Based on comprehensive data from 10,837 middle-aged and elderly Chinese rural individuals, this cross-sectional study provided evidence that current smokers tended to have a lower BMI than non-smokers and thus were less likely to be overweight or obese, but the relationship might be U-shaped between amount of smoking and risk of being overweight or obese. Interestingly, individuals with normal weight would have the lowest levels of smoking behaviours among the four BMI categories, including the oldest age at starting smoking, the lowest number of cigarettes smoked per day, the shortest duration of smoking and the minimum cumulative amount of smoking. Additionally, this study also found that smokers, especially heavy smokers, were more likely to be drinkers and eat less when compared with non-smokers, whereas alcohol drinking behaviours were positively associated with the risk of being overweight or obese. These results highlighted the roles of alcohol drinking behaviour and dietary factors in influencing the relationship between smoking status and BMI, which could have implications for public health interventions aiming at reducing the burdens from weight issues and harmful use of alcohol and tobacco.
Relation to published evidence
A previous epidemiological study also found that young American adults aged less than 30 years who were trying to lose weight were more likely to smoke[29]. However, the mechanism that smoking is associated with reduced body weight is still complex and remains incompletely understood. Basically, cigarette smoking is often thought to have appetite-suppressing actions and therefore control weight[12, 10]. As shown in figure 3, it is obvious that smokers tended to eat less than non-smokers, including most foods of vegetables, fruits, nuts, cereals, bean food, animal product, fried food, pickled food and food made with yeast. However, they may consume more coriander, onion, garlic, hawthorn and fermented bean curd, which are generally considered as the flavorants or foods with a strong flavour having positive effects on appetite or digestion [30-32]. Therefore, it can be suggested that cigarette smoking may negatively influence subjects' appetite, and result in lower consumption of food and consequent reduced body weight. Previous study held that on the one hand, cigarette smoking might serve as a behavioral alternative to eating, and on the other hand, nicotine in cigarettes may increase the resting metabolic rate while controlling the expected growth in food consumption in response to the growth in metabolic rate, which would tip the balance of caloric intake and daily energy expenditure[10]. Interestingly, the results shown in table 2 first reports that subjects with normal weight would have the lowest levels of smoking behaviours among the four BMI categories, which suggests that normal weight individuals may be more likely to have a relatively healthier lifestyle. In other words, subjects rarely or never smoked may be typically more likely to be within a normal BMI range than regular or heavy smokers, because they were less influenced by weight loss or weight gain caused by smoking or smoking cessation.
Among subjects a U-shaped relationship between the number of cigarettes smoked per day/cumulative amount of smoking and the OR for being overweight or obese can be found in figure 1, with 10-˂20 cigarettes smoked per day and 15-˂30 pack-years being the lowest ORs (except the one which aged 51-≤60 and had 15-˂30 pack-years). This finding is exactly consistent with several early studies, which have concluded that individuals smoked about 10-20 cigarettes per day were the leanest, with the trend being similar by stratum of baseline age groups[3, 33-40, 4]. Therefore, studies have reported that there is not just simply negative relationship between smoking amount and BMI or body weight, whereas the observed U-shaped relationship is by no means coincidental. Although this may seem paradoxical in consideration of the metabolic effects of smoking, it has been assumed that heavy smokers may be more likely to have some other unhealthy lifestyles and habits such as heavy alcohol use[4].
In order to verify the above hypothesis, the associations between alcohol use status and risk of being overweight or obese, and between smoking behaviour and alcohol drinking were also assessed. Adjusted results in table 3 illustrate strong positive associations between alcohol drinking behaviour parameters and risk of being overweight or obese, with the statistically significant increasing trends found in alcohol units consumed per day, duration of drinking, earlier age at starting drinking, as well as cumulative amount of drinking. Thus, alcohol use is likely to be a risk factor of overweight and obesity in this middle-aged and elderly Chinese rural population with a significant dose-response relationship. An updated research summarized cross-sectional, longitudinal and experimental studies examining the link between alcohol consumption and obesity also concluded that it is reasonable to say that use of alcohol may be a risk factor for obesity in some populations, likely based on a multitude of factors[41]. Furthermore, table 4 shows that heavy smokers were much more likely to be drinkers. Compared with non-smokers, heavy smokers smoked more than 20 cigarettes, experienced more than 35 years of smoking history, started smoking before 25 years old and had more than 30 pack-years of cumulative amount smoking were significantly associated with 4.88, 4.89, 4.96 and 5.17 times increased risk of becoming a drinker, respectively. By integrating the two parts of results, it is logical to draw that heavy smokers and smokers starting smoking early may be more likely to have the habit of alcohol drinking than light smokers, and past a certain point, the increase in consumption of alcohol and its related high-fat diet, which is highly calorific, offsets the appetite suppression effects of smoking, and this is what leads to the high BMI and obesity levels in heavy smokers, and explains the U-shaped relationship between smoking behaviours and BMI. In addition, heavy drinking is usually accompanied by unhealthy dietary behaviour in China such as high intake of saturated fat and calories, because residents may consume alcoholic drinks frequently when having dinner or midnight snack together.
Strengths and limitations
A strength of the study was that it sampled a relatively unstudied demographic, and additionally, that demographic was one where potential variables were relatively stable, which means the sample group has the additional advantage of being relatively similar, thereby mitigating the impact of variables such as socio-economic status and diet, and improving the accuracy of the result. Furthermore, this study not only analysed the association between smoking and BMI, but also considered the roles of dietary factors and alcohol consumption, which may partly explain the U-shaped relationship between smoking and BMI: up to a certain point, smoking is associated with a decreased BMI, but beyond this, heavy smoking is associated with increased BMI and obesity.
Our study was limited by the possible recall bias brought by the questionnaire. The questionnaire collected the data of food frequency, but the information of portion size was unable to be obtained. In addition, this study was focused on current smokers, which means former smokers were excluded and not analysed. Furthermore, although dietary factors and alcohol use were taken into consideration, and several potential confounders were adjusted in the multivariate analyses, smokers and non-smokers may still differ with respect to other factors such as physical exercise. It should also be pointed out that cross-sectional study has limitations in establishing a causal association.