MATERIAL AND METHODS

Study population

This study is based on the Malmö shoulder-neck study (MSNS), which is a subcohort of the Malmö diet and cancer study (MDCS). Malmö is a city in southern Sweden with about 250 000 inhabitants. In 1990 all subjects born in 1926 to 1945 and living in Malmö were defined as a cohort (n = 53 325) for the MDCS. The recruitment to the MDCS took place from March 1991 until September 1996. The MSNS took place between February 1992 and December 1994, and included 14 555 subjects (6489 men and 8066 women) from this cohort. Detailed information concerning the MSNS and the MDCS is given in two other studies.^22,23 The study cohort was approached in two ways, by postal invitation (with respondents randomly selected from the 45–69 year age brackets) or by direct contact taken by the proband after a media campaign. The focus in the information given in the invitation was on the relation between diet and cancer and not on smoking or musculoskeletal problems.

All who participated in the MSNS baseline study were also invited to participate in a second examination one year later (median 12.6 months, interquartiles 12.3–13.3 months). A questionnaire was sent to all participants in the baseline study still registered in the municipality of Malmö. Information letters introduced the questionnaire and two written reminders, and finally a telephone call followed, if needed. In total 12 507 participated in the second examination, giving a response rate of 86%. On their return, the questionnaires were immediately checked for missing values and completed by telephone, if necessary.

Definitions

Outcome variable

The smoking item (“Do you smoke?”) in both the baseline and the one year follow up questionnaires contained four alternatives: daily smoker, intermittent (non-daily) smoker, stopped smoking, and never smoked. The baseline intermittent smokers had at the one year follow up either become daily smokers (intermittent/daily), remained intermittent smokers (intermittent/intermittent), or had stopped smoking (intermittent/stopped). The three groups of baseline intermittent smokers are separately compared to the aggregate sum (reference population) of daily smokers, former smokers (stopped), and respondents who had never smoked in the univariate and multivariate analyses.

The smoking item includes cigarette, cigar, and pipe smoking, but the vast majority (80.9% of all men and 97.3% of all women) were cigarette smokers.

The reliability of the smoking item was assessed by investigating the test-retest stability of 200 respondents within two weeks after the baseline examination. The test-retest stability was very high. The κ coefficient was 0.96 for all 200 respondents, 0.99 for the men, and 0.94 for the women. No age differences in reliability were observed, since the κ coefficients for the smoking item was 0.97 for the < 58.1 year group and 0.96 for the > 58.1 year group.

Independent variables

The age of the participants was computed from birth to the first visit to the MDCS centre and categorised into five groups.

For country of origin, all participants born in countries other than Sweden were merged into a single category.

Four categories were used for marital status: married, never married, divorced, and widow/widower. In the final analyses, the married category was compared to all unmarried.

Education was categorised by length of education. The respondents were classified into three groups: university degree, medium level (university studies without degree or less than three years of university studies, senior high school), and basal level (primary school, nine years or less).

Social participation (during the past year) describes how actively the person takes part in the activities of formal and informal groups in society. Respondents were asked whether in the previous 12 months they had been involved in any of the following activities: study circle/course at workplace, other study circle/course, union meeting, meeting of other organisations, theatre/cinema, arts exhibition, church, sports event, letter to editor of a newspaper/journal, demonstration, night club/entertainment, large gathering of relatives, private party. It was measured as an index consisting of 13 items and dichotomised. If three alternatives or less were indicated, the social participation of that individual was classified as low.

Social anchorage (five items) described belonging to formal and informal groups and the feeling of membership in these groups (familiarity with neighbourhood, sense of belonging to friends and relatives, membership in organisations or clubs, feeling of belonging at the place of work, and feelings of importance to other people). If three or more items denoted low social anchorage, the whole index variable was regarded as low.

Emotional support (three items) reflected the opportunity for care, encouragement of personal value, and feelings of confidence and trust. Each item had four alternatives: “Yes, I am absolutely sure to get such support”, “Yes, possibly”, “Not certain”, and “No”. The three latter alternatives were classified as low emotional support. If two or three items were low, emotional support was considered low.

Instrumental support (one item) reflected the individual's access to advice, information, practical services, and material resources from other persons. This item was measured in the same way as the three emotional support items. The three latter alternatives were classified as low instrumental support.

The reliability and validity of the four psychosocial variables have been assessed in several other studies.^24,25 The different items showed a good or acceptable validity and reliability. The test-retest stability was high. The κ coefficients for the social support variables social participation and social anchorage were 0.70 and 0.66, respectively. The κ coefficients for the emotional and instrumental support variables were 0.57 and 0.47, respectively. The construct validity analysed by Cronbach's α was highest for emotional support (0.63) and social participation (0.61), while social anchorage scored the lowest (0.40). The analysis of construct validity indicated that the different indices measure different aspects of the psychosocial environment.

Nicotine consumption in the form of oral snuff is a common habit in Sweden with a prevalence of approximately 15–20%.^26,27 The prevalence of snuff intake (yes/no) was assessed.

Statistics

Three groups of baseline intermittent smokers that either had remained intermittent smokers, had become daily smokers, or had stopped smoking at the one year follow up were compared to the reference population in a logistic regression model according to sociodemographic, psychosocial, and snuff consumption characteristics. A multivariate logistic regression model was used to assess differences in psychosocial conditions, adjusting for age, sex, country of origin, marital status, education, and snuff consumption. The statistical analysis was performed using the SPSS software package.²⁸

RESULTS

Table 1 shows that the proportions of daily and intermittent smokers at baseline were very similar among men and women. The proportion of never smokers was much higher among women (44.6%) than among men (28.1%). On the other hand, the proportion of individuals that had stopped smoking was much higher among men (42.0%) than among women (26.8%). The distribution according to age, country of origin, social participation, and social anchorage did not differ between men and women. Men were married to a somewhat higher extent than women, and women were divorced and widows to a higher extent than men. A higher proportion of women had high emotional support and high instrumental support. On the other hand, a higher proportion of women also had only a basal level of education. Only 0.4% of all women were snuff consumers, compared to 7.0% of all men.

The prevalence of daily smoking decreased from 23.8% to 21.7% (p < 0.001) at the one year follow up among the 86% that participated at both the baseline and the one year follow up, while the prevalence of intermittent smoking increased from 4.8% to 5.4% (p < 0.001). The proportion that had stopped smoking increased from 33.7% to 35.1% (p < 0.001), while the proportion of never smokers remained at 37.7% (not shown in tables).

A 59.9 % majority of all baseline intermittent smokers remained intermittent smokers at the one year follow up, while 15.9% had become daily smokers and 19.2% had stopped smoking. An interesting 5.0% of all baseline intermittent smokers stated that they had never smoked at the one year follow up. This group was regarded in the following analyses as baseline intermittent smokers that had stopped smoking (not shown in tables).

Tables 2 and 3 show that the long term intermittent smokers (intermittent/intermittent) were significantly younger and had a significantly higher level of social participation than the reference population. The transitional intermittent smokers that had become daily smokers (intermittent/daily) had low social participation to a significantly higher extent (odds ratio (OR) 1.79, 95% confidence interval (CI) 1.19 to 2.69) than the reference population. The transitional intermittent smokers that had stopped smoking (intermittent/stopped) were younger and unmarried to a significantly higher extent than the reference population. They also had a significantly lower proportion of persons with basal level education (OR 0.69, 95% CI 0.49 to 0.97) than the reference population. The odds ratios of being a snuff user were higher in all the three baseline intermittent smoker groups compared to the reference population.

Table 4 shows that the long term intermittent (intermittent/intermittent) smokers still had a significantly lower proportion of persons with low social participation compared to the reference population after adjustment for age, sex, country of origin, marital status, education, and snuff consumption in the multivariate logistic regression analysis. In contrast, the odds ratio of low social participation among transitional intermittent/daily smokers was 1.79 (95% CI 1.17 to 2.75) compared to the reference population in the multivariate analysis.

The exclusion of the baseline intermittent smokers that reported that they had never smoked at the one year follow up from the intermittent/stopped group yielded the same results as those presented in tables 3 and 4 for the intermittent/stopped group.

DISCUSSION

A 60% majority of all baseline intermittent smokers had remained intermittent smokers, 16% had become daily smokers, and 24% had stopped smoking at the one year follow up. The long term intermittent smokers and the transitional smokers who had stopped smoking were young, unmarried, highly educated, and snuff consumers to a higher extent than the reference population. They also seem to have more psychosocial resources, especially social participation, than the total population. In contrast, the social participation of the baseline intermittent smokers that had become daily smokers was poorer than in the reference population.

The present results could be biased by selection bias, misclassification, and confounding.

A comparison with another investigation conducted in the city of Malmö during the same time period with a higher participation rate (71%) showed a good correspondence in the same age groups concerning socioeconomic status, smoking, and social participation.²¹ Some studies have shown that non-participants differ from study participants in terms of smoking habits.^29,30 The smoking prevalences in these studies have been shown to be somewhat higher among non-participants. This is confirmed in our study by the finding that the prevalence of daily smoking at baseline was 24.4% among all baseline participants, but only 23.8% among those 86% of the baseline participants who participated both at baseline and at the one year follow up. However, we do not find any plausible reason for assuming that the tendency of non-participation would be higher for intermittent smokers compared with daily smokers. Accordingly, the difference between the one year follow up participants and the non-participants in our study are probably not biased to any important extent by selection.

The validity of items assessing smoking has previously been analysed several times. The results have consistently shown that self reported tobacco smoking is a valid and reliable way to measure smoking habits in a population.^9,31–35 The test-retest stability of the smoking item within two weeks was very high, the κ coefficient indicating an extremely high reliability. Differential misclassification is not likely to have been present. Non-differential misclassification seems to be a problem of less importance in this study, since non-differential misclassification tends to attenuate true differences, and the main results of this study show clear differences between different long term and transitional intermittent smoker groups and the reference population. The reliability and validity of the psychosocial variables showed an acceptable validity and reliability,³⁴ although the generally not very high Cronbach α coefficients of the psychosocial index variables are to some extent a limitation of this study.

Age, sex, country of origin, marital status, education, and snuff consumption could be confounders of the associations between the psychosocial variables and long term and transitional intermittent smoking. Adjusting for these variables, however, only marginally affected the estimates.

The 7% prevalence of snuff consumption among men may be regarded as low compared to the prevalence sometimes reported for Sweden.^26,27 However, other unpublished data from Scania in southern Sweden reveal the same prevalence of snuff use in this part of Sweden.

The longitudinal study design may be considered the most important strength of this study, because it makes it possible to follow the smoking history of the individual long term and transitional intermittent smokers for at least somewhat more than a year.

The 45–69 year age restriction is to some extent a limitation, since younger smokers (those who recently started smoking) are probably more likely than older smokers to exhibit changes in smoking patterns.

Despite the very high reliability of the smoking item, 30 baseline intermittent smokers reported at the one year follow up that they had never smoked. There is no certain explanation for this phenomenon. One possible explanation could be that intermittent smokers that have never smoked daily do not regard themselves as ever-smokers when they quit, because they have a much weaker identity as smokers than daily smokers during their smoking career. This notion is supported by the fact that only two out of 217 baseline daily smokers that had stopped smoking reported that they had never smoked at the one year follow up.

Previous studies have suggested that intermittent smokers could be either long term intermittent smokers—that is, in a long term stage (more than one year) of intermittent smoking—or in transitional stages from daily smoking to smoking cessation or from non-smoking to daily smoking.¹ Intermittent smokers appear to be younger, more highly educated, and have higher occupational status than daily smokers.^1,6 Their level of social participation is also higher than among daily smokers, and more similar to the level of social participation among non-smokers. However, this study has shown that the level of social participation differs between the long term intermittent smokers (intermittent/intermittent), and the transitional intermittent/daily and the intermittent/stopped groups. The long term intermittent smokers seem to have a lower proportion of individuals with low social participation compared to the reference population. Such a tendency, although not significant, was also observed for the transitional intermittent/stopped group. In contrast, the baseline transitional intermittent smokers that had become daily smokers had a significantly higher proportion of individuals with low social participation.

Social participation may be a factor that protects against the transformation of intermittent smokers into daily smokers. There are at least two mechanisms that can explain this statistical pattern. Firstly, social participation—that is, many different contacts with different forms of social life—may be a psychosocial asset for the individual intermittent smoker that works as a barrier against the development into daily smoking. The development to daily smoking might otherwise be facilitated by, for example, psychological, social or economic stress. Secondly, high levels of social participation may also be viewed as a source of knowledge, innovation, and transmission of certain values that affect smoking behaviour. The values transmitted may of course theoretically either be such values that affect the intermittent smoker to become a daily smoker, to remain an intermittent smoker, or to stop smoking. However, the generally declining trends in smoking in Sweden and other western countries during the past decades would imply that high levels of social participation generally would serve to facilitate the continuation of intermittent smoking or smoking cessation, rather than the initiation of daily smoking. The theory of diffusions of innovations suggests that some segments of the population adopt to changes in society earlier than others.³⁶ One non-material resource that could explain the decision of an intermittent smoker to stop smoking or remain an intermittent smoker in contrast to becoming a daily smoker may be a high level of social participation as defined in this study.

The finding that all the groups of intermittent smokers have much higher proportions of individuals with snuff consumption partly contradicts the notion that intermittent smokers may be less nicotine addicted than daily smokers. Snuff consumption is a specifically Swedish phenomenon. This may to some extent affect the estimations of the proportion of all intermittent smokers that are long term intermittent smokers compared to other western countries. Snuff consumption is comparatively high in Sweden compared to many other western countries, and the use of oral snuff is a very particular trait that characterises snuff consumption in Sweden. However, the vast majority (approximately 90%) of the intermittent smokers in all the three long term and transitional intermittent smoker groups do not use snuff, a fact that indicates that no far reaching inferences should be drawn from these findings.

CONCLUSION

The majority of intermittent smokers are long term intermittent smokers. The study results indicate that long term intermittent smokers have both sociodemographic and psychosocial characteristics—that is, high levels of social participation—that significantly differ from both transitional intermittent smokers that initiate daily smoking and the reference population. The results further support the notion that long term intermittent smokers are a specific group of smokers with other characteristics than daily smokers.