Association between tobacco use and body mass index in urban Indian population: implications for public health in India

Body mass index [BMI, weight (kg)/height (m²)], a measure of weight adjusted for height, is a simple and inexpensive index that is often used as a proxy for overall health of populations. [1] Body habitus, as described by BMI, is related to skeletal size, muscle mass, and adiposity. As such, it is influenced by diet, other aspects of lifestyle, and other environmental factors. The association between low BMI and smoking is well documented. [2, 3] However, recent studies show that the nature of this relationship depends on educational level, gender, ethnicity, and frequency (e.g., number of cigarettes/day) of smoking. Often, the relationship changes with time. In Finland, for example, the inverse association between smoking and BMI weakened between 1982–1987 and became positive later. [4] Educational attainment is known to influence the relation between smoking and BMI in North European populations. [5, 6] In these populations, smoking is inversely associated with BMI at lower levels of education and positively associated with BMI at higher levels of education. A U-shaped relation between frequency of smoking and BMI has been observed by some investigators. [7]

In India, tobacco is used in various forms. [8] In addition to smoking cigarettes, bidis are commonly smoked, as they are much cheaper. Bidis are handmade by rolling a dried rectangular piece of temburni leaf (Diospyros melanoxylon) with 0.15–0.25 g of sun-dried, flaxed tobacco. In addition, use of smokeless tobacco, in a variety of forms, is widespread among both men and women. The most common form of smokeless tobacco use is mishri, a black powder obtained by roasting and powdering tobacco, which is then applied to the gums using a finger. Another common form is chewing of betel-quid, a combination of betel-leaf, areca nut, slaked lime, tobacco, and condiments; combinations of ingredients are altered according to individual preferences. Use of all forms of tobacco is associated with higher all-cause mortality in the Indian population. [9–11] Previously, we reported that tobacco using (smoking and smokeless) is associated with low BMI in an Indian population. [12] The high prevalence of tobacco use and its association with low BMI raises important questions about its impact on public health in India, a country which has a high prevalence of low BMI among adults. The focus of this paper is to provide a more detailed analysis of the relation between different forms of tobacco use and BMI and discuss the public health implications of these associations.

The data presented in this report were obtained from a baseline cross-sectional survey conducted between 1992–1994 for a cohort study on tobacco-attributable mortality. [13] The survey was carried out in the main city of Mumbai, formerly Bombay – the largest city in India. A sampling frame was constructed from the electoral rolls. The sampling unit was a 'polling station,' consisting of 1000 to 1500 eligible voters. Rolls were assumed complete, as they are updated before every major election through house-to-house visits. Electoral rolls were organised by geographic areas. The selection of polling stations was done in a non-random manner to exclude those with apartments having high security, as it became evident during the pilot study that it would not be possible to gain access to these apartments.

Investigators approached all individuals aged ≥ 35 years (cut off chosen because of the overall goal of studying tobacco-attributable mortality in the cohort) listed in the selected polling stations for interview and anthropometric measurements. Individuals not present on the voters' list also were interviewed and included in the sample if their residence status was confirmed by their having a 'ration card.' These cards, issued by the Bombay Municipal Corporation, act as a proxy for residence cards and permit access to all city and state government services (including receiving certain food items at subsidized prices). Such individuals comprised about 5% of the sample. Less than 1% of individuals approached refused for interview and/or allow anthropometric measurements to be taken. A total of 99,958 adults, 40,071 males and 59,527 females, were recruited and surveyed. The study satisfied all criteria of ethical treatment of human subjects; especially those formulated by the Indian Council of Medical Research.

The survey included two components: 1. measurement of height, weight, blood pressure; and 2. interviewer-administration of a structured questionnaire to obtain information on age, occupational history, education, religion, language, and tobacco-related behaviour. Weight was measured using a bathroom scale accurate to 0.5 kg. The scale was kept on a flat surface and the subject was requested to step on it in bare feet without holding on to anything. Subjects were measured in normal apparel, which in Mumbai is light cotton because of the tropical weather year around. The weight was recorded to the nearest kg. Height was measured using a specially constructed instrument consisting of a steel platform to which was attached a steel measure tape. With the subject standing erect on the steel platform, the tape was pulled vertically above the head, and then brought down to touch the flat ruler placed horizontally on the crown of the head. Height was recorded to the nearest cm.

All data were entered directly in a handheld computer in the field and transferred to a PC once a week in the Project office. Respondents were classified according to present and past tobacco use as (a) having never used tobacco, (b) ex-smoker, (c) ex-smokeless tobacco use, (d) ever smoker, (f) ever smokeless tobacco user, and (g) ever mixed habit (smokeless and smoking). Ever smokers and ever mixed users clubbed together were further divided into those who smoked i) cigarettes ii) bidis and iii) cigarettes + bidis. The smokeless-tobacco use further divided into -i) mishri ii) mishri + other forms of tobacco iii) betel quid iv) other tobacco and v) areca nut. For analysis of educational background, respondents were classified as (a) illiterate - received no education (b) primary - up to five years of education (c) middle - 6 to 8 years of education (d) secondary - 9 to 12 years of education (e) college - those who have received education past secondary level. Religion and mother tongue was used for analysis of cultural background. For mother tongue, the categories used were- Marathi, Hindi, Gujarati, Tamil, Urdu and other. For religion, the respondents were classified as Hindu, Muslim, Christian, Jew, Buddhist, and other.

Descriptive statistics (mean, standard deviation, and proportion within categories) were calculated for the total survey population and by tobacco use for both men and women separately. Because the number of ex smokers and cigarette smokers among women was small, separate analyses were not conducted on these two categories.

Multivariable analysis was performed using logistic regression. The response variable, BMI, was converted into a dichotomous variable by using two cut points: 18.5 and 25.0 kg/m². These are the conventional cut points indicating underweight and overweight, respectively. Three possible models corresponding to these two cut-off points were fit. Age (in 5-year age groups), education, mother tongue, religion and tobacco use were fit as independent variables in the final model. For dose response, the number of times (per day) tobacco was smoked, chewed or applied was refereed as frequency of habit per day and grouped as 1 to 5/day, 6 to 10/day and ≥ 11/day.

On comparing adjusted odds ratios with unadjusted odds ratios (data not shown), the adjusted odds ratios were always smaller indicating that controlled variables were confounders. The differences however, were not large; the highest reduction in odds ratio was from 2.79 to 2.19 for women with BMI < 16.0 and frequency of smokeless tobacco use > 10 per day; suggesting that there was little possibility of residual confounding affecting the results.

In India, nearly half of all rural adults and a quarter of urban adults have a low BMI (i.e., <18.5 kg/m²). [14]. Although, chronic energy deficiency due to inadequate diet may be the main factor placing the population at risk of low BMI, factors other than diet may play a significant role in explaining the low BMI within this population. These factors may act directly (by affecting appetite or other aspects of physiology) or indirectly (by decreasing the purchasing power for food). From a public health perspective, unmasking these non-dietary determinants of low BMI in the population will help in understanding the impact of such exposures. The present study has found that all forms of tobacco use are associated with low BMI independent of (i.e., after accounting for) age, education, mother tongue, and religion in this population. Further, there exists a dose-response gradient and the response at every dose was higher in women compared with men, a relationship observed in many tobacco-health disease analysis, including cancer related end points [15, 16]. This finding raises important questions about the magnitude of the adverse impact of tobacco use on the health status of the population.

The effects of tobacco use on the incidence of certain diseases, particularly cancers of the aerodigestive tract and urinary bladder are well documented. The findings of this study that all forms of tobacco use are associated with low BMI (a proxy for nutritional status) suggest a strong need for further research on as tobacco use may have even more far-reaching public health implications in India than previously thought. If the association is evaluated as causal then tobacco control research and intervention also will benefit other public health goals on improved nutritional status and consequential health benefits,. These results have potential to affect the population living in the developing world.