• Epidemiologic studies
• health services
• incidence
• prognosis
• chronic obstructive
• pulmonary disease

We appreciate the issues raised by Dr Hoesein regarding our paper1 and respond as follows.

First, the NELSON study is a randomised screening trial for lung cancer and therefore targets subjects at high risk. The focus of our paper is different. We evaluated the prognostic meaning of a change in prebronchodilation spirometry—a common outcome in environmental epidemiology—in terms of respiratory health, not specifically bound to chronic obstructive pulmonary disease (COPD). We showed that this outcome has value in epidemiological research, but that understanding of the early stages of airflow obstruction, where non-persistence seems highest, remains poor. These early stages are a commonly affected outcome by inhaled environmental toxicants (eg, air pollutants) and therefore of public health relevance. Second, age was not predictive of non-persistence in our study (see table 4 in the online supplement). Third, the non-persistence rate was lower in heavy smokers (14.0 cases/1000 person-years in heavy baseline smokers vs 19.2 cases/1000 person-years in all participants, see table 4 in online supplement) and approaches that observed in the more homogeneous NELSON study which included only smokers. Fourth, it is well established that additional risk factors beyond active smoking contribute to age-related lung function decline. COPD is acknowledged as being more than a smoking-related disease.2 In the SAPALDIA cohort one-third of subjects with airflow obstruction at follow-up examination were never smokers.3 Previous estimates from developed countries also ranged from 17.0% to 38.8%.2 Fifth, it is important to differentiate between actual measurement error and true measurement variation. Lung function may show considerable variation over time due to different reasons, including changes in environmental conditions. This explains some of the observed non-persistence and incidence. From a population-based perspective it is, however, the average distribution of health parameters which matters and not the classification of individuals into fixed clinical categories. Sixth, previously reported values and SDs of longitudinal lung function change in the SAPALDIA cohort4 were very comparable to those found in other studies—for example, the ECRHS.5 SAPALDIA has applied very stringent quality control procedures including different ‘round robin’ studies comparing devices and fieldworkers, as referenced in our paper.

However, we thank Dr Hoesein for pointing to the unusually low SDs in the percentage predicted values of forced vital capacity (FVC) and forced expiratory volume in 1 s (FEV₁)/FVC. This is indeed a mistake in table 1 and the corrected values are provided in the revised table 1 shown here. Further, we would like to correct the third author's name which is PO Bridevaux and not B Pierre-Olivier.