Change in the Prevalence and Social Patterning of First- and Second-Hand Smoking in PORTUGAL: A Repeated Cross-Sectional Study (2005 and 2014)

Between 2005 and 2007, important reinforcements of the tobacco legislation have been implemented in Portugal, which may have affected smoking patterns. The aim of this study was to measure the change in prevalence of first- and second-hand smoking (SHS) among adults, and its socio-demographic patterning in Portugal from 2005 to 2014. Data from the last two Portuguese National Health Interview Surveys (2005 and 2014) were used. The changes in daily smoking and SHS were measured using Poisson regressions, stratifying by sex and survey year. The inequalities were measured using relative inequality indexes (RII). From 2005 to 2014, there was a reduction in SHS (75%–54% among men, and 52%–38% among women), and a reduction in smoking among men (27%–26%), and an increase among women (9%–12%). SHS reduction was more marked among less privileged people. Among Portuguese men, inequalities in daily smoking have increased slightly, while among women the gap favoring low-educated reduced. Between 2005 and 2014, SHS decreased, but not daily smoking, particularly among women. Additionally, socioeconomic inequalities in smoking increased. Future policies should simultaneously tackle smoking and SHS prevalence, and their socioeconomic patterning. More comprehensive policies such as comprehensive national (non-partial) bans, combined with price increases could be more effective.


Introduction
Altogether, first-and secondhand smoking are the second highest risk factors contributing to the global burden of disease in many cases being responsible for 156,838 disability-adjusted life years [1]. However, the epidemiological burden is not equally distributed in the population. Premature death is more likely among lower social classes, and smoking largely contributes to it. A study from 2006 in England and Wales, Poland, and North America at the ages of 35-69 years found that the mortality rate in the lowest social strata was twice the mortality in the highest social strata [2]. This same study estimated that smoking-attributed mortality was responsible for more than half of those differences.
Evidence on tobacco control policies (TCP) among adults showed that several measures, such as tax increases, smoke-free legislation, high-intensity media campaigns, stronger advertising bans, and health warnings, comprehensive cessation treatment, and youth access laws, can reduce smoking prevalence and deaths due to smoking [3], and that smoking bans have the power to reduce exposure to second-hand smoking (SHS) [4]. However, non-targeted interventions at population level may ultimately widen socioeconomic (SE) inequalities in smoking [5]. For example, cessation programs that do not target low SE level persons might have a harmful effect on inequalities, once high SE will benefit more from cessation opportunities than low SE status persons [6]. In the same line, the literature shows that TCP have an effect on SHS [4], but the effects on inequalities are still inconclusive [6].
In ). This legislation also reinforced regulation about cigarettes composition and labels, banned advertising, and prohibited sales to minors. The law included a partial smoking ban in bars and restaurants, leading to an improvement of the Portuguese position in the Tobacco Control Scale (TCS) ranking by four places [7].
Studying SHS is extremely important since it is an externality affecting people surrounding smokers due to the burning of tobacco. It contains a mixture of compounds that were found to be to be carcinogenic on subjects of a previous studies [4]. The SHS in Portugal was evaluated mainly among children [8,9], for specific outcomes [10] or among small study samples [11].
The Portuguese case is relevant because of its still relatively high smoking prevalence, namely among men (23.5% daily smoking among men according to OCDE [12]), and because last data showed that, during the 1987-2006 period, smoking was more concentrated among low SE status men, but still more prevalent among high SE women [13], indicating that Portugal was at an earlier stage of the tobacco epidemics. Thus, the aim of this study was to measure the change in prevalence of daily smoking and SHS exposure among adults, and its socio-demographic patterning, before and after implementation of tobacco control legislation (2005-2007).

Sample
We used data from the last two Portuguese National Health Interview Surveys, carried out in 2005/06 and 2014 [14,15]. These surveys were cross-sectional studies designed to be representative of non-institutionalized population living in Portugal. Both samples result from a multi-stage stratified probabilistic sampling scheme, were designed to represent the seven regions of the country, and included sampling weights computed as the inverse of the probability of selection in each sampling unit, adjusted for non-response and post-stratified to best represent the target population. Only individuals aged more than 25 years old were included in this analysis, in order to avoid young people who had not completed their education; in addition, individuals older than 79 years old were not considered, to reduce the selective mortality bias. Final database included 28,433 observations for 2005/06 and 15,196 observations for 2014.

Variables
Variables for smoking and exposure were dichotomous. Daily smoking equaled one if the participant smoked daily and zero if they did not smoke or smoked less than daily. For SHS, the question slightly changed from the 2005/06 to the 2014 survey. The question in 2005/06 was «how many time do you spend in closed spaces around smokers, throughout the week?», while in 2014, the question was «how often are you in closed spaces while other people smoke?». Thus, a dichotomous variable for SHS was created, equaling one if exposed and zero if not exposed. The participant was considered exposed when reported being always, most of time, quite amount of time, some time, or short amount of time in close spaces around smokers throughout the week in 2005/06, and when reported being daily or occasionally exposed to smoking in 2014. The participant was considered not exposed when reporting never spending time in close spaces around smokers throughout the week in 2005/06, and never been in close spaces while other people smoke in 2014.

Covariates
As regards SE variables, education included five categories: no education (zero years of education), primary first education (six years of education), primary education (nine years of education), secondary or post-secondary education (twelve years of education or post-secondary non-tertiary education), and tertiary education (bachelors, master or doctoral level). Income was defined as five income quintiles of family income. Family income was the household income per equivalent adult, computed with modified equivalence scale from the Organisation for Economic Co-operation and Development (OECD), which attributes different weights to different family members [16].

Statistical Analysis
Prevalence ratios representing the ratio between the prevalence in 2014 and in 2005-06, and respective confidence intervals, were estimated using Poisson regressions, recommended when the log-binomial has convergence problems [17]. The change in daily and SHS from 2005/06 to 2014 was computed as follows. Firstly, daily smoking was modelled using Poisson regressions [18] with education as variable of interest, adjusting for age and stratifying by sex and survey year. The same analysis was done for income. Secondly, the educational and income inequalities were measured using relative inequality indexes (RII). RII is an indicator that not only takes into account social differences in smoking but also considers the distribution of the population over the educational levels (ridit, a score that results from the midpoint of the population's cumulative distribution in the education/income categories) [19,20]. The analyses were adjusted for age, and stratified by sex. We estimated the RII in two separated Poisson regression models for each year, and then pooled the two surveys in a third regression by including an interaction term between the ridit term and the survey year [18]. Table 1 presents the descriptive statistics for the two surveys. There were 47% women in 2005, and 44% in 2014. Around 40% of the participants were 45-64 years old. The education levels changed between the two surveys, with a decrease in the percentage of individuals with no education, from 15% to 10%. The increase was also higher among the low educated, and among the lowest income quintile. Among women, the prevalence of SHS decreased from 51.93% in 2005/06 to 38.39% in 2014. The decrease was highest among the oldest, the lowest educational groups, and the lowest income quintile.

Results
The age-adjusted prevalence ratios (PR) for daily smoking and SHS are presented in Table 3. Among men, the daily smoking prevalence was higher among the men with lower education, comparing with those with tertiary education both in 2005 (a) Lowest quintile of income. The income was provided in quintiles, for a question of data protection. So, the authors did not access information on individual income. Therefore, the quintiles may not be representing each 20%, when we excluded some age categories.

Discussion
This study showed that, in Portugal, the number of people that reported exposure to second hand smoking largely decreased, while the number of daily smokers slightly decreased among men and increased among women, from 2005/06 to 2014. Also, among Portuguese men, socioeconomic inequalities in daily smoking have increased in magnitude against the worse-off, while among women the gap reduced, due to an increase in the prevalence among less-educated subjects. SHS was large but not consistent across sub-populations, being more marked among less privileged individuals.
In most European countries smoking is more concentrated in low social status persons [21]. Southern European countries were an exception to this behavior, since the higher educated women smoked more [22]. However, in the recent past, these countries moved to a later stage in the smoking epidemic, and that behavior was considered a slight deviation from the norm [22,23]. Yet, Portuguese women still lagged behind men in what respects the above-mentioned transition. The last study, from 1987-2005/06, showed that among men the inequalities increased while among women the gap reduction was not enough to observe the expected reversal in inequalities, despite the decrease of inequalities in cessation in both sexes. This finding was observed notwithstanding the important reinforcement of TCPs [13]. Indeed, between 2005 and 2007, the Portuguese government approved the WHO FCTC and a new legislation was enacted mainly to protect individuals against involuntary tobacco exposure and to reduce the demand for cigarettes, in particular regarding dependence and smoking cessation. This legislation also reinforced regulation about cigarette composition and labels, banned advertising, and prohibited sales to minors.
In Portugal, after the introduction of the TCP, the SHS decreased but this reduction was more marked among less privileged people. We can hypothesize that the 2009 economic crisis might have changed the meaning of income quintiles, and the exclusion of people under 25 from the sample, might have left out the ones that face the higher burden of the SHS exposure. In fact, the low socioeconomic Portuguese children are exposed to high levels of SHS at home [24]. Additionally, the number of smokers slightly decreased among men but increased among women, although remaining unequally distributed. The evidence of TCP in other countries showed that, despite the positive effect of decreasing prevalence, non-comprehensive policies can harm the social distribution of smoking. Partial smoke free policies increase socioeconomic inequalities in terms of protection to SHS, while national, comprehensive smoke free policies had the power to decrease inequalities [25]. This might be the case for Portugal, where the smoking ban implemented was only partial [26]. The legislation implemented in 2005 and 2007 was not subject to inspections, for example in what concerns sells to minors, and allowed several exceptions, as for example in the case of eating and drinking establishments (like restaurants or bars) with separated non-smoking areas and air extracting systems. Due to the study design, we cannot attribute the effect to the partial ban; but we also cannot ignore that the SHS reduction was more marked among the less privileged individuals. Thus, we hypothesize that the ban made smokers move outside (or to specific smoking areas within the establishments) to smoke, reducing the involuntary exposure to tobacco smoke, but this was not enough to make them quit smoking. Similar studies were made in Spain, following a comprehensive smoke free legislation. Authors concluded that comprehensive bans decreased smoking in settings targeted for the legislation, but the reductions were also observed in other settings not covered by the legislation (as bus stops, homes, and train stations) [27,28]. Unfortunately, we do not have enough data to investigate those effects.
This study also has some limitations. The question for SHS changed from 2005/06 survey to 2014 survey. However, we computed a dichotomous variable for exposure to SHS (high versus low), in order to minimize the impact of this change. Also, in Portugal, the existent surveys do not allow to build a trend to infer the real impact of recent legislation on smoking and SHS. In this period there were other changes in the legislation in place that might interfere with the effect (for example a ban to the sale of tobacco products to individuals younger than 16 years was approved in 2005). Therefore, this study could not infer directly on the impact of recent TCP on prevalence and inequalities. However, it was still possible to investigate the evolution of prevalence and inequalities in smoking from 2005 to 2014. Further studies could measure whether the legislation has been equally enforced across all the Portuguese municipalities and across different settings.

Conclusions
Among Portuguese men, inequalities in smoking against the worse-off have increased, while among women the gap reduced, but at the cost of an increase in the prevalence among the low-educated. From 2005 to 2014, there was a large reduction in SHS, and a significant reduction in daily smoking among men but not among women. SHS reduction was more marked among less privileged people. Future policies should be more comprehensive, to tackle simultaneously smoking and SHS prevalence, and their socioeconomic patterning. For example, comprehensive national (non-partial) bans, combined with price increases showed a consistent positive impact on inequalities [19]. Funding: This study is part of the project SILNE-R "Enhancing the effectiveness of programs and strategies to prevent smoking by adolescents: a realist evaluation comparing seven European countries", which was funded by the European Union's Horizon 2020 research and innovation program under the SILNE-R Grant Agreement number 635056.