Background In the UK, chronic obstructive pulmonary disease (COPD) causes approximately 25,000 deaths annually. Around three million people are estimated to have COPD, of whom fewer than one-third have been diagnosed. Using spirometry, fixed thresholds (FT) and lower limit of normal (LLN) criteria define airflow obstruction (AO) differently. FTs use cut-offs for lung function measurements regardless of age, sex, height and ethnicity; LLN criteria use cut-offs (e.g. 5th centile) based on normalised values for a healthy person of a given age, sex, height and ethnic group. Although FTs are widely used, LLN criteria are recommended in epidemiological studies as FTs overestimate AO in the elderly due to age-related changes in lung function.

Methods The Health Survey for England 2010 focused on respiratory health and included spirometry (without bronchodilators) in a nationally representative random general population sample. For participants aged ≥45years without asthma (n = 1635), we estimated the prevalence of self-reported doctor-diagnosed COPD. Using the Global Lungs Initiative 2012 multi-ethnic reference equations, prevalence rates of probable AO were estimated using four definitions:

FT: FEV₁/FVC<0.70

FT + : FEV₁/FVC<0.70 and FEV₁<80% predicted

LLN: FEV₁/FVC

LLN + : FEV₁/FVC and FEV₁

These definitions are overlapping. We estimated the prevalence of COPD within each probable AO group. Potential risk factors for COPD or for probable AO were analysed using logistic regression (five separate outcomes).

Results 5% of participants reported doctor-diagnosed COPD. The prevalence of probable AO was 24% (FT), 12% (FT +), 14% (LLN) and 7% (LLN +). 17% of participants in the FT + group reported COPD compared with 24% of participants in the LLN + group. Pack-years smoked and respiratory symptoms significantly increased the odds of COPD and of AO after adjustments for age, sex, socioeconomic status and exposure to passive smoking. Compared with never regular smokers, those with 50 + pack-years had 5.8 (95% CI 2.6, 13.0) times higher odds of reporting diagnosed COPD, and 5.5 (3.0, 10.3) and 6.8 (3.2, 14.8) times higher odds of probable AO by FT + and LLN + criteria, respectively. Participants with respiratory symptoms had 2.5 (1.1, 5.7), 2.0 (1.0, 4.0) and 2.8 (1.3, 5.7) times higher odds of COPD/AO respectively than those without respiratory symptoms. Participants meeting LLN probable AO criteria had higher odds of diagnosed COPD: FT 2.0 (1.2, 3.5); FT + 40 (2.2, 7.1); LLN 4.2 (2.4, 7.3); LLN + 5.7 (3.0, 10.8).

Conclusion As post-bronchodilator spirometry was not used in HSE 2010, our results may overestimate the true prevalence of AO, but this large population survey confirms that substantial underdiagnosis of COPD remains likely. LLN criteria appear to identify fewer, more severely affected individuals than FT criteria.