Study design

EMENO, as mentioned above, is a cross-sectional health status survey conducted in Greece from May 2013 to June 2016 (field study conducted during 2014–2016). Ethics approval and consent to participate in the EMENO study was approved by the Ethics and Deontology Committee of the National and Kapodistrian University of Athens (Date: 8 November 2012, Protocol number: 1742) and by the Hellenic Data Protection Authority (Date: 7 December 2012, Protocol number: GN/ΕX/1069-1/07-12-2012). A modified version of the informed consent form (ICF) was approved by the Ethics and Deontology Committee of the National and Kapodistrian University of Athens (Date: 6 March 2013, Protocol number: 6315).

The survey was conducted by ‘door-to-door’ interviews using a Computer Assisted Personal Interview (CAPI) and augmented with data and physical examinations performed by trained personnel (reviewers and physicians), in a randomly selected sample representative of the non-institutionalized adult (≥18 years) population of Greece. The detailed EMENO study protocol has been described elsewhere¹¹. The survey sample consisted of 6006 individuals aged ≥18 years and covered urban (>10000 inhabitants), semi-urban (2000–9999 inhabitants) and rural areas (<2000 inhabitants) of the country. The overall response rate to interviews was 72%.

Outcomes of interest

Covariates

Statistical analysis

All EMENO participants with known smoking status (current, former, or never smokers) were included in the analysis. Sampling weights, being the reciprocal of the selection probabilities, were applied to adjust for the sampling design; sampling weights were multiplied by a correction factor to adjust for the discrepancies in the gender, age, and geographical area distribution of the sample from the corresponding one of the adult (≥18 years) general population (based on the 2011 census, provided by the National Statistics Agency).

Weighted means and standard deviations for normally distributed, medians and interquartile range (IQR) for non-normally distributed continuous variables and weighted percentages for categorical variables were provided. Weighted multivariable logistic regression models were fitted to investigate factors associated with ever smoking (ever vs never) or being current smoker (current vs former smoker). Demographic and socioeconomic variables (age, gender, educational level, partnership status, job status, ‘blue or white collar’ classification, place of residence), lifestyle factors (alcohol consumption, Mediterranean diet, physical activity) and BMI were investigated as potential covariates. Backward elimination procedure was implemented to select the statistically significant covariates and possible interaction effects were also investigated. In non-binary categorical variables, categories were collapsed if there were no indications for significant differences in the estimated ORs based on the Wald test.

General features of smoking prevalence in Greece

Out of 6006 individuals participating in the survey, 13 had missing values for age and gender and 131 (2.2%) for smoking status, yielding a total of 5862 study participants. The characteristics of the study population, expressed also as weighted averages, are presented in Table 1.

The majority of the study population (62.7%) was living in urban areas, 47.1% had finished at least secondary school and 40% had less than 900 euros monthly family income. Forty-three percent were non-drinkers and 42.1% moderate drinkers, whereas nearly 80% had moderate adherence to MD (79.5%) and 71.3% moderate/high physical activity. Regarding BMI, 29.2% of the study population were overweight. The overall prevalence of current smoking was 37.8% (95% CI: 36.3–39.3) and of former smoking 16.1% (95% CI: 15.1–17.2) (Table 1). Descriptive characteristics by smoking status are also shown in Table 1, with the distribution of all presenting factors being different among smoking groups. There were 98 (3.2%) participants with missing data for pack-year consumption of manufactured and hand-rolled cigarettes. Among those with available information, the median number of pack-years (cigarettes and hand-roll cigarettes) of ever smokers was 18.6 (IQR: 7.5–36.5). Boxed cigarettes are the tobacco product most used among smokers (91% of all tobacco products) (data not shown).

Current smoking is prevalent during adulthood (18–29 years: 44.6%; 30–39 years: 47%; 40–49 years: 48.1%; 50–59 years: 42.9%; p<0.001), but declines sharply in the elderly (60–69 years: 29.4%; >70 years: 12.7%; p<0.001). This generational difference is much more evident in women (Figure 1a). Among former smokers, the findings were exactly the reverse (18–29 years: 6.4%; 30–39 years: 11.3 %; 40– 49 years: 14.3%; 50–59 years: 18.7%; 60–69 years: 26.3%; >70 years: 23.11%; p<0.001). Both genders had high prevalence of current smoking, but men were more likely than women to be current (44.3% vs 31.6%, p<0.001) (Figure 1a) and heavier smokers [median pack-years (IQR): 22.5 (10–45) vs 13.5 (5.5–25), p<0.001] (results not shown). Conversely, women were less likely to be former smokers (8.2% vs 24.5%, p<0.001) than men (Figure 1a).

Figure 1

Prevalence of current and former smoking by age and gender (a) and mean age at onset of smoking by gender (b) in Greece during 2014–2016

For 49 participants (1.6% of ever smokers), the age at initiation of tobacco smoking was unknown. Among those with available information, the median age of daily smoking initiation for ever smokers (current and former smokers) was 18 years for women (IQR: 16–21) and 18 years for men (IQR: 15–20). It is remarkable that one in two males (48.7 %) and one in three females (36.2%) reported starting smoking before the age of 17 years. Moreover, 31.3% of ever smokers aged 18–29 years reported smoking initiation at age <15 years (results not shown). However, the crucial ages for smoking initiation in both genders appears to be the four-year period between 18 and 22, during which a substantial proportion take up smoking (41.9% of males and 44.5% of females). Finally, there seems to be generational differences regarding the age of smoking initiation for both sexes. Younger individuals reported to start smoking earlier than older ones. This difference is particularly evident in older women (Figure 1b).

Multivariable analysis to assess the factors associated with being ever smoker

All factors presented in Table 1 were assessed for their independent association with ever smoking. Results from multivariable logistic regression showed that among all investigated factors, only alcohol consumption and unemployment status were independently associated with ever being a smoker (Table 2). More specifically, the heavier the alcohol consumption the higher were the odds of ever smoking; compared to non-drinkers, those who were drinking 1–7 glasses per week had 41% higher odds, and those who were drinking >7 glasses had 152% higher odds of being ever smokers.

Additionally, unemployed participants had 42% higher odds of being ever smokers compared to employed participants (Table 2). Significant interactions were found between age and gender (p<0.001), age and educational level (p=0.001), and gender and educational level (p=0.001). As there were no significant differences between those who had completed secondary and those with higher education across gender and age groups, these two groups were combined and compared with those who had only primary education. Educational level was not significantly associated with the odds of ever smoking among women aged 18–65 years, whereas for women aged >65 years having secondary or higher educational level compared to having a primary education level was associated with significantly higher odds of ever being a smoker (OR=2.19; 95% CI: 1.57–3.06) (Table 2). On the other hand, among men, having a higher educational level was associated with lower odds of ever being a smoker for all age groups (OR=0.44; 95% CI: 0.24– 0.84 for those aged 18–34.9 years; OR= 0.67; 95% CI: 0.44–1.03 for those aged 35–49.9 years, and OR=0.52; 95% CI: 0.35–0.79 for those aged 50–64.9 years), apart from those being ≥65 years.

Females were less likely to have ever been smokers compared to males with the difference between the two genders being more pronounced for those who had only primary education compared to those having higher education as well as for those aged >50 years (Table 2).

Multivariable analysis to compare current with former smokers

To investigate factors associated with giving up smoking, we restricted the analysis to those who had ever been smokers and compared current with former smokers. Current smokers reported lower levels of adherence to MD [compared to those with low adherence levels, the adjusted ORs (95% CI) of being current smoker were 0.55 (95% CI: 0.36– 0.83) and 0.35 (95% CI: 0.21–0.58) for those with moderate and high adherence levels, respectively]. Furthermore, they were less likely to be either overweight (OR=0.71; 95% CI: 0.54–0.95) or obese (OR=0.60; 95% CI: 0.45–0.80), and more likely to be single (OR=1.47; 95% CI: 1.15–1.88) compared to former smokers (Table 3). Two significant interactions, one between age group and educational level (p=0.019) and one between gender and age group (p=0.024), were found. For those aged <50 years, those having secondary or higher education were associated with lower odds of being current than former smokers compared to participants having only primary education. There were no significant differences from the odds of being current smokers in relation to their educational level for those aged >50 years (Table 3). The gender effect differed across age groups, but it was the same across educational levels within the same age group. Whereas there were no significant differences in the odds of being a current smoker between women and men of the younger age group (18–35 years), in the other age groups the older the age, the more likely were women to be current smokers compared to men (Table 3). In general, the odds of being current smokers decreased with increasing age, with age differences being more pronounced among men of primary education.

Secondhand smoking in Greece

In Table 4, information on the prevalence of reported exposure to SHS during the last six months is presented. Overall, 58.3% of the respondents were exposed to SHS. The prevalence of exposure was 38.8% at work, 30% at home, and 44.6% in public places. The majority of the population was exposed to SHS for less than five hours per day regardless of place of exposure. Tobacco ever smokers reported higher percentages of passive smoking either overall (71.1%) or in public and working places (54.7% and 45.7%, respectively), compared to never smokers (43.5%, 32.9%, and 28.5%, respectively).

Future perspectives

In Greece, the introduction of legislative measures to limit tobacco exposure began in 1952. For more than 60 years, a multitude of laws have been enacted that aimed at both a smoking ban in public places as well as the banning of any form of advertising campaign. However, looking at the high rates of current smoking and SHS exposure levels that are still present, we understand that the existing smoking bans have not been satisfactorily implemented and enforced. It would seem appropriate that a similar study should be conducted in Greece, to confirm or disprove our results, in the near future. In this way, the impact on the percentage of both smoking and passive smoking exposure can be ascertained following the enhanced anti-smoking campaign that was initiated in the beginning of 2020. Moreover, based on the findings of the EMENO that smokers were more likely to combine other lifestyle risk factors, (unhealthy diet and harmful alcohol use) it will be interesting to plan future studies on these health-related behaviors and mortality rates, not only nationally but also across Europe.

Strengths and limitations

The representative sample of the adult Greek population, the satisfactory response rate, the collection of data by trained interviewers, and the low rate of missing values are the strengths of our study and offer the opportunity to draw some generalizable conclusions regarding smoking prevalence and smokers’ lifestyle in Greece. The most significant limitation, as with other similar studies, is the self-reported nature of the data collected, which may result in under-reporting of smoking. However, a review study on self-reporting of smoking behavior, concluded that self-reports of smoking were quite accurate³⁴. It is worth noting that as our study methodology was quite different from those implemented in other national and international studies on smoking prevalence, our results may not be comparable.