Tobacco Chewing and Adult Mortality: a Case-control Analysis of 22,000 Cases and 429,000 Controls, Never Smoking Tobacco and Never Drinking Alcohol, in South India

Tobacco consumption in any form, smoking or smokeless, is a major source of premature mortality. About 20% of global tobacco related mortality occurs in India (World Health Organization, 1999). Globally, among adults, 27% of males and 9% of females use smokeless tobacco tobacco (Global progress report 2010). Among those aged 15 or older, 28% of males and 12% of females were earlier found to use smokeless tobacco. The higher prevalence of tobacco chewing is noted less educated group, in rural population than in urban and in women compared to men (Gajalakshmi et al., 2012; Thakpur et al., 2013). For this study, we have used data from two large household surveys conducted in South India. One is a retrospective enquiry of the habits of 80,000 adults (48,000 urban, 32,000 rural) who had died a few years earlier (cases) (Gajalakshmi and Peto, 2004). The other is the baseline survey for an ongoing prospective enquiry of 600,000 adults (500,000 urban, 100,000 rural) in the same locations (controls). Follow-up information from


Introduction
Tobacco consumption in any form, smoking or smokeless, is a major source of premature mortality. About 20% of global tobacco related mortality occurs in India (World Health Organization, 1999). Globally, among adults, 27% of males and 9% of females use smokeless tobacco tobacco (Global progress report 2010). Among those aged 15 or older, 28% of males and 12% of females were earlier found to use smokeless tobacco. The higher prevalence of tobacco chewing is noted less educated group, in rural population than in urban and in women compared to men (Gajalakshmi et al., 2012;Thakpur et al., 2013).
For this study, we have used data from two large household surveys conducted in South India. One is a retrospective enquiry of the habits of 80,000 adults (48,000 urban, 32,000 rural) who had died a few years earlier (cases) (Gajalakshmi and Peto, 2004). The other is the baseline survey for an ongoing prospective enquiry of 600,000 adults (500,000 urban, 100,000 rural) in the same locations (controls). Follow-up information from RESEARCH ARTICLE Tobacco Chewing and Adult Mortality: a Case-control Analysis of 22,000 Cases and 429,000 Controls, Never Smoking Tobacco and Never Drinking Alcohol, in South India Vendhan Gajalakshmi*, Vendhan Kanimozhi the prospective study is not yet available, but those in the baseline survey can be used as controls for the cases in the survey of adult deaths.
Our aim is to assess the relationship, causal or otherwise, of chewing tobacco with mortality at ages 35-69, after excluding from cases and controls who had ever smoked tobacco or drunk alcohol.

Materials and Methods
The study took place in the late 1990s in the state of Tamil Nadu in South India. There were two study areas, one urban (the city of Chennai, which is the state capital, with a population of 4 million) and one rural (the district of Villupuram, with a population of 2.5 million in about 2000 villages). In both, efforts were made to identify deaths that occurred at the age of 25 and over in urban area during 1995-97 and at the age of >1 day in rural area during 1997-98. Made arrangements to visit the home of the dead person(cases for the study) to get information from the family about the educational level, circumstances that led up to the death, smoking, chewing tobacco and 1202 drinking (alcohol) habits of the dead person before they became ill. The controls were from our surveys of the general population at ages 35 years or above in the same two study areas (from were cases were recruited for the study) during 1998-2001.

Cases for the study
Urban cases: interviewed 48,000 families: Registration of the fact of death is almost complete in the city of Chennai. Necessary data were abstracted from the death registers in the Vital Statics Department (VSD) in Chennai city to locate the home of the deceased. Since cause of death stated on the death certificate was non-specific in about 50% of certificates, verbal autopsy(VA) was done, to arrive at underlying cause of death, at the time of visiting home of the deceased to collect required data for the study. The details of the novel verbal autopsy methodology developed and used in Tamil Nadu study was dealt somewhere else (Gajalakshmi et al., 2002;. This VA methodology consists of training non-medical graduates with at least 15 years of formal education on VA tool to interview the spouse, and/or close associates, and/or neighbours of the deceased and to write the verbal autopsy report. The verbal autopsy report is a narrative description of symptoms and events that led to death and written in local language to enhance the accuracy of the underlying cause of death . All VA reports were reviewed centrally and independently by two medical doctors, unaware of risk factor(s) data, to arrive at the probable underlying cause of death and coded the underlying cause of death according to the 9th International Classification of Diseases (ICD-9) (World Health Organization, 1977).
There were 72,000 deaths among adults aged 25 or older at the time of death during the study period 1995-97 in Chennai city. Of these, 5000 deaths were attributed to external causes (unintentional injuries, suicide or homicide) and 67,000 deaths to medical causes in the VSD records. Deaths due to external causes were excluded in the study. We were successful in tracing and interviewing 48,000 out of 67,000 households during 1998-1999. 19000 houses could not be visited because the address was missing or inadequate, the house no longer existed, or the family had moved. The cause of death of 1000 deaths that were attributed to medical causes in the VSD records were reclassified to external causes based on the VA report diagnosis, hence excluded from the study. Finally we left with 47,000 cases (27,000 men and 20,000 women) aged 25 or over at the time of death for the urban study. Of this, 5206 men and 8260 women were lifelong non-smoking non-drinkers of age 35-69 at the time of death.
Rural cases: interviewed 32,000 families: The registration of the fact of death is less than 60% complete in rural Tamil Nadu. Hence efforts were made to identify deaths at all ages (except deaths at age ≤1 day) irrespective of cause of death during 1997-98 in the study area from various sources such as records in the Village Administrative Offices in the study district, enquiring village health nurses /health care workers and village leaders in the study area. Field interviewers were natives of the study area and sought the help of the village leaders and/or village health care workers in obtaining introductions to the relatives, neighbours, or associates of the deceased. Verbal autopsy methodology and, the method of assigning and coding the probable underlying cause of death were as in urban case-control study (Gajalakshmi et al., 2002;. The total number of deaths identified during 1997-98 was 40 763 . Of these, 1927 deaths could not be traced because the addresses were missing/ incomplete in the Village Administrative Offices or the occupants had moved out after death. Of the 38 836 households traced and interviewed, 27,000 deaths (cases: 16,000 men and 11,000 women) were due to medical causes and 5000 were due to external causes at ages 25 or older. Of this, 3278 men and 5716 women were lifelong non-smoking non-drinkers of age 35-69 at the time of death.

Controls for the study are from population surveys conducted in urban (Chennai city) and rural (Villupuram) study areas in Tamil Nadu
Urban population survey: 500,800 individuals: A population survey (Gajalakshmi et al., 2007) was undertaken in Chennai city during 1998-2001. The men and women aged 35 years or over residing in the randomly chosen study area were interviewed at home. Precautions similar to those in the case-control studies were taken to ensure strict quality control of fieldwork, coding, and data entry. Details on the following variables were collected in this population survey: age, sex, educational status, tobacco smoking, tobacco chewing and alcohol drinking. Of 500,816 interviewed, 138,928 men and 219,241 women were lifelong non-smokers and non-drinkers aged 35-69 at the time of baseline survey.
Rural population survey: 100,000: A similar population survey was performed at the same time (1998)(1999)(2000)(2001) in the rural study area. Interviewed all people aged 35 years or over resident in seven of the 22 rural administrative blocks that make up the study area. Of 105,837 interviewed, 23,254 men and 47,883 women were lifelong non-smokers and non-drinkers aged 35-69 at the time of baseline survey.

Quality control
The survey reports and verbal autopsy reports submitted by the field Interviewers were validated by selecting randomly 5% of the households for re-interview by the senior investigator. This was done one week after receiving the output from the field Interviewers and blind to its results. The random checking was done partly because knowledge that a revisit might well take place would ensure reliably motivated fieldwork at the initial survey and to identify any systematic defect in the interview techniques. The questionnaires were checked centrally for consistency and missing values by coding clerks, and were double-entered into the computer.

Statistical methods
In this study chewing was defined as daily chewing of tobacco or related tobacco products, either alone or in combination, for at least 6 months. The term chewer was consistently used to mean ever chewer (former +current).
Asian Pacific Journal of Cancer Prevention, Vol 16, 2015 1203 DOI:http://dx.doi.org/10.7314/APJCP.2015.16.3.1201 Tobacco Chewing andAdult Mortality in Never Smoking Non Alcohol Drinkers in South India This study uses the controls from the population surveys in urban and rural areas since their exposure distribution should have been reasonably representative of the population at risk of becoming cases. In the general population, chewing habits may be associated with smoking and drinking habits. As the risk of chewing is likely to be much smaller than those of smoking and drinking, it is easiest to study them unbiasedly in people who never smoke or drink alcohol. Because the number of cases and of controls, in this study ,is so large, it is possible to restrict attention only to those who never smoked tobacco or drank alcohol.
The urban and rural case-control studies were analysed separately as well as combined. For each category of disease the cases were those who died of it and were compared with population controls; for each different category of disease the control group was always the same. Logistic regression models in STATA (version 8) statistical software (Statacorp LP 2005) were used to calculate mortality odds ratios. The excess death caused by chewing tobacco was calculated by multiplying over all number of deaths among chewers by 1-1/RR, in which RR is the adjusted mortality odds ratio.

Results
Detailed analyses of mortality among lifelong nonsmoking non-drinkers are reported here at the age range of 35-69 years because the underlying cause of death assigned by verbal autopsy is more reliable in middle age (35-69) than at older age (70+). A total of 22 460 cases (urban: 5206 men and 8260 women; rural: 3278 men and 5716 women) and 429 306 controls (urban: 138 928 men and 219 241 women; rural 23 254 men and 47 883 women) were analysed. Table 1 shows key characteristics of 22,460 cases and 429 306 controls. The cases were on average 9-11 years older than the controls. A higher proportion of women compared to men did not have formal education in both urban and rural study areas and a higher proportion of participants, both men and women, had no formal education in rural area compared to those in urban area. A higher proportion of cases compared to controls in both sexes and in both study areas were ever chewers of tobacco.

Discussion
Tobacco chewing is not considered as stigma in India. Hence we do not expect any misclassification of tobacco chewing habit in this study, even though the data were collected on dead people from the surviving family members. Since the analyses excluded smokers and drinkers the associations seen in this study could not be confounded by tobacco smoking or alcohol drinking, but they could, despite adjustment for educational level, be residually confounded by social factors.
The present study shows that ever-chewer mortality odds ratio adjusted for age and education for tuberculosis was 1.5-fold in rural men, 2-fold in rural and in urban women compared to never chewers. When both study areas were combined the age, sex, education and study area adjusted ever-chewer mortality odds ratio for tuberculosis was 1.7(1.5-1.9) and for respiratory diseases other than tuberculosis was 1.4(1.2-1.6). The present study results are consistent with the Cancer Prevention Study I in USA (Henley et al., 2005) that observed elevated hazard ratios among smokeless tobacco users for respiratory diseases combined (HR:1.28, 95% CI:1.03-1.59) and the Mumbai cohort study (Gupta et al., 2005) in India that noted about 40-46% higher risk for death from tuberculosis among tobacco chewers in both genders. The reason for this increased risk among tobacco chewers for respiratory diseases, including tuberculosis, is not clear.
In the urban study, cause of death due to cancer based on VA reports was confirmed with medical records (Gajalakshmi et al., 2004) and this was not feasible in rural study area. Among tobacco chewers the risk of death from cancer was about 40-50% higher in men and 60-70% higher in women compared to never chewers. The age and education adjusted mortality odds ratio associated with tobacco chewing was significant for deaths from upper aerodigestive, stomach and cervical cancers. Of the cancers mentioned above, except cervical cancer (McCann et al., 1992;Gajalakshmi et al., 2012), cancers in other sites are well known to be caused by chewing tobacco (Stockwell and Lyman, 1986;Sankaranarayanan et al., 1989;Gupta et al., 1980;Rao et al., 1994;Dikshit and Kanhere, 2000;Balaram et al., 2002;Znaor et al., 2003;;IARC Monographs 2004Gupta et al., 2005;Henley et al., 2005;Phukan et al., 2005;Razmara et al., 2013).
Noted increased heart rate and high blood pressure in tobacco chewers (Stockwell and Lyman, 1986;Bolinder and de Faire, 1998) and high levels of total cholesterol, low-density lipoprotein cholesterol and triglycerides in tobacco smokers and in tobacco chewers compared to never users of tobacco (Nanda and Sharma, 1988). A study conducted in Sweden (Bolinder et al., 1994) among men noted the relative risk of 1.4(1.2-1.6) for cardiovascular diseases in smokeless tobacco users. Mumbai cohort study (Gupta et al., 2005) in India found elevated risk of death from vascular diseases among women tobacco chewers only. Both in CPS I and II elevated risk of death from cardiovascular disease was found among smokeless tobacco users (CPS-I: Hazard ratio(HR):1.18, 95%CI:1.11-1.26 and CPS-II:HR:1.23, 95%CI:1.09-1.39) (Henley et al., 2005). The present study shows higher age and education adjusted mortality risk, of death from stroke among tobacco chewers compared to never chewers in rural men 2.2(1.6-3.0) , in rural women 1.3(1.0-1.6) and in urban women 1. 3(1.1-1.7).However, the mortality odds ratio (men, women and study areas combined) adjusted for age, sex, education and study area for stroke was 1.4(1.2-1.6) and for vascular diseases 1.1(1.0-1.2).
The strengths of the study are large sample size, inclusion of all deaths that occurred in the study areas, having used novel verbal autopsy method (with strict supervision and quality control) to assign the cause of death for all cases(deaths), use of population controls from general population surveys conducted in the areas from where cases were recruited for the study, and exclusion of smokers and alcohol drinkers from the analyses to avoid confounding by tobacco smoking and drinking alcohol.
In conclusion, the present study is the first large study in India on lifelong non-smoking non-drinkers at ages 35-69. The age, sex, education and study area adjusted mortality odds ratio was about 30% higher in ever-chewers in south India and the risk is higher among those in rural area compared to those in urban area. Chewing tobacco is the cause of 7.1% (n=1595) of deaths from all medical causes among non-smoking non-drinkers at ages 35-69 in south India. Of the cancers, ever-chewer mortality odds ratio adjusted for age and education was significant for upper aerodigestive cancers in urban men, urban women and in rural women, stomach cancer in rural men and urban women and, cervical cancer in both rural and urban women. The reason for increased risk among ever tobacco chewers for stroke and respiratory diseases including tuberculosis found in this study are not clear.