Introduction: The proportion of chronic airflow obstruction (CAO) due to smoking (Population Attributable Fraction, PAF) is often estimated at around 50% and thought to be lower in poorer countries.

Aims: To estimate the impact of smoking on CAO in rich and poor countries.

Methods: A cross-sectional survey of adults >40 years old estimated the prevalence of CAO (FEV1/FVC < lower limit of normal, LLN) and the prevalence of smoking in 33 sites. Local estimates of relative risk adjusted for confounders, whose accuracy was improved using hierarchical Bayesian models, were combined with smoking rates to derive local estimates of PAF. Prevalence of CAO caused by smoking (Population excess risks, PER) were also estimated.

Results: Prevalence of ever-smokers ranged from 2% (Cotonou, Benin) to 68% (Cape Town, S. Africa). In a third of sites prevalence of CAO among non-smokers was <5% (the expected value defined for a healthy non-smoking population by the LLN). The relative risk from smoking was similar across sites (2.15; 95%CI: 1.9 to 2.4; I2=0%), and so PAF was strongly related to smoking prevalence, and ranged from <10% (Pune, India; Ile-Ife, Nigeria; Blantyre, Malawi) to >50% (Cape Town; London, UK). PER values ranged from <1% (Benin; Ile-Ife; Pune; Blantyre; Penang, Malaysia; Riyadh, Saudi Arabia) to over 10% (Lexington, USA; Maastricht, Netherlands; London).

Conclusions: Smoking is the most important cause of CAO in both rich and poor regions of the world. PAF values are mostly <50%, but this is not easy to interpret in areas with low CAO prevalence and where 5% of “normal” people are defined as “diseased” by the LLN. PER values are less dependent on reference criteria.