Background

Cigarette smoking rates among adult women have been found to be very low (<10%) in East Asian countries, including China, Hong Kong and South Korea (hereinafter ‘Korea’),1 which often lead national anti-smoking policies to concentrate on male smokers. Although the absolute smoking prevalence among Asian women is likely to be lower than in Western women, one of the potential explanations for the low rates of female cigarette smokers in East Asian countries is under-reporting of smoking.

Comparisons of smoking rates determined by self-reporting and by biochemical verification have found a general trend of underestimation in self-reported smoking rates.2 ,3 However, the magnitude of underestimation was found to vary by population studied and by the biological samples tested.2 ,3 Furthermore, most of these studies were from Western developed countries, where gender differences in smoking rates are smaller than in countries at less advanced stages of economic development and/or in countries with other socio-cultural backgrounds.2 ,3

Only a few studies comparing self-reported and biochemical smoking rates have assessed gender differences in under-reporting of smoking. For example, studies using national representative and community data in the USA showed negligible gender differences in under-reporting of smoking,2 ,3 and a study from Finland found that 2.5% of men and 2.7% of women self-reported non-smokers had detectable serum concentrations of cotinine.4

Outside the west, especially in East Asian countries with very low female smoking rates, there is a paucity of studies comparing self-reported and biochemically verified smoking rates. To our knowledge, no national study in Asian countries to date has estimated the true levels of underestimation of smoking rates. One very small study (34 people) from Iran showed that under-reporting of smoking was slightly greater among women than among men, with differences in cotinine-verified and self-reported rates of 5.4% and 2.5%, respectively.5

Smoking prevalence based on self-reporting may well be underestimated when these behaviours are considered socially undesirable. Thus, self-reported prevalence of smoking may be especially underestimated in certain population subgroups such as pregnant women, adolescents and people with smoking-related diseases.6–8 This under-reporting may also be affected by cultural backgrounds in which smoking is stigmatised among certain subgroups. For example, the influence of Confucianism, resulting in social repression of women's smoking, is considered a major reason for low tobacco use rate among Korean women.9 The low prevalence of smoking among women in many Asian countries has been associated with low levels of national efforts towards curbing smoking prevalence in women. Thus, there is an urgent need to estimate true national levels of female smoking in Asian countries using biochemical verification methods such as measurements of cotinine concentration. Cotinine is a principal metabolite of nicotine, with a half-life of about 20 h and is therefore the best available biochemical measure of current nicotine consumption.10 Cotinine can be measured in blood, urine, saliva and hair.

Using a nationally representative sample from an Asian population, we have compared self-reported and cotinine-verified smoking rates in Korean men and women. We also investigated gender difference in under-reporting of smoking in Korea, a country in which the female smoking rate is very low and the gender difference in smoking rate is very high.

Results

Of the 5455 study subjects, 56.2% (n=3068) were women and 43.8% (n=2387) were men (Supplementary table 1). Mean (±SD) subject age was 49.4 (±16.2) years (range, 20–92 years). Approximately 30% of men and 20% of women had completed college or higher education, while more than one-half of the men and one-third of the women had manual jobs. Nearly 80% of men and 70% of women were married and living with their spouses, whereas fewer than 25% of each were living with their parents. About 16%–17% of the respondents were living in Seoul, the capital city of Korea.

Supplementary table 1 also presents prevalence rates of cigarette smoking, both by self-report and by urine cotinine verification. Self-reported smoking prevalences were 44.7% for men and 5.9% for women. Mean numbers of cigarettes per day were 16.8 for men and 9.7 for women. Cotinine-verified smoking prevalences for men and women were 50.0% and 13.9%, respectively. Cotinine-verified smoking prevalences were greater than self-reported prevalences by 5.3% points in men and 8.0% points in women. SHS exposure at work and at home was reported by 49.1% of men and 34.7% of women.

Table 1 presents cross-tabulation results on cotinine-verified smoking status by self-reported smoking status and SHS exposure by cotinine-verified smoking status. Of the total of 1620 cotinine-verified male and female smokers, 24.4% were self-reported non-smokers (never-smokers plus ex-smokers). This under-reporting, however, differed markedly by gender. Of male cotinine-verified smokers, 12.0% were self-reported non-smokers (4.0% never-smokers and 8.0% ex-smokers), whereas of female cotinine-verified smokers, 58.9% were self-reported non-smokers (54.2% never-smokers and 4.7% ex-smokers). In comparison, 1.5% and 0.2% of cotinine-verified male and female non-smokers, respectively, reported being current smokers.

Of the 3084 subjects without SHS exposure, 42.5% of men and 8.5% of women were cotinine-verified smokers. The proportion of cotinine-verified smokers was higher in subjects with SHS exposure. Of the subjects exposed to SHS for ≥1 h/day, 70.7% and 29.4% were cotinine-verified smokers in men and women, respectively (table 1).

Table 2 shows absolute differences between cotinine-verified and self-reported smoking rates and ratios of cotinine-verified to self-reported smoking rates according to the study variables. Absolute differences between cotinine-verified and self-reported smoking rates were greater in women (8.0%) than in men (5.3%), which decreased to 5.2% for women and 4.5% for men among subjects without SHS exposure (Supplementary table 2). The ratios of cotinine-verified to self-reported smokers were 2.36 for women and 1.12 for men. For men, percentage differences and ratios between cotinine-verified smoking and self-reported smoking did not significantly differ with study variables except for marital status, in that the percentage difference and ratio were slightly higher for men cohabiting than not cohabiting with a spouse. For women, the percentage differences and ratios between cotinine-verified and self-reported smoking were greater among those of younger age, higher education and non-manual occupations, and among women living with parents or a spouse than those living alone or with others. When we performed sensitivity analysis using urinary cotinine cut-off of 100 ng/ml, absolute differences between cotinine-verified and self-reported smoking rates were 3.9% for women and 2.8% for men (Supplementary table 4). The ratios of cotinine-verified to self-reported smokers were 1.66 for women and 1.06 for men. Our sensitivity analysis for 75 ng/ml presented greater absolute differences and relative ratios for men and women than the analysis for 100 ng/ml (Supplementary table 4).

Table 3 shows the percentages of self-reported male and female non-smokers among cotinine-verified smokers and gender differences and ratios of percentages of under-reporting by cotinine-verified smokers. Of the 1194 cotinine-verified male smokers, 12.1% responded that they currently do not smoke compared with 58.9% of 426 cotinine-verified female smokers. The gender difference in self-reported non-smoking prevalence among cotinine-verified smokers was 45.4% and the gender ratio was 4.85. These gender differences and ratios of under-reporting tended to be greater among respondents in their 30s, among married respondents cohabiting with a spouse, among respondents cohabiting with their parents and among residents in rural areas, but lower among those with education levels less than elementary school (table 3). When we performed similar analysis with subjects without SHS exposure, gender differences and ratios of self-reported non-smoking among cotinine-verified smokers were 50.1% and 4.90, respectively, which were not significantly different from the results among all subjects (Supplementary table 3).

Table 4 shows gender differences in self-reported non-smoking among cotinine-verified smokers. The PR of under-reporting of smoking among women was 4.85 (95% CI 4.02 to 5.86) in the baseline model (model 1) and was 5.35 (95% CI 4.34 to 6.60) in the model fully adjusting for age, socio-demographic variables and SHS. PR was greater among those with higher education, married cohabiting with a spouse and cohabiting with parents. PD results, showing absolute per cent differences in under-reporting of smoking by gender among cotinine-verified smokers, showed patterns similar to those for PR. Prevalence of self-reported non-smoking among cotinine-verified smokers was around 45% greater among women than among men (95% CI 38.1 to 52.7 in the baseline model) and did not change significantly after further adjustment for age, socio-demographic variables and SHS. PD was significantly greater for respondents with college or higher education (PD=10.6, 95% CI 3.5 to 17.8 in the fully adjusted model) than for those with elementary or less education. PR (6.02, 95% CI 4.39 to 8.24) and PD (49.2, 95% CI 37.4 to 61.0) of under-reporting of smoking among subjects without SHS exposure were some greater than total cotinine-verified smokers.

When we performed sensitivity analysis using urinary cotinine cut-off of 100 ng/ml, PR of under-reporting of smoking among women was 5.00 (95% CI 3.85 to 6.49) in the age and age square adjusting model, which was not different from cotinine cut-off of 50 ng/ml. Prevalence of self-reported non-smoking among cotinine-verified smokers was 33% greater among women than among men (95% CI 25.3 to 40.7) (Supplementary table 4).

Discussion

Our results indicate that there may be a large number of hidden female smokers in Korea. Using the age-specific prevalence rates of smoking, both by self-report and cotinine verification, based on the 2008 Korean registration population, and the findings reported here, we estimate that, in 2008, there were 9.32–9.79 million (with urinary cotinine cut-off of 50 or 100 ng/ml) cotinine-verified male and 2.01–2.87 million cotinine-verified female smokers, whereas there were only 8.76 million self-reported male and 1.21 million self-reported female smokers. Subtracting the number of self-reported smokers from the number of cotinine-verified smokers indicates that, in 2008, there may have been 0.56–1.03 million hidden male and 0.8–1.65 million hidden female smokers in Korea. In addition to there being more hidden female than hidden male smokers, 42.7%–57.6% of all female smokers were hidden. This result suggests that, in Asian countries that share similar patterns of huge gender differences in self-reported smoking prevalence and similar socio-cultural backgrounds that frown upon female smoking, more than half of female smokers may be hidden.

To our knowledge, this is the first study to compare self-reported and biochemically verified smoking rates among a representative national sample in an East Asian country. We found that the ratio of biochemically confirmed to self-reported smoking was 2.36 times greater among women compared with only 1.12 times greater among men. The PR of self-reported female and male non-smokers among cotinine-verified smokers was greater than five, suggesting that Korean women smokers hide their current smoking five times more often than do men. This finding differs considerably from in Western populations, in which there were either no gender differences or a higher non-disclosure rate among men than among women.2 3 16 17 Under-reporting of smoking, however, was more prevalent among women than among men in a Muslim country, Iran, which has a similar pattern of gender difference in smoking rate as East Asian countries.5

Three possible explanations for greater gender differences in under-reporting rate of smoking should be considered: (1) gender differences in smoking patterns and metabolism of nicotine, (2) gender differences in exposure to SHS and (3) social repression of reporting of smoking as well as smoking among women.

In general, female smokers were found to have lower urinary cotinine concentrations than men who smoke the same numbers of cigarettes.18 This may be due to more rapid metabolism of nicotine and its metabolite cotinine by women19 and/or to female preferences for cigarettes with lower nicotine concentrations.18 Because we did not have a gender-specific cotinine cut-off for Asian populations, we used the same concentration for men and women to differentiate smokers from non-smokers. This decision may have led to some smokers with lower cotinine concentrations being classified as cotinine-verified non-smokers, resulting in reduced, rather than increased, gender differences in PR and/or PD.

Gender differences in SHS exposure at home or at work could influence gender differences in PR and/or PD considering the positive relation between under-reporting of cigarette smoking and exposure to SHS exposure among both genders (table 1). Cotinine concentrations in smokers and in non-smokers exposed to SHS can overlap, and there is no clear cotinine cut-off to differentiate these two groups.20 However, we found that self-reported SHS exposure was greater among men than among women, thus reducing the gender difference or gender ratio (PR or PD). Furthermore, the addition of SHS exposure to multiple regression models did not change PR or PD. When we analysed subjects without SHS exposure, gender differences and ratios in percentages of self-reported non-smoking did not change significantly (46.3%–50.1% for gender differences and 4.70–4.90 for gender ratios, respectively; table 3 and Supplementary table 3). All these findings suggest that SHS contributes non-disclosure rate of cigarette smoking among both genders; however, these do not seem to contribute gender difference of under-reporting of smoking.

We found that the PR of self-reported non-smokers among cotinine-verified smokers was greater among married women than among men cohabiting with a spouse and among women than among men living with their parents. Women living with a spouse or parents in a patriarchal society with veneration of elders, such as those influenced by Confucianism, may be under strong social pressures not to smoke or to hide their smoking status. Previous studies in Korea showed that the smoking rate was lower for married than for unmarried (including divorced/widowed) women.9 We found, however, that the one Finnish study did not show any difference of misclassification of smoking status according to age and marital status.4 PR of self-reported non-smokers among cotinine-verified smokers did not change after full adjustment for variables considered related only to social repression among women (ie, marital status, cohabiting with parents and cohabiting with spouse). This result suggests that the variables used in this study as proxies for social repression among women did not fully capture the construct of societal repression (especially related to health behaviours) originating from the Confucian influence in Korea. Thus, other socio-cultural factors not included in this analysis may have prevented women from correctly reporting their smoking status.9

The urinary cotinine cut-off we used may have affected the rate of under-reporting of cigarette smoking. The optimal cut-off distinguishing cigarette smokers and non-smokers has been found to differ among racial and ethnic groups.21 Because no cut-off has been specified for Asian populations, we used the urinary cotinine cut-off of 50 ng/ml proposed by the Society for Research on Nicotine and Tobacco.12 Changing the cut-off to 100 ng/ml, as proposed in one Korean study,22 resulted in decreases in cotinine-verified smoking rates, from 13.9% to 9.8% in women and from 50.0% to 47.5% in men. This change also resulted in decreases in differences (8.0%–3.9% for women vs 5.3%–2.8% for men) and ratio (2.36–1.66 for women vs 1.12–1.06) of under-reporting, and PR and PD of non-reporting of smoking among women than among men was materially the same as in the original analysis (Supplementary table 4).

The strength of this study is that, by using a nationally representative sample, we could determine the point prevalence of cotinine-verified smoking rate and could estimate the level of under-reporting in an Asian population with a relatively low female smoking rate.

Several limitations of our study need to be considered. First, respondents using nicotine replacement therapy (NRT) could not be excluded because the question on NRT use did not specify its timing. Only 43 ex-smokers, however, reported using NRT for smoking cessation in the past, indicating that NRT did not significantly affect our results. Second, we did not determine dietary nicotine exposure, which may differ by gender. However, the dietary contribution to serum cotinine is very small,23 making it unlikely to have affected our results.

Our findings have important implications for future studies and tobacco control policy. We found that under-reporting of smoking status was widespread among Korean women. Because measurements of cotinine are better indicators of tobacco exposure than self-reporting, biochemical validation of cigarette smoking in Asian countries should be considered as well as being monitored by self-reporting. However, smoking prevalence surveillance using biochemical verification can be expensive in some developing countries and cannot be feasible in all settings. Even though self-reporting of smoking has a problem of underestimation of actual smoking, it can give invaluable information on subgroups at greater risk (eg, lower socioeconomic group, ethnic minorities). In addition, biochemical verification of the representative subgroup of all survey subjects could be a cost-effective alternative. Additional studies, using more thorough measurements of social repression among Asian women, are needed to investigate factors related to gender differences in under-reporting of smoking. Particularly, policy makers should be aware that large numbers of female smokers in Asian countries with low self-reported smoking rates may be hidden. Smoking control policies specifically targeting women are therefore urgently needed in Asian countries, including Korea.