Incidence trends of airflow obstruction among European adults without asthma: a 20-year cohort study

Investigating COPD trends may help healthcare providers to forecast future disease burden. We estimated sex- and smoking-specific incidence trends of pre-bronchodilator airflow obstruction (AO) among adults without asthma from 11 European countries within a 20-year follow-up (ECRHS and SAPALDIA cohorts). We also quantified the extent of misclassification in the definition based on pre-bronchodilator spirometry (using post-bronchodilator measurements from a subsample of subjects) and we used this information to estimate the incidence of post-bronchodilator AO (AOpost-BD), which is the primary characteristic of COPD. AO incidence was 4.4 (95% CI: 3.5–5.3) male and 3.8 (3.1–4.6) female cases/1,000/year. Among ever smokers (median pack-years: 20, males; 12, females), AO incidence significantly increased with ageing in men only [incidence rate ratio (IRR), 1-year increase: 1.05 (1.03–1.07)]. A strong exposure-response relationship with smoking was found both in males [IRR, 1-pack-year increase: 1.03 (1.02–1.04)] and females [1.03 (1.02–1.05)]. The positive predictive value of AO for AOpost-BD was 59.1% (52.0–66.2%) in men and 42.6% (35.1–50.1%) in women. AOpost-BD incidence was 2.6 (1.7–3.4) male and 1.6 (1.0–2.2) female cases/1,000/year. AO incidence was considerable in Europe and the sex-specific ageing-related increase among ever smokers was strongly related to cumulative tobacco exposure. AOpost-BD incidence is expected to be half of AO incidence.


Internally-derived LLN equations
At each examination, pre-bronchodilator FEV1/FVC values (corrected for the change in spirometer according to Bridevaux  Three-level linear regression models (measurement: level 1 unit; subject: level 2 unit; centre: level 3 unit) were used to calculate the LLN equations separately for males and females in the ECRHS and SAPALDIA studies. The models had random intercept terms at levels 2 and 3, a random slope for age at level 2, an unstructured variance-covariance matrix of the random effects at level 2, a 1 st order autoregressive error at level 1, and examination (1 st , 2 nd or 3 rd measurement), the age × examination interaction term and height as fixed effects. Age was included as a fixed effect in the model for males in ECRHS because the variance of the random slope was negligible. The LLN equations were computed as follows: where: is the predicted FEV1/FVC obtained from the fixed part intercept (b0, b1) and slope coefficients (b2-b4; age is the age of a subject at a given examination and height is the height of a subject at the 2 nd examination);  z-score0.05 is the 5 th percentile of the z-score of FEV1/FVC;  (1) 2 = resi dual 2 / (1   2 ) is the variance at level 1 (resi dual 2 is the residual variance and  is the 1 st order autocorrelation coefficient);  (2) 2 = i ntercept 2 + sl ope 2 • age 2 + 2 • i ntercept, slope • age is the variance at level 2 (i ntercept 2 , sl ope 2 and i ntercept, slope are the components of the variance-covariance matrix of the random effects at level 2; sl ope 2 = i ntercept, slope = 0 for males in ECRHS);  (3) 2 is the variance at level 3.
The estimates of the fixed and random parameters were obtained using the restricted maximum likelihood (REML) method.

Year of age at AO onset
For each new case, the year of age at AO onset (tAO) was estimated by linear interpolation as follows:  FEV1/FVC was obtained for each year of age (t) between the year of age at the last examination with normal spirometry (t1) and the year of age at the subsequent examination with AO (t2) assuming a linear decrease:  LLN was obtained for each year of age (t) between t1 and t2 assuming a linear decrease:  tAO was the first year of age between t1 and t2 with FEV1/FVC(tAO) <LLN(tAO).
As a result of the interpolation process, the incident cases were aged 26 or older.

Reshaped datasets
When computing the IRs of AO, we used a reshaped dataset by age ("long" format) for each sex. The "original" dataset, which contains one record for each subject, was reshaped by replicating each individual record in as many records as the number of age years from the subject's first participation in ECRHS or SAPALDIA (or from age 25 if his/her age at the first study was <25) to the end of follow-up (or to age 64 if his/her age at the end of follow-up was >64). The end of follow-up was the calendar year of AO onset for the incident cases, and the calendar year of the last examination for the remaining subjects. The "new" record of a subject at a given age contained the subject's age and health indicator at that age (i.e. the dummy variable indicating if he/she was an incident case at that age). Therefore, a subject contributed one person-year to the denominator of the IR for each age year without AO, and one unit to the numerator and one person-year to the denominator of the IR for the age year when he/she developed AO. When estimating the trends in AO incidence as a function of lifetime pack-years among ever smokers, reshaping was carried out by pack-years instead.

Stability of AO
We evaluated how many incident cases of AO at ECRHS/SAPALDIA II had one of the following conditions at ECRHS/SAPALDIA III: • pre-bronchodilator FEV1/FVC <LLN; • pre-bronchodilator FEV1/FVC ≥LLN AND ever smoking (≥10 pack-years) AND key respiratory symptoms for COPD (chronic cough, chronic sputum production, dyspnoea, shortness of breath following strenuous activity in the last 12 months); • pre-bronchodilator FEV1/FVC ≥LLN AND self-reported diagnosis of chronic bronchitis, emphysema or COPD; • pre-bronchodilator FEV1/FVC ≥LLN AND father and/or mother with a diagnosis of chronic bronchitis, emphysema or COPD.
Among the incident cases of AO at the 2 nd examination, the subjects who had fulfilled at least one of these conditions at the 3 rd examination were: 57/93 (61.3%) males and 44/66 (66.7%) females.  * Subjects aged ≥25 years at the ECRHS -stage 2 or SAPALDIA I (baseline), who were born in 1945 or later. † Subjects who had valid lung function measurements at baseline and who reported never having been diagnosed with asthma at baseline. ‡ Subjects who had pre-bronchodilator FEV1/FVC LLN at baseline, who participated and had valid lung function measurements at ECRHS/SAPALDIA I-II or ECRHS/SAPALDIA I-II-III, and who reported not having been diagnosed with asthma at the follow-up.  * Subjects who had valid lung function measurements and reported never having been diagnosed with asthma at the ECRHS I -stage 2 (baseline). † Eligible subjects who participated in ECRHS II or ECRHS II-III, and who had valid lung function measurements and reported information on asthma at the follow-up. ‡ Eligible subjects who did not participate in ECRHS II-III, or who did not have valid lung function measurements or did not report information on asthma at the follow-up. * Subjects who had valid lung function measurements and reported never having been diagnosed with asthma at SAPALDIA I (baseline). † Eligible subjects who participated in SAPALDIA II or SAPALDIA II-III, and who had valid lung function measurements and reported information on asthma at the follow-up. ‡ Eligible subjects who did not participate in SAPALDIA II-III, or who did not have valid lung function measurements or did not report information on asthma at the follow-up.