Underdiagnosis of Obstructive Lung Disease

Findings From the French CONSTANCES Cohort

Marie-Christine Delmas; Laetitia Bénézet; Céline Ribet; Yuriko Iwatsubo; Marie Zins; Rachel Nadif; Nicolas Roche; Bénédicte Leynaert


BMC Pulm Med. 2021;21(319) 

In This Article


Study Population

The methodology of the CONSTANCES cohort study has already been described.[10] For each year of inclusion, individuals aged 18–69 years, affiliated to the main national health insurance covering around 85% of the population, and living in selected administrative areas in France (called "départements"), were randomly chosen to participate in the study according to an unequal probability sampling stratified by gender, age, social category and area. Participants who gave informed consent completed a self-administered questionnaire and underwent a health examination at one of the health prevention centres (HPCs) in the selected areas. Additionally, data were collected for participants and a random sample of non-participants through a linkage to two national databases: the National Health Database (SNDS) that covers all reimbursements for outpatient and hospital healthcare, and the National Retirement Insurance Database that gathers occupational data throughout life. The longitudinal follow-up of participants is still ongoing. This includes annual self-administered questionnaires, health examinations every 4 years, and passive data collection by linkage to the two national databases.

The present analysis was conducted using data collected at inclusion in 2013–2014. The study population comprised all participants with spirometry results (Figure 1).

Figure 1.

Flow chart of the study population

Data Collected at Inclusion

The self-administered questionnaire included data on sociodemographic characteristics, occupational history, lifestyle, and health. Questions on respiratory health, taken from the European Community Respiratory Health Survey (ECRHS) questionnaire, covered asthma-like symptoms, chronic bronchitis, and dyspnoea.[11] Dyspnoea was quantified using the modified Medical Research Council scale.[12]

Participants were interviewed by a physician about their medical history, including respiratory and cardiovascular diseases, and underwent a comprehensive health examination that included anthropometric measurements, blood pressure, electrocardiogram, vision and hearing tests, spirometry, and blood and urine sampling for biological testing. For participants aged 45 years and older, a specific work-up of functional, physical, and cognitive capacities was performed. To maintain high-quality standards in the measurements, strict quality management was implemented as previously described.[13]

Spirometry and Airflow Limitation

Spirometry was performed following ATS/ERS guidelines without administering a bronchodilator, as French HPCs are not allowed to administer medication even for diagnostic purposes.[14] For the present study, repeatability criteria were extended to a 200 mL threshold for both forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC). FEV1 and FVC were expressed as percentages of predicted values (FEV1% predicted and FVC% predicted) using the 2012 Global Lung Initiative (GLI) predictive equations.[15] Airflow limitation was defined using the Global Initiative for Obstructive Lung Disease (GOLD) criteria as FEV1/FVC < 0.70 and the lower limit of normal (LLN) criteria with GLI equations. Airflow limitation severity was determined using GOLD criteria (mild: FEV1 ≥ 80% predicted; moderate: [50–80%[; severe: [30–50%[; very severe: < 30% predicted).[12]

Statistical Analysis

The outcome was undiagnosed OLD, defined as spirometry-confirmed airflow limitation with no previous diagnosis of OLD (asthma, COPD, emphysema, chronic bronchitis, or bronchiectasis) reported by participants during the medical interview. Participants with undiagnosed OLD were compared to those with diagnosed asthma and COPD for demographic and anthropometric characteristics, education level (coded with the International Standard Classification of Education (ISCED)[16]), lifetime tobacco consumption (none; > 0 and < 10 pack-years; ≥ 10 pack-years), respiratory symptoms (wheezing in the past 12 months, chronic cough or sputum, dyspnoea), cardiovascular history, and spirometry results. Multivariate analysis was performed using robust Poisson regression models to estimate adjusted prevalence ratios of undiagnosed OLD according to the factors studied. Two models were built: the first was adjusted for sociodemographic characteristics (gender, age, education level in two groups of similar size), tobacco consumption, and clinical data (respiratory symptoms, cardiovascular comorbidities), and the second for these features plus FEV1% predicted.

All analyses incorporated appropriate weights. Using the sampling weights calculated for each year of inclusion, two consecutive steps of reweighting were performed: the first took into account non-participation in the clinical exam, and the second the unavailability of spirometry results in participants who attended the clinical exam. Reweighting was performed with the equal quantile score method using demographic, socioeconomic, and health-related data, along with inclusion data (tobacco status, presence of respiratory symptoms, body mass index, blood pressure) for the second step.[17] Finally, as 2013 and 2014 samples were independent, annual weights were combined.[18] Statistical analyses were performed using Stata v14 (Stata Corporation, College Station, Texas, USA).