In this study, we analysed the associations of CPD, urinary cotinine, and SES in an elderly German population of current smokers. Most men and women reported to smoke one pack of cigarettes per day (20 CPD). While the reported median smoking intensity did not vary by SES, CPD reports were more variable in men and women with high SES. The association of CPD and cotinine levelled off at about 20 CPD, and the logarithm of CPD was the best functional form to predict cotinine. The correlation between CPD and cotinine increased with high SES, indicating a higher precision of self-reports. In regression models, CPD and creatinine were the main predictors of cotinine. In women, cotinine slightly decreased with higher SES, as confirmed in the multiple regression. All findings for cotinine were similar for 3OH-cotinine.
For our study we utilised data of a population-based cohort with largely available information on smoking and occupational and educational SES, and we were able to apply highly sensitive urinary markers for smoking.
Although self-reported average daily cigarette consumption is the common measure of smoking intensity in health studies, a more precise measure of current smoking exposure would have included the exact number of cigarettes in the days before examination, and/or the time of the last cigarette. We used the last meal and the last consumption of coffee or tea before sampling as a proxy for the latter, but did not find associations with cotinine. However, average and recent CPD were found to be highly correlated [6]. We also did not have information on individual puffing behaviour, which was found to be a mediator for the association of CPD and salivary cotinine [18], and thus could have contributed to a better predictive model for urinary cotinine. We also lacked data on the magnitude of nicotine dependence beyond CPD. In particular, the time to the first cigarette after awakening was found to be an additional predictor for cotinine [19, 20]. We did not consider cigarette brands, which may vary by nicotine content. However, it is not clear to which extent different nicotine contents are reflected in cotinine concentrations, as they might be compensated by inhalation behaviour [21].
We observed a well-known, wide variation of cotinine concentrations at each level of CPD which partially can be attributed to genetic differences in nicotine metabolism [4, 22]. The nicotine metabolism rate is commonly displayed by the ratio of 3OH-cotinine/cotinine, additionally by the ratio of cotinine glucuronide/cotinine [23]. Our additional adjustment for the ratio of 3OH-cotinine/cotinine increased the model fit. However, urinary nicotine metabolite ratios varied in previous studies, probably dependent on the type of measurement [24]. Differences in nicotine metabolism also likely account for the reduced cotinine values in women [4]. The sum of cotinine and 3OH-cotinine might more comprehensively reflect nicotine uptake than single metabolites [25], but we found a slightly higher correlation with CPD for cotinine. As mentioned, variation of nicotine metabolites might be reduced by additional information on puffing behaviour, nicotine dependence, and more detailed recent cigarette exposure.
Further, retrospective self-reports of CPD tend to show a digit preference, i.e. reports of multiples of 10 or 5, which increase at higher CPD [26] and were also apparent in our data. More frequent reports of one daily pack and decreased correlation of CPD and cotinine indicated lower precision of CPD reports for lower SES [5].
The flattening we observed in the association between CPD and cotinine confirms results of several studies, regardless of the body fluid (plasma, saliva, or urine) that was used to determine cotinine [6,7,8, 18, 27]. Investigating non-linearity, we selected the logarithm of CPD to appropriately depict this association, while most other studies remained with visual evidence of the plateau effect. Some studies found improved model fits with an additional quadratic term of CPD [5, 27, 28], or setting a cut-off at 20 CPD in regression analyses [18, 27].
In our multiple regression model, urinary creatinine was confirmed to be an important predictor for cotinine [28]. In contrast to our results for CPD, creatinine, and to results of some other studies, we found only weak effects for BMI/weight [29], and also age, the latter a possible consequence of the narrow age range in our study population. A gender effect was not investigated here, as all models were stratified by gender.
We found a slightly negative association of occupational and educational SES and cotinine in women, but no obvious association in men. A negative association of SES and cotinine was observed in other studies, though varying by SES indicator: Higher cotinine concentrations were found for lower education, but not income or occupation in the FINRISK study [10], and deprivation, but not lower occupation in an English survey [11]. A Czech study on CPD and thiocyanate, another biomarker for smoking, showed higher levels for low education [30]. Results in these studies were adjusted for CPD, however, as linear variable. Further, an income/wealth-based SES index was negatively associated with cotinine when adjusting for nicotine dependence (including CPD), but only in the subgroup of unemployed subjects [31].
Different causes for the widely observed flattening of cotinine values at higher CPD may be considered:
First, there is possibly a biological maximum for the uptake and metabolism of nicotine. A biological saturation was also discussed to explain flattening cancer risks in heavy smokers [9, 32]. However, increasing lung cancer risks were also reported at higher cotinine concentrations [33, 34], and specific associations of other (carcinogenic) tobacco smoke constituents with CPD were observed [8]. Thus, it seems problematic to infer a biological saturation for cotinine from the association of CPD and cancer.
Another possible cause is compensatory smoking behaviour, i.e. less inhalation in heavy smokers. Similar associations of CPD and cotinine including a flattening were shown for smokers of cigarettes with regular compared to reduced nicotine content [27]. Most studies on compensation investigated effects of nicotine reduced cigarettes, but not separately for light/heavy smokers, and with different results [35,36,37]. In a comparison of large surveys with over two decades in between, cotinine concentrations remained constant while CPD substantially decreased, which was attributed to compensatory smoking [38].
Finally, information bias of self-reported CPD could lead to flattening CPD-cotinine-curves, assuming an underreporting of CPD in the middle CPD section with corresponding high cotinine values. An independent effect of SES on cotinine would have indicated information bias, whereas in particular biological saturation should not differ by SES. Stronger compensation could also be associated with lower SES due to higher nicotine dependence [10, 39] or financial reasons [30, 40]. I.e., being financially restricted could increase pressure to satisfy nicotine needs with a relatively lower number of cigarettes per day, with 20 CPD being more frequently reported for low SES in our study. However, we found only a weak association between SES and cotinine either with or without adjustment for CPD.