The purpose of this work is to model the relationship between smoking-related variables and the risk of lung cancer by using parametric and non-parametric models. A hospital-based case-control study was conducted to ascertain the influence of smoking on risk of lung cancer. We used parametric logistic regression with a series of categorized independent variables and non-parametric logistic regression models. Such models allow for variables to be treated as continuous, since they avoid arbitrariness in the selection of cut-offs and furnish information on the dose-risk relationship. The results point to the possible existence of a saturation effect for a lifetime tobacco consumption of around 25 000-30 000 packets. Duration of habit and years of abstinence show a linear relationship marked by opposite, though similar, slopes, which would seem to indicate that for every year of smoking, risk rises by an amount (8.00%, 95% confidence interval (CI) 5.94-10.06) equal to the decline in risk for every year of abstinence (6.98%, 95% CI 2.53-11.84). Lastly, a lower age of smoking initiation appears to have an influence, although non-significant, on the appearance of the disease. The risk of lung cancer due to duration of the habit would seem to be proportional to years of abstinence, and there could be a saturation effect with respect to lifetime tobacco consumption.