In studies done in HICs, the risks of all-cause mortality were about 2–3 times higher in current than in never smokers. For example, in the British Doctors’ Study of 34 439 men, followed up from 1951 to 2001, the relative risk (RR) was 2·19 for the cohort (aged ≥60 years) born in the 20th century. In the UK's Million Women Study of 1 180 652 women (median age 55 years), followed up from 1996 to 2011, the RR was 2·76 (95% CI 2·71–2·81). In the US National Health Interview Survey of 216 917 individuals (aged ≥18 years), followed up from 1997 to 2006, the HR was 2·8 (99% CI 2·4–3·1) in men and 3·0 (2·7–3·3) in women. By contrast, studies in MICs and LICs have reported lower hazards for all-cause mortality than those seen in HICs. For example, in a cohort study of 224 500 men (aged ≥40 years) in China, followed up from 1990 to 1996, the RR was 1·19 (95% CI 1·13–1·25). In a prospective study of 118 840 adults (aged 30–69 years) in Cuba, followed up from 1996 to 2017, the RR was 1·66 (95% CI 1·58–1·74). In a nationally representative case-control study of 152 058 adults (aged ≥20 years) done in India from 2001 to 2003, the risk ratio was 1·7 (99% CI 1·6–1·8) in men and 2·0 (1·8–2·3) in women.
The variations in risks associated with smoking between country income groups are not fully explained by the differences in risk factors, smoking patterns (ie, age at initiation, quantity, duration, and pack-years of smoking), the tobacco products used, or competing risks. In our analyses, we accounted for heterogeneity in smoked tobacco products by adjusting for pack-years, which included converting non-cigarette smoked products to cigarette-equivalents using standard methods. In PURE, bidi smoking was more common in LICs than in MICs and HICs, and smokeless tobacco use was more frequent in HICs (mostly snuff) and LICs (mostly chewing tobacco) than in MICs. However, compared with never smokers, exclusive bidi smokers and exclusive cigarette smokers had similar risks in LICs. Furthermore, the exclusion of smokers who were also using smokeless tobacco did not alter the main results. The cumulative incidence of the competing event for mortality from smoking was low over 11 years in all three country income groups, and so the results were similar to those from Cox models.