis. Ignoring drop-out information, the estimated annual decline in lung function was 0.76 units of FEV1 %p (or 42 ml). Patients with at least 6 years follow-up had significantly lower FEV1 decline than those dropping out earlier (0.70 %p vs. 1.21 %p, or 40 vs. 53 ml per year). This difference was even more pronounced when stratified by GOLD stages 1-4. Differentiating by drop-out categories showed further significant differences. <bold>Conclusions:</bold> Ignoring drop-outs and information on drop-out reasons in the calculation of lung function decline may underestimate COPD progression rates.