ABSTRACT Background Smoking is undoubtedly a major risk factor for morbidity, disability and premature death. Its use results in grave health and economic losses not only to the individual but also to the population and the world at large. Many surveys have been done in South Africa to estimate the prevalence of smoking. It is therefore imperative and expedient to have an overall impression of the prevalence rates over time. And also it is important to assess how subgroups affect the prevalence and trends in the national population. This will be of help in determining which subgroups have achieved reduction in smoking prevalence and which have not; evaluating the tobacco control policies in the country; and in designing specific interventions. This research was undertaken to determine the trends and patterns of smoking in the South African adult population Objectives The objectives for this study were: Regarding the South African adult population during 1995 – 1998, to: 1. Compute the prevalence of smoking and assess the trends of smoking prevalence. 2. Assess the patterns and trends of smoking prevalence in subgroups by sex, age, marital status, race, locality (urban or rural), education and province. 3. Identify factors in the population that may account for patterns and trends in smoking prevalence over time 4. Make recommendations regarding the public health implications of the findingsMethods This was an analytical study involving secondary analysis of existing datasets from four South African representative national surveys. From 11 surveys, which measured smoking in the South African population, four surveys were selected using some inclusion and exclusion criteria. The population of interest was the South Africa adult population (18 – 49), so variables of interest (outcome variable was current smokers) for this group were extracted. Prevalence (frequency) rates estimation of smoking in the national population and in subgroups were then estimated. Unadjusted odds ratios and adjusted odds ratios were computed by bi-variate cross tabulation and multivariate logistic regression respectively. Time-trend analyses (Maentel Haenszel chi-squared test) were computed by logistic regression for trend in proportions Results From 1995 to 1997 about 1/3 of the adult South African population were smokers, but that dropped significantly to about ¼ in 1998. For the period however, there was no significant trend. The prevalence of smoking varied with, and was largely depended on population subgroup; while it was as high as 63.9% among Coloured males, 62.3% among Coloured females, 53.7 % among all males, 52.7% among rural males, it was as low as 11.4% among all females, 6.8% among rural females, 10.83% among Indian females and 5.06% among Black females. The only significant trends was an increasing smoking prevalence among Blacks, Coloured men, people with tertiary education, Free State and Gauteng provinces, age group 35 – 44; urban men and a decreasing smoking prevalence in all women, urban women and black women, age group 18 – 24 and the Eastern Cape, Kwazulu-Natal, Northen Cape and Mpumalanga provinces. Sex, race, age, and education were the major risk factors for smoking in the South African adult population. Locality (rural/urban) though had different smoking rate was not a risk factor for smoking. Marital status was neither a determinant nor risk factor for smoking. Discussion and Conclusion The prevalence of smoking in the South African adult population is very high and did not achieve any significant trend between 1995 and 1998. However the significant drop from 1997 to 1998 probably means that smoking prevalence in the national population may have started declining; therefore, more monitoring is needed to ascertain this. This high prevalence of smoking in the South African population, which may have been for years, may predict a high burden of chronic smoking-related diseases in the near future. The patterns of smoking analyses reveal that smoking in the South African adult population is determined by a complex interplay of different factors.