Factors associated with cessation of smoking among Swiss adults between 1991 and 2011 : results from the SAPALDIA cohort

Ayala-Bernal Dianaab, Probst-Hensch Nicole M.ab, Rochat Thierryc, Bettschart Robertd, Brändli Ottoe, Bridevaux Pierre-Oliviercf, Burdet Lucg, Frey Martinh, Gerbase Margaret W.c, Pons Marcoi, Rothe Thomasj, Stolz Daianak, Tschopp Jean-Marief, Turk Alexandere, Künzli Ninoab, Schindler Christianab a Swiss Tropical and Public Health Institute, Basel Switzerland b University of Basel, Switzerland c Hôpitaux Universitaires de Genève, Service de pneumologie, Geneva, Switzerland d Lungenpraxis Aarau, Hirslanden Klinik, Aarau, Switzerland e Zürcher Höhenklinik, Wald, Switzerland f Service de Pneumologie, Hôpital du Valais, Switzerland g Hôpital intercantonal de la Broye, Payerne, Switzerland h Klinik Barmelweid, Abteilung Pneumologie, Aarau, Switzerland i Ospedale Regionale di Lugano Sede Civico, Lugano, Switzerland j Spital Davos, Davos Platz, Switzerland k Clinic of Pulmonary Medicine and Respiratory Cell Research, Basel, Switzerland


Introduction
The fight against the tobacco epidemic is a global health priority.During the 20th century, tobacco smoking killed about 100 million people and this number is likely to reach 1 billion in the 21st century [1].Despite numerous international tobacco control efforts, smoking is still a top ranking cause of disease and premature death [2].Six million people per year die as a consequence of tobacco use, more than 600 000 of them as a result of exposure to secondhand smoke [3].
In Switzerland, more than 9000 deaths were attributable to smoking in 2007 according to the Federal Office of Statistics.A quarter of these deaths involved persons under the age of 65 years and a higher proportion were men.The three leading causes of smoking-attributable death were cancer (41%), cardiovascular disease (41%) and respiratory diseases (18%) [4].Additionally, environmental tobacco smoke exposure in public places in Switzerland was estimated to cause 32 000 additional hospital days and 3000 YLL (years of life lost) owing to lung cancer and ischaemic heart disease per year.This corresponds to healthcare costs of 330 million Swiss francs [5].
Recent data from the Swiss national health survey point to a decreasing trend in smoking prevalence in both sexes and most age groups [6].According to national and international evidence, the most effective measures for tobacco control at the population level are structural prevention measures, such as cigarette pricing and legislation against smoking in public and work places or against advertisement [1,7].Switzerland was late in responding to international calls for tobacco control and is one of the few countries that have not yet ratified the World Health Organization (WHO) tobacco convention.In 2016, the health committee of the Swiss parliament voted against the prohibition of tobacco advertisement in public places and movie theatres.All European Union member states have in the meantime implemented such laws.Switzerland has, however, increased tobacco taxes and restricted smoking in public and work places in more recent years.A nationwide smoking ban in public places was enforced in 2010.However, less is known about personal factors influencing the likelihood of smoking cessation.Educational level has been identified as an important predictor of quitting among smokers in Switzerland and the European Union [6,8].However, health variables have not been well studied as predictors of cessation among Swiss smokers [6].Research on these factors can help to tailor prevention strategies for further reduction of the burden of tobacco use.The aims of this study were to assess changes in smoking prevalence and predictors of smoking cessation between 1991 and 2010/11 in the population-based cohort of SAPALDIA (which covers culturally and geographically diverse regions of Switzerland) with a specific focus on the role of sociodemographic, lifestyle and health-related factors.

Materials and methods
This study used data from SAPALDIA, a population-based multicentre cohort study initiated in 1991 with the objective of studying effects of air pollution on respiratory health in adults.The SAPALDIA study was conducted in eight different areas of Switzerland (Aarau, Basel, Davos, Geneva, Lugano, Montana, Payerne and Wald).Members of the SAPALDIA team are listed in appendix 1. Participants aged 18-60 years at the onset were recruited through random samples drawn from the respective inhabitant registries.The baseline survey (S1) included 9651 participants.Later, two follow-up assessments were conducted, one in 2002 (S2; 8047 participants) and the second one in 2010/11 (S3; 6088 participants).
At each assessment of the SAPALDIA cohort, detailed information about general health, life-style and living conditions was collected by means of an extensive computerbased interview at the study centre.In the second and third survey, self-administered questionnaires and telephone interviews were additionally used.Pulmonary function of study participants was measured in all three surveys.The lung function parameters assessed were FVC (forced vital capacity), FEV1 (forced expiratory volume in the first second) and various forced expiratory flow measures.The FEV1/FVC ratio was calculated for every subject and a low ratio (defined as <0.7) was considered an indicator of obstruction to air flow in accordance with the GOLD report definition of Chronic Obstructive Pulmonary Disease (COPD) [9].The SAPALDIA methodology has been described in detail [10,11].All participants gave written informed consent and ethical approval was granted by the Swiss Academy of Medical Sciences and the respective cantonal ethics committees.

Study population
For the purpose of this study, two samples were chosen.The first sample consisted of all smokers at SAPALDIA1 with smoking status known at SAPALDIA2 and sufficient covariate information at SAPALDIA1 and the second sample contained all smokers at SAPALDIA2 with smoking status known at SAPALDIA3 plus sufficient covariate information at SAPALDIA2 (see section on statistical analysis).

Outcome variables
The outcome variables were "smoking cessation between S1 and S2" in the sample of smokers in SAPALDIA1 and "smoking cessation between S2 and S3" in the sample of smokers in SAPALDIA2.Subjects who reported being smokers at the beginning of the respective period and nonsmokers at the end were assigned the value 1 in these variables and will be referred to as quitters.
Being a smoker was defined by a positive answer to both of the following questions: "Have you ever smoked for as long as a year?" "Do you now smoke, as of one month ago?"

Statistical analysis
The proportions of smokers and nonsmokers were assessed for each of the three surveys according to age, sex, study area, educational level, nationality and marital status.Predictors of smoking cessation were studied by using a multivariable logistic regression model with the outcome "having quitted between respective surveys".In order to minimise the risk of confounding of main results and to be open to new findings, the selection of independent variables was broad and limited only by data availability.All independent variables were defined on the basis of their values at the beginning of the respective follow-up period.overweight (defined by a body mass index >25 kg/m 2 ), level of environmental tobacco smoke exposure (none; less than 3 hours a day; at least 3 hours a day), asthma (defined by the question "Did you ever have asthma and was this asthma confirmed by a doctor?"),chronic cough (defined by the question "Do you usually cough during the day or at night, on most days for as much as 3 months each year, and for at least two years?"),chronic phlegm (defined by the question "Do you usually bring up any phlegm from your chest during the day or at night, on most days for as much as 3 months each year, and for at least two years?"), and level of chronic obstructive pulmonary disease (FEV1/ FVC ≥0.7, 0.5 ≤FEV1/FVC <0.7, FEV1/FVC <0.5).This model was applied to smoking cessation between S1 and S2 among smokers at S1 (i.e., with predictor variables assessed at S1) and to smoking cessation between S2 and S3 among smokers at S2 (i.e., with the same predictor variables updated at S2). Estimates of both models were then combined using fixed effects meta-analysis.In a second step, the model for smoking cessation between S2 and S3 was extended by the following additional predictor variables only assessed at S2: alcohol consumption ("less than once a week"; "less than once a day"; "once or two times a day"; "three times or more per day"); physical activity ("sufficiently active" = at least 150 min/week of moderate or vigorous activity vs "insufficiently active" = less than 150 min/week of moderate or vigorous activity); an indicator variable for "new dog ownership" (i.e., reported dog ownership in S2 but not in S1); separate indicator variables for self-reported doctor's diagnoses of diabetes, cancer and depression; an indicator variable for cardiovascular disease defined as presence of heart disease, hypertension or stroke, with heart disease having been defined based on prescribed medications (beta-blockers, angiotensin converting-enzyme [ACE] inhibitors, calcium channel blockers or diuretics).In a third model, we additionally included the eight indices of the SF36 questionnaire on self-perceived, health-related quality of life, representing the domains vitality, physical functioning, bodily pain, general health perceptions, physical role functioning, emotional role functioning, social role functioning and mental health [12].As these variables may be on the causal pathway between chronic diseases and the decision to quit smoking, they were not included from the beginning.As FEV1/ FVC is the lung function measure used to define chronic obstructive pulmonary disease (COPD), also known as "smoker's disease", we explored the functional relationship between FEV1/FVC at S1 and the likelihood of quitting smoking between S1 and S2, using a natural cubic spline function.Effect modification by sex was assessed using interaction terms.Gender-specific analyses were also conducted and their results are presented in supplementary tables S6 and S7 (appendix 2).As the interaction between sex and age had a p-value <0.1, effects of age were estimated separately for men and women.To address potential bias due to loss to follow-up, analyses were also performed using inverse probability weighting.Probability weights for the analysis of smoking cessation between two consecutive surveys were derived from logistic regression models of participation in the second of the two surveys as a function of variables from the first of the two surveys which were informative about further participation.The level of statis-tical significance was defined at 0.05 for main effects and at 0.1 for interaction effects.All analyses were conducted in Stata 13.1 (Stata Corp., College Station, USA).

Characteristics of smokers and non-smokers over the time of follow-up
The sociodemographic characteristics of smokers and nonsmokers are shown in table 1.Similar patterns were observed in all three surveys.Current smoking was more prevalent in men, persons aged between 30 and 50 years, persons with low education or of non-Swiss nationality, and divorced persons.Moreover, smoking was most prevalent in Payerne, Basel, Lugano and Geneva.
Overall, the proportion of current smokers declined among most age groups and areas between S1 and S3.Moreover, the prevalence of current smoking decreased for both men (from 38% in S1 to 20% in S3) and women (from 29% in S1 to 17% in S3).However, the prevalence of current smoking was higher in males than in females in all three surveys.The proportion of current smokers also decreased between S1 and S3 across all educational levels (most strongly among persons with medium education, with 33% in S1 and 19% in S3), nationalities (most strongly among Swiss persons, with 33% in S1 and 18% in S3) and categories of civil status (most strongly among married persons with 31% in S1 and 15% in S3).Smoking prevalence remained highest in the lowest educational group, in divorced participants, and in non-Swiss nationals.

Predictors of smoking cessation between S1 and S3
Table 2 shows the odds ratios of smoking cessation between S1 and S2 among smokers at S1 for predictor variables from S1 (first column) and of smoking cessation between S2 and S3 among smokers at S2 for the same variables re-assessed at S2 (second column).The meta-analytic summary estimates of the period-specific results are shown in the third column of the A further analysis studying the relationship between smoking cessation and the Tiffeneau ratio FEV 1 /FVC was performed (fig.1).The probability of quitting showed a Ushaped dependency on the FEV1/FVC ratio with highest probabilities of quitting at the two ends of the spectrum.However, the most important finding was that the probabil-

Discussion
The findings of our study are consistent with results from the Swiss Health Survey, which documented a decrease in smoking prevalence from the early nineties to about 2010 in Switzerland [6,13].However, in our study the decrease was faster, which is partly explained by the stronger loss to follow-up among smokers in our cohort, but possibly also by the fact that our cohort aged by 20   1 Results for the periods 1991 (SAPALDIA1) -2002 (SAPALDIA2) and 2002 -2010 / 11 (SAPALDIA3) are from multivariable logistic regression models (among persons who were smokers at the start of the period) with outcome = "having quit smoking by the end of the period" and covariates including study area and all variables listed in the table (as assessed at the beginning of the period).Results for the entire period 1991 -2010/11 were obtained as meta-analytic summary estimates of the period-specific results. 2 Low = ≤9 years of school; medium = 12 years of school; high = college or university 3 Body mass index >25 kg/m 2 4 Mean duration of daily exposure to environmental tobacco smoke 5  Regular occurrence of respective symptom for ≥3 months per year and since at least 2 years 6 Confirmed by a doctor 7  of former smokers and a decreasing prevalence of current smoking.This trend was previously reported in another Swiss study [6], as well as in other countries such as Spain [14], the UK and Australia [15].Overall, men had a higher proportion of current smokers over time than women, but the sex difference decreased over time.Our findings of a decreasing smoking trend in both men and women with a slower decrease rate among women are in agreement with other studies in high-income countries [14][15][16][17], where this is described as a consequence of the later adoption of smoking as a widespread habit by women.Reasons for this time lag include multiple sociocultural factors such as social disapproval of female smoking in the early twen- ) associated with SF36-scores of "social functioning" and "energy and vitality" are adjusted for age, sex, study area, civil status, educational level, nationality, BMI, exposure to environmental tobacco smoke, number of cigarettes smoked per day, alcohol consumption, physical activity, new dog ownership, presence of chronic cough, chronic phlegm and asthma, depression, diabetes, cardiovascular disease, cancer and pulmonary function.There was no evidence of any association between smoking cessation and the other SF36-scores.
tieth century, gender role norms, less employment among females and a higher religious commitment in women [18].Possible causes of this closing gap between females and males, recently described in the WHO bulletin [19], include increasing societal tolerance toward smoking women [15], growing economic resources among women and the tobacco industry's targeting of women by featuring smoking as a symbol of independence and social desirability [20,21].

Sociodemographic and lifestyle predictors of quitting
Our findings indicate that higher educational level, being divorced or widowed and age in men are significant predictors of smoking cessation.These results are consistent with those of previous research in Switzerland [6,22] Europe [8, 23,24], and the USA [25], where higher odds of quitting were linked to male gender, aging, higher education, marital status and higher socioeconomic status.That overweight smokers were more likely to quit was unexpected.A similar finding was made in two European cohort studies, of which the first was conducted in Germany, where the cessation rate increased significantly with increasing BMI [26], and the second in Ireland where having a BMI ≥25 kg/m 2 was positively associated with quitting [27].In contrast, a study conducted in the USA [28] reported a lower cessation rate among overweight persons and explained this by concerns about gaining weight expressed by overweight smokers, particularly women.New dog ownership as a positive predictor of quitting was another unanticipated finding.Although there is some evidence that having a pet can motivate healthy behaviour changes [29,30], this finding would need to be confirmed by other studies.
In contrast, the likelihood of quitting was lower in women and decreased with increasing number of cigarettes smoked per day, an indicator of dependency.Multiple studies have found that heavier smokers are more likely to continue smoking in the long term [8,11,24,31].This negative relation between nicotine dependence and smoking cessation may be used as a basis for a personalised approach to nicotine replacement therapy (NRT).For example, a previous study in twins showed that 40 to 60% of the individual differences in the ability to quit smoking and, to some extent also the propensity for nicotine dependence, are genetically determined [32].As a consequence, the effectiveness of NRT could be improved by differentiating the dose of nicotine replacement according to the level of dependency and the genotype of the smoker [33].
We additionally performed an analysis looking at the probability of quitting before S1.Interestingly, patients with asthma and women were more likely to quit before SAPALDIA 1, suggesting that persons from these two groups who were still smoking at the baseline survey had increased nicotine dependency.This might provide another explanation for the negative association of these two characteristics with quitting.

Health-related predictors of smoking cessation
Among the health-related variables included in our study, cardiovascular disease and severe airway obstruction showed a positive association with quitting, whereas persons with asthma, chronic phlegm or depression were less likely to quit.A negative association between smoking cessation and chronic sputum production was also found in a recent smoking cessation intervention study conducted among 887 smoking employees in the city of Basel [34].
The coexistence of smoking and depression has been extensively investigated, mainly because depression is the psychiatric disorder most associated with tobacco use [35].Similar to our results, the aforementioned intervention study in Basel showed that smokers with a history of depression or current use of antidepressants were less likely to quit [34].This may be related to common genetic mechanisms behind smoking initiation, nicotine dependence and depressive disorders [36,37].
According to the European Respiratory Society-Task force guidelines, respiratory patients who smoke are a difficult target group for smoking cessation, as they tend to minimise their own perceived risk of disease.Even in advanced stages of COPD, when quality of life is low, a smoking patient could consider cigarettes as an important factor for his/her quality of life [38].Therefore, these patients need strong encouragement and continued counselling support for overcoming their addiction [39].
Our data on the health-related quality of life assessed with the SF-36 questionnaire showed that current smokers with a lower vitality score had a higher probability of quitting, whereas those with a lower social functioning score were less likely to quit.This is consistent with the results of previous studies, where persons who continued smoking had lower social functioning scores than those who quitted smoking [40,41].Moreover, a higher number of cigarettes smoked per day has been significantly associated with a lower vitality score [40,42], possibly reflecting impaired physical health.However, in our study the association was observed after adjusting for chronic diseases and even in a sensitivity analysis where we excluded the participants with chronic diseases (data not shown).

Swiss smoking bans and smoking cessation
The decrease in smoking prevalence between SAPALDIA 2 and SAPALDIA 3 could also be related to the introduction of smoking bans at the cantonal level.However, we could not assess the impact of these smoke-free laws because of the shortness of the follow-up time after the implementation of the bans and the imprecise information on the quitting age of many study subjects.

Strengths and limitations
Our study is based on a large multi-regional population sample, which was followed up over 20 years, and from which an extensive database on sociodemographic, environmental, lifestyle and health characteristics of the study participants could be established.The large number of different predictor variables considered is both a strength and a limitation of the study.It reduces the risk of confounding, but increases alpha error inflation.Loss to follow-up, which is common in long-term cohort studies, is a limitation.To address this concern, we performed inverse probability weighting involving predictor variables of follow-up participation.This led to only minor changes in the observed associations.Although this does not rule out the presence of some remaining bias, it makes major attrition bias less likely.Although the SAPALDIA study includes local populations from urban and rural communities and from the three major language regions of Switzerland, the present results may not be fully generalisable to the adult population of Switzerland, as study areas were not selected randomly.Moreover, as the cohort aged during follow-up, our results cannot be generalised to younger adults.Another limitation is the self-report of smoking status.But self-reported smoking status was shown to have good validity in several population based studies [43,44] and we had found end-expiratory carbon monoxide measurements in SAPALDIA2 to discriminate well between quitters and non-quitters in the first follow-up period, with an area under the receiver-operating-characteristic curve of 0.91.

Conclusion
Our data confirm the decrease in smoking prevalence in Switzerland over the period 1991 to 2011.Our findings regarding individual characteristics associated with smoking cessation are relevant tools for personalised approaches in tobacco control at the policy and at the personal level, as they point to subgroups in need for additional support such as pharmacological treatment of nicotine addiction.Prospective public health policies against tobacco use in Switzerland should be particularly focused to women, younger persons and persons of lower education.Persons with respiratory problems, depression or signs of high nicotine dependence need personalised support.  Reported ownership of a dog in SAPALDIA2 but not in SAPALDIA1 10 Diagnosed by a doctor 11 Hypertension, stroke and/or intake of heart medication 12 Low = score <25; medium = 25 ≤score <75; high = score ≥ 5 * p < 0.05 FVC = forced vital lung capacity, FEV1 = forced expiratory volume in one second * p <0.05 Original article Swiss Med Wkly.2017;147:w14502 Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Figure 1 :Figure 2 :
Figure 1: Estimated probability of quitting smoking between 1991 and 2002 as a function of FEV1/FVC in 1991, with 95% confidence limits*. 1 Defined according to the GOLD report definition*derived from a logistic regression model of smoking cessation between 1991 and 2002 containing a natural cubic spline function of baseline FEV1/FVC and adjusting for age, sex, study area, civil status, educational level, nationality, BMI, parental smoking, exposure to environmental tobacco smoke, number of cigarettes smoked per day and presence of respiratory conditions (chronic cough, chronic phlegm and asthma) at baseline

table .
smoking cessation, with odds ratios above 2.5 in both cases.Moreover, smokers with asthma were significantly less likely to quit than those who did not report having asthma.On the other hand, the gender difference decreased.
16% CI 0.69-1.27),chronicphlegm(OR 0.77, 95% CI 0.51-1.16)or asthma (OR 0.77, 95% CI 0.54-1.08)were also less likely to quit.No significant differences were found according to area or nationality.Some associations got considerably stronger in the second period.Both high and medium as compared to low education became sig-nificant predictors of (OR 1.78, 95% CI 1.12-2.83)were significantly more likely to quit smoking.A history of cancer was also a positive albeit not significant predictor of quitting by S3 (OR 1.54, 95% CI 0.79-2.99).Associations of smoking cessation between S2 and S3 with the variables re-assessed at S2 showed similar odds ratios with and without adjustment for the variables newly assessed in S2, and patterns of statistical significance were identical (supplementary table S2 in appendix 2).

Table 1 :
Characteristics of the cohort at the three surveys of SAPALDIA in 1991, 2002 and 2010/11 by smoking status.
Swiss Medical Weekly • PDF of the online version • www.smw.chPublished under the copyright license "Attribution -Non-Commercial -No Derivatives 4.0".No commercial reuse without permission.See http://emh.ch/en/services/permissions.html.

Table 2 :
Estimated independent associations of smoking cessation between surveys with different personal characteristics 1 .

Table S1 :
Description of the samples included in the analyses of smoking cessation.FVC = forced vital lung capacity, FEV1 = forced expiratory volume in one second 8 At least 150 min/week of moderate or vigorous physical activity