Authors: Sitas F, Egger S, Bradshaw D, Groenewald P, Laubscher R, Kielkowski D, Peto R

Source: The Lancet, Volume 382, Issue 9893, Pages 685-693


Background: The full eventual effects of current smoking patterns cannot yet be seen in Africa. In South Africa, however, men and women in the coloured (mixed black and white ancestry) population have smoked for many decades. We assess mortality from smoking in the coloured, white, and black (African) population groups.

Methods: In this case-control study, 481 640 South African notifications of death at ages 35—74 years between 1999 and 2007 yielded information about age, sex, population group, education, smoking 5 years ago (yes or no), and underlying disease. Cases were deaths from diseases expected to be affected by smoking; controls were deaths from selected other diseases, excluding only HIV, cirrhosis, unknown causes, external causes, and mental disorders. Disease-specific case-control comparisons yielded smoking-associated relative risks (RRs; diluted by combining some ex-smokers with the never-smokers). These RRs, when combined with national mortality rates, yielded smoking-attributed mortality rates. Summation yielded RRs and smoking-attributed numbers for overall mortality.

Findings: In the coloured population, smoking prevalence was high in both sexes and smokers had about 50% higher overall mortality than did otherwise similar non-smokers or ex-smokers (men, RR 1·55, 95% CI 1·43—1·67; women, 1·49, 1·38—1·60). RRs were similar in the white population (men, 1·37, 1·29—1·46; women, 1·51, 1·40—1·62), but lower among Africans (men, 1·17, 1·15—1·19; women, 1·16, 1·13—1·20). If these associations are largely causal, smoking-attributed proportions for overall male deaths at ages 35—74 years were 27% (5608/20 767) in the coloured, 14% (3913/28 951) in the white, and 8% (20 398/264 011) in the African population. For female deaths, these proportions were 17% (2728/15 593) in the coloured, 12% (2084/17 899) in the white, and 2% (4038/205 623) in the African population. Because national mortality rates were also substantially higher in the coloured than in the white population, the hazards from smoking in the coloured population were more than double those in the white population.

Interpretation: The highest smoking-attributed mortality rates were in the coloured population and the lowest were in Africans. The substantial hazards already seen among coloured South Africans suggest growing hazards in all populations in Africa where young adults now smoke.