Breath analysis has a powerful potential for disease diagnostics and metabolic status monitoring. A sol-gel SnO2-based micromachined sensor array was developed and tested for potential application in breath analysis. A suitable breath sampling system was used to sample the alveolar air volume from the air volume of one expiration. Breath tests on alveolar air sampled by some volunteers, i.e. smokers and non smokers individuals, were carried out. Principal Component Analysis applied to gas sensor responses showed good properties of discrimination between smokers and non smokers individuals.