Surprisingly different frequencies and patterns of K-ras mutations are observed in human adenocarcinomas of the pancreas, colorectum and lung. Their respective relationships with smoking are apparently paradoxical. We evaluated all the available types of clinical and epidemiological studies on the relationship between tobacco smoking and the occurrence of K-ras mutations in human adenocarcinomas of the pancreas, colorectum and lung. We identified 8, 7 and 12 studies that analyzed the relationship between K-ras mutations and tobacco smoking in human neoplasms of the pancreas, colorectum and lung, respectively. A meta-analysis was undertaken for each site separately. In pancreatic adenocarcinomas lifetime history of tobacco consumption was not significantly associated with the frequency of K-ras mutations (OR=1.26; 95% CI=0.82-1.94). Similarly, no association was observed between smoking and K-ras mutations in colorectal adenocarcinomas (OR=0.94; CI=0.79-1.12), neither when colorectal adenomas and adenocarcinomas were jointly analyzed (OR=0.96; 95% CI=0.83-1.13). In lung adenocarcinoma, where only 15-25% of cases harbor a K-ras mutation, tumors from smokers were more likely to have K-ras mutations than tumors from non-smokers (OR=3.67; 95% CI=2.47-5.45). Furthermore, in lung adenocarcinomas K-ras mutations have a pattern different from that in pancreatic and colorectal adenocarcinomas. Results support the hypothesis that smoking influences the risk of pancreatic cancer - and possibly colorectal cancer - through events other than K-ras mutations. In adenocarcinoma of the lung, smoking may play a role in the occurrence of K-ras mutations. If the influence of tobacco products in the induction, acquisition and persistence of K-ras mutations had some tissue specificity, or was dependent on different factors in different organs, the corresponding mechanisms would deserve detailed research.