This is the first study on smoking behavior in a Spanish adult population using both a questionnaire and a biomarker of tobacco exposure. Salivary cotinine concentration was associated with the number of cigarettes smoked in the last 24 hours. This relation was better explained with a quadratic function and in separate strata for men and women. The greater the number of cigarettes smoked, the greater the cotinine concentration in a linear scale up to 20 cigarettes per day, after which the association plateaus. A similar relation between salivary [6–9, 20] or serum  cotinine concentrations and number of cigarettes smoked has been observed in other studies. Abrams et al. dichotomized tobacco consumption, and found that for smokers of less than 25 cigarettes per day, salivary cotinine concentration was highly correlated with tobacco consumption, while among heavier smokers the relation was not evident .
Other studies have found that cotinine concentration plateaus at different numbers of cigarettes: 25 cigarettes per day , 5 cigarettes per day , and 4 cigarettes per day in adolescent smokers . The evidence suggests that cotinine concentration rises in a non-linear fashion with increasing number of cigarettes smoked, but the point where concentrations level off may vary across different populations. This finding suggests a difference in how people regulate their intake of nicotine to reach the desired dose , even for adolescents, who may be more susceptible to nicotine than adults and require only 4-5 cigarettes per day to satisfy their nicotine cravings .
We observed that cotinine concentrations differed by sex, regardless of the number of cigarettes smoked. Some studies reported similar findings of higher cotinine concentrations in men than in women [7, 23, 26, 27], but other studies did not find differences by sex [6, 28]. Association between urinary cotinine and cigarettes smoked according to sex was found in a study in the USA: urinary cotinine concentrations in men increased up to 34 cigarettes per day and then declined, while no flattening was observed in women . The differences we observed by sex could reflect not only differences in tobacco consumption by sex, i.e., men usually smoke more cigarettes than women, but also a difference in the metabolism of nicotine between men and women [30, 31].
Our data showed a higher cotinine concentration in smokers of black tobacco that did not persist with control for the number of cigarettes smoked. Whereas uptake of carcinogens is higher among black tobacco smokers [32–34], differences in nicotine uptake by type of tobacco smoked have not been reported . In our study, use of filter, frequency, and depth of inhalation were not related to cotinine concentrations. An explanation of these results could be that smokers tend to maintain the same intake level of nicotine by drawing in more smoke per cigarette when they try to smoke fewer cigarettes. Benowitz et al. reported that among people who reduced from 37 to 5 cigarettes per day on average, the intake of tobacco toxins per cigarette increased roughly threefold . This could also explain how cotinine concentrations level-off in smokers of more than 20 cigarettes per day, when a certain intake of nicotine is achieved [8, 9].
The role of age in cotinine concentrations is still not clear, since our results, as well as previous studies [7, 37], indicated no association between cotinine concentrations and age, while others have found a significant association [6, 18, 29, 38]. Some studies have modeled the relation between cotinine concentrations and cigarette consumption by taking into account several of these variables. The fit of the multivariate model improved once age, BMI, educational level, and a quadratic term for cotinine were included.
Study limitations and strengths
To our knowledge, this is one of the few studies in which information about tobacco exposure was obtained in a representative random sample of the general population with a simultaneous use of a questionnaire and a biological marker. In the USA, the National Health and Nutrition Examination Survey (NHANES) provides population-based national estimates of smoking prevalence using both standard questionnaire and serum cotinine concentration . Most previous studies were based on selective samples from existing observational studies [6, 8, 18, 25, 37] or smoking cessation trials [21, 22, 29]. Other factors affecting validity also need to be considered. Although the use of self-reported data from questionnaires could be a source of bias, self-reports on smoking are accurate and have acceptable validity [40, 41]. Cotinine is a specific biomarker of tobacco exposure [1, 2], and the laboratory methods are highly sensitive .
Some potential limitations deserve consideration. We found that the model fit could be affected by the measurement scale of the number of cigarettes. Smokers tend to round up the number of cigarettes smoked, particularly heavy smokers , and hence some information bias due to digit preference cannot be disregarded. While some loss of representativeness due to non-response might also be possible, the sample did not differ by sex, age, and district of residence from the Barcelona population. Moreover, the prevalence of smokers in the sample (28.6% of men and 18.2% of women) was similar to that derived from the 2006 Health Interview Survey of Barcelona (27.3% of men and 20.6% of women) . The participation rate was almost complete because the study design allowed replacement of non-respondents by subjects in the same strata of sex, age, and district of residence.