Data from two clinical studies involving smokers and snuff users were analysed to address the estimation of nicotine intake using urinary and salivary nicotine metabolites. Comprehensive regression modelling is performed to determine which combinations of urinary nicotine metabolites provide better estimation of nicotine intake in these subjects than the predominant practice of basing nicotine intake on urinary cotinine analysis alone. Within-subject and between-subject variability is examined with regard to reliability of measurement and replicate sampling. Salivary cotinine models are compared to urinary metabolite models. Results suggest that estimation of nicotine intake is greatly improved by measuring urinary cotinine and additional metabolites (trans-3´-hydroxycotinine, and glucuronide conjugates) rather than measuring only cotinine. Analyses also indicate that replicate sampling on subjects greatly improves the reliability of the measurement. Based on these data, a model to predict nicotine equivalents based solely on saliva cotinine was severely inferior to any of the urinary models, including that of urinary cotinine alone.