The main objective of the study was to compare the sociodemographic features, psychiatric history, and substance use patterns, of CSs and FSs at the time of initial diagnosis of a first lung or H&N cancer. The main results of the multivariable regression models were that the CS status was associated with younger age and ever use of hand-roll cigarettes in subjects with a first lung cancer, and increased AUDIT score in subjects with a first H&N cancer. Overall, we found a prevalence of 43.9% CSs, i.e., smokers over the preceding month. This figure is relatively consistent with the estimation provided by a recent literature review, which estimated approximately 50% the rate of CSs over the year preceding the diagnosis of a lung or H&N cancer [10]. Though most of the studies have used a one-year period prior to cancer to define the CS status [10], we chose to follow the recommendations of the National Cancer Institute that deem a one-month period to be more precise [20].

Regarding the sociodemographic features of CSs, we have found in bivariable comparisons that being CS was significantly associated with younger age, being inactive, and being unmarried. Previous studies that explored the smoking status among subjects with a first lung or H&N cancer found relatively similar results. For example, Schnoll et al. found that only a single marital status was significantly associated with increased CS status [21]. In subjects with H&N cancer only, Duffy et al. found that smoking at cancer diagnosis was associated with younger age, single status, and lower education level [22]. Another study by the same team of authors also found that active smoking at the time of diagnosis of an H&N cancer was associated with a single status, but not with age or education level [23].

Though younger age was not consistently found as a substantial contributor of the CS status in previous studies among subjects with lung or H&N cancer, it is a well-known factor in the general population, both in the US and in Europe [24, 25]. In this respective, our result could thus be reflective of this general finding. Moreover, the CS status was significantly more important among subjects with a H&N cancer, compared to those with a lung cancer.

Another important result was that the CS rate was much higher in the subjects with a first H&N cancer than in those with a first lung cancer. To our knowledge, our study was the first to directly compare these two types of populations. The risk of experiencing a lung cancer, especially adenocarcinoma, is still significantly increased compared with never smokers more than 30 years after smoking cessation [26]. In contrast, quitting tobacco smoking fosters a more rapid decrease in the relative risk of H&N cancer [27]. The gap observed in the CS rates between lung and H&N cancer populations may thus be attributable to this difference in longitudinal risk reduction among FSs.

More importantly, it was recently noted that almost no previous studies have addressed the relative rates of the types of tobacco ever used among CSs and FSs with a lung or H&N cancer and that it was a significant issue to explore [10]. In this regard, our study is also the first to provide important findings on this subject. We have notably found that reporting ever smoking of roll-up cigarettes was highly associated with being a CS at the time of cancer diagnosis of a first lung cancer. In the French population, it has been previously found that 24.3% of smokers were used handrolling tobacco, while 7.5% of smokers use only this type of tobacco [28]. These figures are difficult to compare to other countries, as the prevalence of roll-you-own smokers is very variable depending on countries and their specific regulations on tobacco [29]. Regardless, in France, compared to smokers of manufactured cigarettes, handrolling smokers reported much lower average personal incomes and higher rates of unemployment [28]. This finding is consistent with other data out of France [30]. In the general population, it was previously shown that quitting smoking is inversely associated with impaired social conditions (10). Our findings in patients with cancer could thus be explained by unexplored social factors, even if our analyses were adjusted for the level of education.

Moreover, as early alterations in daily life activities and the presence of metastases at the time of cancer diagnosis were significantly associated with the FS status in bivariable comparisons, it could be suggested that the FS status is actually partially explained by symptom burden, in particular the fact that some people have become too sick to smoke before cancer was diagnosed. However, these two variables were involved in the step-by-step regression model, but they were not retained as relevant explanatory parameters by the modeling.

Finally, we identified that concurrent alcohol misuse, reflected by the AUDIT score, was significantly associated with the CS status only in patients with a H&N cancer (see Table 2), which could be underlain by the fact that alcohol and tobacco are combined risk factors for this type of cancer.

Several limitations should be acknowledged with regard to the present study. First, the study was multicenter, but the recruitment was regional. Some findings could thus be skewed by local features. Moreover, patients with esophageal cancers were not included in the study, and thus, the findings cannot be applied to all types of H&N cancers. Finally, we did not assess the use of non-smoking types of tobacco, notably chewing or snuff tobacco. However, this type of tobacco use is very rare in Europe [25]. The main objective of our study was to assess the lifelong types of tobacco used by the patients. This retrospective assessment could be found to be rather imprecise. However, a detailed and quantitative assessment of the specific periods of time during which each type of tobacco was used over the entire lifetime period would have been very difficult to carry out and would have suffered from memory bias and imprecision.