|
|
| |||||||||||||||||||||||||||||||||||||||||
Background The nationwide effects of smoking on mortality in India have not been assessed reliably.
Methods In a nationally representative sample of 1.1 million homes, we compared the prevalence of smoking among 33,000 deceased women and 41,000 deceased men (case subjects) with the prevalence of smoking among 35,000 living women and 43,000 living men (unmatched control subjects). Mortality risk ratios comparing smokers with nonsmokers were adjusted for age, educational level, and use of alcohol.
Results About 5% of female control subjects and 37% of male control subjects between the ages of 30 and 69 years were smokers. In this age group, smoking was associated with an increased risk of death from any medical cause among both women (risk ratio, 2.0; 99% confidence interval [CI], 1.8 to 2.3) and men (risk ratio, 1.7; 99% CI, 1.6 to 1.8). Daily smoking of even a small amount of tobacco was associated with increased mortality. Excess deaths among smokers, as compared with nonsmokers, were chiefly from tuberculosis among both women (risk ratio, 3.0; 99% CI, 2.4 to 3.9) and men (risk ratio, 2.3; 99% CI, 2.1 to 2.6) and from respiratory, vascular, or neoplastic disease. Smoking was associated with a reduction in median survival of 8 years for women (99% CI, 5 to 11) and 6 years for men (99% CI, 5 to 7). If these associations are mainly causal, smoking in persons between the ages of 30 and 69 years is responsible for about 1 in 20 deaths of women and 1 in 5 deaths of men. In 2010, smoking will cause about 930,000 adult deaths in India; of the dead, about 70% (90,000 women and 580,000 men) will be between the ages of 30 and 69 years. Because of population growth, the absolute number of deaths in this age group is rising by about 3% per year.
Conclusions Smoking causes a large and growing number of premature deaths in India.
The effects of prolonged smoking of bidis or cigarettes on mortality in India have been assessed reliably in only two specific localities5,6,7,8,9 in which the numbers of female smokers were too small to study. To assess the hazards of smoking in India nationwide among both women and men, we have conducted a case–control study that collected information on all adult deaths from 2001 to 2003 in a nationally representative sample of 1.1 million homes.
Methods
Study Design
Details of the study sample, case and control definitions, assessment of exposures, assignment of the underlying causes of death, and statistical methods are in the Supplementary Appendix (available with the full text of this article at www.nejm.org). In brief, the study was conducted in 1.1 million homes in 6671 small areas chosen randomly from all parts of India (about 1000 persons per area); the Sample Registration System was established by the Registrar General of India to monitor all births and deaths in these areas.10,11 Each home in which a death had been recorded between 2001 and 2003 was visited by 1 of 900 nonmedical field-workers to collect information about the cause of death, the history of tobacco and alcohol use, and educational status. The underlying causes of all deaths were sought by verbal autopsy (a structured investigation of events leading to the death).6,12,13,14 Two trained physicians independently assigned codes to the causes of death on the basis of the nonmedical field-worker's written narrative of the death. If the two physicians did not agree on the assigned three-digit code from the International Classification of Diseases, 10th Revision,15 a senior physician adjudicated. A random sample of about 10% of the areas was resurveyed independently, generally with consistent results. Details of the methods, quality-control checks, and validation results have been reported previously.12,13,14,16,17
The field teams asked respondents (typically, household members) whether the deceased person had been a smoker within the previous 5 years and, if so, the usual number of bidis or cigarettes the person had smoked per day. Since smoking cessation is uncommon in India,5,18 the key comparisons were between persons who had smoked in the past 5 years and those who had not. Questions were also asked about other tobacco smoking, quid chewing, alcohol consumption, and years of education. Adult respondents were asked similar questions about themselves.
Subjects
Potential case subjects were adults 20 years of age or older who had died between 2001 and 2003 and whose deaths had been recorded in the earlier fieldwork of the Sample Registration System. Among case subjects, data were available for 33,069 women and 41,054 men after the exclusion of deaths for nonmedical or maternal causes and of deaths for which data regarding smoking status were missing (3%). Adults who provided information about someone who had died were also asked about their own smoking status, and 97% of them provided a response (34,857 women and 43,078 men). This population of respondents served as the control group. The analyses do not match particular case subjects with particular control subjects; therefore, the study design does not adjust for household.
Women and men were analyzed separately. In the major analyses, we used logistic regression to adjust for age, educational level, and use or nonuse of alcohol.
Calculation of Absolute Risk
In the calculation of absolute risk, we used the World Health Organization (WHO) age-specific death rates for India to correct for any slight undercounts in the Sample Registration System.19,20 Deaths of persons between the ages of 30 and 69 years are presented separately because deaths at these ages involve substantially more years of life lost than do deaths at older ages. In addition, the assignment of underlying causes of death is substantially more reliable for persons between the ages of 30 and 69 years than for older persons,12,14,16,17 and the main effects of smoking on mortality occur after the age of 25 years (the approximate median age at which smoking begins among men in India4).
Results
Characteristics of the Subjects
Persons who died were older and less educated and had a higher prevalence of smoking, tobacco chewing, and alcohol use than did living control subjects (Table 1, and Table 1 of the Supplementary Appendix). In the group of control subjects, the prevalence of smoking among men rose with age, from 8% among those between 15 and 19 years to 27% among those between 30 and 34 years. The prevalence then remained approximately constant at 35 to 40% for subjects between the ages of 35 and 69 years (Figure 1, and Table 2 of the Supplementary Appendix). Between the ages of 40 and 59 years, smoking was more common among men without primary education (44%) than among other men (35%). The age-specific prevalence of smoking among female control subjects between 30 and 69 years of age rose fairly steadily from 3 to 6%. Between the ages of 40 and 59 years, the prevalence of smoking was 6% for women without primary education and 2% for other women.
|
|
For women between the ages of 30 and 69 years, 9% of those who died from medical causes were smokers; 5% of control subjects were smokers (Table 2). The mortality risk ratio comparing smokers with nonsmokers was 2.0 (99% confidence interval [CI], 1.8 to 2.3), after adjustment for age, educational level, and use or nonuse of alcohol. This mortality risk ratio corresponds to an excess of 783 deaths among female smokers between the ages of 30 and 69 years, constituting 5% (99% CI, 4 to 6) of deaths from any medical cause in the study. Since further adjustment for status with respect to tobacco chewing, residence (urban or rural), and religion did not substantially alter the risk ratios, all subsequent analyses were adjusted only for age, educational level, and use or nonuse of alcohol (see the Supplementary Appendix). For women 70 years of age or older, the risk ratio was lower than that for women between the ages of 30 and 69 years, and the difference was not significant (risk ratio, 1.3; 99% CI, 0.9 to 1.7).
|
Underlying Causes of Death
Among women between the ages of 30 and 69 years, smokers accounted for 13% of those who died from tuberculosis (risk ratio, 3.0, 99% CI, 2.4 to 3.9) and 14% of those who died from respiratory disease (risk ratio, 3.1; 99% CI, 2.5 to 3.8). Among the 783 excess deaths of women that were associated with smoking, tuberculosis accounted for 127 (16%) and respiratory disease accounted for 221 (28%); the proportions were lower for stroke (6%) and heart disease (10%).
Among men in this age group, smokers accounted for 66% of those who died from tuberculosis (risk ratio, 2.3; 99% CI, 2.1 to 2.6) and 60% of those who died from respiratory disease (risk ratio, 2.1; 99% CI, 1.9 to 2.3). Thus, of the 5751 smoking-related excess deaths from medical causes among men between the ages of 30 and 69 years, 1174 (20%) were from tuberculosis and 1078 (19%) were from respiratory disease. The risk ratio for death from stroke among men was 1.6 (99% CI, 1.4 to 1.8) and that for death from heart disease was also 1.6 (99% CI, 1.5 to 1.8), with deaths from heart disease accounting for 1102 of the smoking-associated excess deaths (19%) in men in this age group. Women and men who had smoked also had an increased risk of death from neoplastic disease and peptic ulcer.
Relative Risk in Subgroups
Among women, the absolute number of deaths associated with smoking was too small to be statistically reliable after stratification for location of residence (urban or rural), educational level, use or nonuse of alcohol, and level of tobacco use, even though no anomalies were apparent (Figure 2, and Fig. 1A and 1B of the Supplementary Appendix). Thus, the results for these subgroups are described here only for men.
|
In a comparison of smokers and nonsmokers, the relative risk of death from any medical cause in men between the ages of 30 and 69 years was similar among those without primary education (1.6), those with primary or middle school education (1.7), and those with secondary or post-secondary education (1.7); the relative risk was the same among users of alcohol and nonusers of alcohol (1.6).
Most male subjects smoked only bidis, but some smoked only cigarettes, and there was a dose–response relationship between smoking and mortality among men who smoked only bidis (risk ratio for one to seven bidis per day, 1.3; risk ratio for eight or more, 2.2) and among those who smoked only cigarettes (risk ratio for one to seven cigarettes per day, 1.8; risk ratio for eight or more, 2.9; P<0.001 for both trends) (Table 3 of the Supplementary Appendix). Even among those who smoked only one to seven bidis per day (mean, four per day), the smoking-associated excess deaths accounted for a quarter of all deaths from any medical cause (risk ratio, 1.3). Among those who smoked only one to seven cigarettes per day (mean, four per day), smoking-associated excess deaths accounted for almost half of deaths from any medical cause (risk ratio, 1.8).
Absolute Risk
To help estimate absolute hazards, we used WHO estimates of Indian national death rates and annual numbers of deaths, subdivided according to 5-year age groups. Within each age group, our study provides the proportion of deaths from any medical cause; the relative risk of death from any medical cause among smokers, as compared with nonsmokers; and the prevalence of smoking among control subjects. The combination of these data with the WHO data yields for each age group the death rates for smokers and nonsmokers (adjusted for any differences in educational level, alcohol use, or death from nonmedical causes) and the number of deaths that would have been avoided if smokers had had the same death rates as nonsmokers of similar age, education, and use or nonuse of alcohol.
From these age-specific death rates among smokers and nonsmokers, we calculated the cumulative rate of death among subjects between the ages of 30 and 69 years (Figure 3). These rates suggest that about 62% of female smokers who were 30 years of age will die before the age of 70, as compared with only 38% of otherwise similar nonsmokers (absolute difference, 24 percentage points) (Figure 3A). A substantial hazard was evident even before the age of 50 years: 15% of female smokers will die between the ages of 30 and 49 years, as compared with only 7% of nonsmokers. The median ages at death suggested that female smokers will die an average of approximately 8 years (99% CI, 5 to 11) earlier than their nonsmoking counterparts. Likewise, about 61% of male smokers who were 30 years of age will die before the age of 70 years, as compared with 41% of otherwise similar male nonsmokers (absolute difference, 20%) (Figure 3B). The median ages at death suggested that male smokers will die an average of approximately 6 years (99% CI, 5 to 7) earlier than their nonsmoking counterparts. Again, a substantial hazard was evident before the age of 50 years: 15% of male smokers, as compared with 9% of male nonsmokers, will die between the ages of 30 and 49 years. Figure 2 of the Supplementary Appendix provides estimates of absolute risks of death for smokers and nonsmokers for three major causes of death: respiratory disease, heart disease, and tuberculosis.
|
Table 3 provides estimates of the numbers of adult deaths in India for 2010 and the numbers that would be avoided if smokers had the same rate of death as otherwise similar nonsmokers. These 2010 projections assume that as the population grows, the age-specific distributions of the underlying causes of death and of the proportions associated with smoking will be similar over the next few years to those suggested by this study. Among women between the ages of 30 and 69 years, the excess number of deaths associated with smoking will be approximately 5% (99% CI, 4 to 6) of all deaths. Among men in this age group, the excess will be 20% (99% CI, 19 to 22) of all deaths. This proportion differs from the 23% shown in Table 2 and Figure 2, since deaths from injuries were excluded from those denominators. The excess number of deaths among adult smokers in 2010 will be about 930,000, including 580,000 deaths among men and 90,000 deaths among women between the ages of 30 and 69 years. Because of population growth, the annual number of smoking-associated deaths among adults will be about 1 million during the 2010s.
|
In this large, nationally representative case–control study, we found that in both rural and urban India, among men between the ages of 30 and 69 years, the rate of death from any medical cause in smokers was 1.7 times that in nonsmokers of similar age, educational level, and alcohol status (use or nonuse). Among female smokers, mortality from any medical cause was double that among their nonsmoking counterparts. Most of the excess mortality was from tuberculosis or from respiratory, vascular, or neoplastic disease.
Although the hazard associated with any smoking (even only a few bidis a day) was substantial, the hazard of cigarette smoking (risk ratio, 2.2) was even higher and corresponded to more than a doubling of the risk of death in middle-aged men, suggesting about a 10-year gap in median survival between cigarette smokers and nonsmokers. Such risks are similar to those seen in a study in the United Kingdom in which most of the male smokers had been smoking substantial numbers of cigarettes since about 18 years of age.21 We could not study smoking cessation, since quitting is uncommon in India18 and often occurs only after a smoker becomes ill. However, studies elsewhere have shown that cessation can substantially reduce a smoker's risk.21,22
Although the assessment of smoking-associated risks for women is subject to more uncertainty, the risk of smoking among women appears to be at least as large as that for men. For both sexes, such extreme risks in India are surprising,7 since the age at which persons generally start smoking is older23 and the amount that is smoked per person is lower5 than in Europe or North America; India also appears to have a relatively low rate of lung cancer.9 Indeed, in our study, the number of smoking-associated deaths from tuberculosis was more than 10 times the number of smoking-associated deaths from lung cancer. These high overall risks reflect the high background rates of death from tuberculosis, respiratory disease, and heart disease among nonsmokers in India.
Our study has some limitations. First, differences between case subjects and control subjects, in factors other than the few variables that we measured, may have affected the relative risks. Second, tobacco use and alcohol use are strongly correlated, so residual confounding by the use of alcohol could explain some of the excess mortality among smokers. However, the relative risk of death from any medical cause among smokers, as compared with nonsmokers, was similar when analyses were restricted to those who did not drink alcohol. Third, verbal-autopsy methods misclassify the underlying causes of some deaths.12,13,14,16,17 Although a more precise classification of the underlying disease responsible for each death would probably sharpen the findings by increasing the risk ratios for some conditions and decreasing it for others, it would have little effect on the risk ratio for the aggregate of all diseases. Fourth, we did not have information on the smoking status of people who were homeless or living alone when they died; overall, we obtained information on smoking for only 88% of all enumerated deaths. However, persons who were homeless or living alone were also excluded from the control subjects. Finally, the tendency of smokers to live with other smokers could have inflated the rates of smoking among control subjects. In addition, control subjects were self-selected by their willingness to be interviewed. However, we had few refusals, and the age-specific prevalence of smoking among control subjects was reasonably similar to that in recent independent, nationally representative surveys.18,24 Differential reporting of the daily amount smoked between case subjects and control subjects could have distorted the dose–response relationships, but the overall hazards associated with smoking should not be increased, since it is unlikely that many deceased nonsmokers were misreported as having been smokers.
Previous studies of men in India6,7,8,9,25,26,27 and of men and women elsewhere21,28,29,30,31 have provided reasonably robust evidence that smoking can actually cause various types of respiratory, vascular, and neoplastic disease — in other words, that smoking increases the risk of developing such diseases in the near future among otherwise similar people of a given age. Hence, it is reasonable to conclude that the observed association between excess mortality and such diseases after adjustment for age and use or nonuse of alcohol was mostly causal, especially since the relative risks were not materially altered by adjustment for educational level or by restriction of the analyses to persons who did not consume alcohol. In contrast, the association between smoking and death from liver disease (which is greatly reduced after adjustment for use or nonuse of alcohol) is probably largely or wholly due to the tendency of habitual drinkers to smoke tobacco.
Tuberculosis is a special case because the tubercle bacillus is obviously a cause of all deaths from tuberculosis. Nevertheless, smoking could also be a cause of many of the deaths from tuberculosis. Subclinical infection is widespread,32,33 and smoking could facilitate progression to clinical disease.6 If so, smoking might also contribute to the spread of tuberculosis to others.
If the associations we observed in our study were mostly causal, we estimate that smoking will cause about 930,000 deaths in 2010 in India. Of these deaths, about 580,000 among men and 90,000 among women will occur between the ages of 30 and 69 years (about 1 in 20 deaths from any cause among women and 1 in 5 deaths from any cause among men in this age group). The excess mortality among smokers in India will account for much of the gap between male and female mortality in middle age.
Because of population growth, the absolute numbers of deaths are rising by about 3% per year among persons between the ages of 30 and 69 years and by a higher rate among persons who are older.2 During the 2010s, therefore, the annual number of deaths from smoking in India will be about 1 million, which is similar to the annual number in China.34 Of the million annual deaths from smoking in India, approximately 70% (100,000 among women and 600,000 among men) will occur in middle age, rather than old age.
Supported by grants from the John E. Fogarty International Center of the National Institutes of Health (R01-TW05991–01 and TW007939-01, to Dr. Jha; and R01-TW05993-02, to Dr. Peto); the Canadian International Development Research Centre (102172); the Canadian Institute of Health Research (IEG-53506); the Li Ka Shing Knowledge Institute and Keenan Research Centre (to Dr. Jha); Cancer Research UK and the United Kingdom Medical Research Council (to Dr. Peto); and the Canada Research chair program (to Dr. Jha).
No potential conflict of interest relevant to this article was reported.
The opinions expressed in this article are those of the authors and do not necessarily represent those of the Government of India or the Office of the Registrar General.
We thank the Office of the Registrar General for the productive collaboration on the Million Death Study; K.T. Shenoy for hosting a meeting in January 2004 to plan the survey of deaths; J. Moore for assistance with the Web-based physician-coding system; D. Corsi, J. Godwin, and W. Suraweera for their expert assistance with the tables and figures; C. Mathers and A.D. Lopez for providing WHO estimates of mortality; and D. Bassani, H. Gelband, and A.S. Slutsky for their helpful comments on the manuscript.
Source Information
From the Centre for Global Health Research, Toronto (P.J., B.J., R. Kamadod); and the Epidemiological Research Centre, Chennai (V.G.); Healis-Sekhsaria Institute for Public Health, Mumbai (P.C.G.); the Rural Health Training Centre, Najafgarh, New Delhi (N.D.); the School of Public Health, Post Graduate Institute of Medical Education and Research, Chandigarh (R. Kumar); the School of Preventative Oncology, Patna (D.N.S.); Tata Memorial Hospital, Mumbai (R.P.D.); and North Eastern Indira Gandhi Regional Institute of Health and Medical Sciences, Shillong (D.K.P.) — all in India; and the Clinical Trial Service Unit and Epidemiological Studies Unit, University of Oxford, Oxford, United Kingdom (J.B., R.P.).
This article (10.1056/NEJMsa0707719) was published at www.nejm.org on February 13, 2008.
Address reprint requests to Dr. Jha at the Centre for Global Health Research, St. Michael's Hospital, University of Toronto, Toronto, ON M5C 1N8, Canada, or at prabhat.jha{at}utoronto.ca.
References
| |||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
