Table 17. Continued Reg;ion Survey akl Country Year Central America' 1986 Sample area Number Belize National Drug Use Survey 12,595 Costa Rica Honduras 1984 1986 Nicaragua 1988 Panama 1984 1989 Mexico 1989 1988 1980 San Jose Preuniversity students 487 15-20 Calderon et al. 17 10 13 694 15-30 National University 29 4 17 High school students 468 in Managua Nationwide 11,383 Private college 464 Secondary students 9,967 First year 88,735 university students Mexico City sec- 3,408 ondary students Latin Caribbean++ Cuba 1988 Nationwide 1,067 Selected Caribbean countries Bahamas 1987 Areawide 4,838 Out-of-school 74 youths In-school youths 4,767 Cayman 1985 Areawide 2,077 Islands French Guiana 1986 Areawide Jamaica 1987 Secondary students Suriname 1988 Seven cities and rural areas Aruba and 1988 Aruba Netherlands Antilles Trinidad and 1985 All secondary Tobago students U.S. Virgin 1988 Household Islands Behavioral Risk Factor Survey `Given for current daily smokers. `Smoked during the previous year. *Ever smoked. "Smoked during the previous month. 2,192 Age Sponsor lo-20 Pride Belize 15-18 University of Nicaragua 1 l-l 8 National Cancer Association 15-19 Department of Health National University Mexican Insitute of Psychiatry 473 13-17 Consumer Institute 8 3 6 United Nations Fund for Drug Abuse 20$ 103 32$ 10-l 7 Drug Advisory Committee 11-13 1 l-21 National Council on Drug Abuse 13-21 24 12 Prevalence* (%) Men Women Total 12$ 40 52 46 10 4 7** 3 3 6 6 42$ 9+ 15$ 23$ 2% 40 19 29$ 7 3 5" 36 12 23 12 17+ 1 l-19 Trinidad and Tobago Government Drug Survey 12-17 U.S. Virgin Islands Government _Excludes El Salvador and Guatemala. **Smoked during the previous week. `+Excludes Dominican Republic, Haiti, and Puerto Rico. "SOccasional smoker. Table 18. Prevalence of smoking among women of childbearing age, selected Latin American and Caribbean countries, 1979-1987 - Survev 1 Number Sponsor Prevalence (%) Country Argentina Brazil Chile Colombia Costa Rica Ecuador Guatemala Mexico Panama Paraguay Puerto Rico Suriname Uruguay Venezuela Year Sample area 1987 Nationwide 1981 Southern Brazil CDCS 1982 Piaui State CDC 1982 Amazonas State CDC 1987 Nationwide CLAP 1983 Santiago 1987 Nationwide 1986 Nationwide 1987 Nationwide 1983 Nationwide 1987 Nationwide 1979 U.S. border 1987 Nationwide 986 1987 Nationwide 1,935 1982 Entire territory 1985 Urban areas 1987 Nationwide 1987 Nationwide 980 CLAP 4,605 CLAP* 1,480 CLAP 2,009 4,187 CLAP EP CDC CLAP CLAP 5,169 CLAP Source: Pan American Health Organization (1992). `Centro Latinoamericano de Perinatologia y Desarrollo Human0 de la Organizacibn Panamericana de Salud. %ix months before pregnancy. *enters for Disease Control. BBefore pregnancy/during pregnancy. Table 19. Public knowledge and attitudes on smoking and health in Latin America and the Caribbean, 1982-1990 38+ 25 27 22 36 58/26_ 21 12 8 7 3 19 4 7 16 26 44+ 34 Country Bolivia Year Sample 1983 344 daily smokers 1983 120 adolescents 1987 72 physicians Question Response (%I Is smoking dangerous? (yes) 83 Is smoking harmful to health? (somewhat 96 or very) Is smoking harmful to health? (somewhat 94 or very) Brazil 1988 P&-to Alegre 1988 PBrto Alegre Is the life expectancy of smokers decreased by smoking? (yes) Is environmental tobacco smoke harmful to children? (yes) 48 100 78 Prevalence and Mortality Table 19. Continued Country Costa Rica Year 1984 Sample Urban students Cuba 1988 Nationwide Guatemala 1989 Treasury employees Honduras 1986 1987 Preuniversity students aged 15-30 Ministry of Health employees in Tegucigalpa Mexico Panama 1989 Students Paraguay Peru 1990 Physicians 1982 Adolescents 1989 Adult smokers Puerto Rico 1989 San Juan Uruguay 1984 Montevideo Venezuela 1984 Nationwide 1984 Caracas 1988 Nationwide 1988 Nationwide 1984 Nationwide 1986 Caracas 1986 Caracas 1986 Caracas 1989 Caracas Question Response (X) Are health risks associated with smoking? (adequate knowledge of such risks) Do you approve of a ban on indoor smoking? (yes) Are health risks associated with smoking? (low level of knowledge) Does smoking cause lung cancer and other diseases? (yes) Do you favor a worksite smoking regulation? (yes) Are you bothered by smoking at your worksite? (yes) Is smoking harmful to health? (yes) Is smoking less harmful than use of other drugs? (yes) Are you bothered when other people smoke? (yes) Is smoking undesirable? (yes) Is smoking harmful? (yes) What is the most important reason to stop smoking? (health) Do you believe that smoking is harmful to the health of smokers? (yes) Does smoking affect health negatively? (yes) Is smoking harmful to health? (yes) Should smoking be restricted in public places? (yes) Should all forms of tobacco advertising be banned? (yes) Is smoking harmful to others? (yes) Are some cigarettes less harmful than others? (yes) Should smoking be restricted in public places? (yes) Should radio and television advertising of tobacco be banned-including indirect advertising? (yes) 81 98 64/56* 50 70 77 90 55 60 30 95 66 89 81 94 83 72 75/81' 53 89 60 so urce: Pan American Health Organization (1992). `Smokers/nonsmokers. Table 20. Modified stem-and-leaf display of prevalence of smoking (%) among adults, selected countries of Latin America and the Caribbean, 1980s and 1990~~ - Men o-9 10-19 20-29 30-39 40-49 50-59 60-69 Women o-9 2 3 4 4 5 6 6 8 8 9 10-19 11 11 11 11 12 13 13 13 14 14 14 14 15 16 17 17 18 18 18 18 20-29 20 20 22 23 23 23 23 24 24 25 26 26 27 27 27 28 29 30-39 30 31 31 32 32 33 33 34 36 36 38 40-49 40 40 45 49 6 10 10 15 18 19 20 23 25 25 25 26 27 27 28 28 32 32 32 33 34 34 35 35 35 35 35 36 36 36 37 37 38 38 38 39 39 40 41 41 42 42 43 43 44 45 45 45 46 48 48 48 48 49 49 51 52 52 53 56 66 68 Median = 37 Median = 20 *Prevalence data from Table 16 are grouped by decile (stem) and listed in ascending order (leaf). The data are from different sources and derive from various methodologies. This display provides a visual overview of the range of measured values. Table 21. Prevalence of smoking (%) among Hispanic persons in the United States, aged 20-74, by ethnic group and sex, selected years Ethnic group and sex 1982-1984* 1987+ Mexican origin (southwest United States) Men 43.6 31.8 Women 24.5 17.4 Cuban origin (Miami area) Men 41.8 23.3 Women 23.1 20.4 Puerto Rican origin (New York City area) Men 41.3 38.6 Women 32.6 24.1 *Hispanic Health and Nutrition Examination Survey, 1982- 1984 (Escobedo, Remington, Anda 119891). `Schoenborn (1989). 80 Prevalence and Mortality Smoking-Attributable Mortality in Latin America and the Caribbean Introduction Births and deaths are the most widely collected and reported health events, and mortality is a stan- dard measure of the health status of a population. Mortality has traditionally been used as an indicator of socioeconomic status and standard of living, espe- cially in countries for which measures of economic productivity are inappropriate. Mortality is a useful measure when setting health priorities, communicating health-related infor- mation, and marshalling political support for a health initiative. It is a measure easily understood by the public, and it can affect the public's perception of risk. For example, the following statement about the United States has a po\sJerful simplicity: "cigarette smoking, alone, causes more premature deaths than do all the following together: acquired immunodefi- ciency syndrome, cocaine, heroin, alcohol, fire, auto- mobile accidents, homicide, and suicide" (Warner 1987, p. 2081). Yet the data that allow such a statement are difficult to assemble, and the methodologies used to determine the number of deaths attributable to smoking are complex (USDHHS 19891. Although useful, mortality data do not indicate the full effect of a disease or set of diseases on a community. They do not describe the pain, morbidity, disability, economic costs, and decreased quality of life of persons who live with an illness, nor do they describe the secondary effects on family members who lose a close relative. However, other measures of the effect of a dis- ease have limitations as well. For example, life expec- tancy, which can express the health status of a population, may be misleading. For developing coun- tries, life expectancy is strongly inff uenced by infant and childhood mortality and much less so by disease prevention or therapeutic advances that affect adult health. People who have died from a smoking-related disease would have lived approximately 15 years longer if they had not been smokers (Warner 1987). This powerful effect is diluted if the improvement in smokers' life expectancy is averaged over the whole population. In the following discussion, an attempt is made to specify the number of deaths in Latin America and the Caribbean attributable to smoking, while keeping in mind the limitations of common disease measures. The result is an approximation, an early step in an iterative process for determining the health impact of tobacco use in the Americas. The methodology, which applies the concept of attributable mortality, is com- plicated by the need to estimate and adjust data to compensate for missing or insufficient data. A step- by-step description of the methodology is provided in Table 22. The effects of the empirical decisions made are discussed at the end of the chapter (see "A Com- ment on the Methodology"). Mortality Data The data in this section are from the PAHO Technical Information System, a data base that in- cludes mortality information. PAHO collects mortal- ity data (by age, sex, and cause of death) from source jurisdictions by using questionnaires, national publi- cations, and other methods. Most of the data are from civil registries, which rely on death certificates com- pleted by health personnel in the field. These mortal- ity data have several problems: the coverage of the population is incomplete, the quality of some data is questionable, and the cause-of-death groupings of the World Health Organization (WHO)/PAHO data col- lection questionnaire limit comparability with other data. Coverage PAHO has estimated that the underregistration of mortality is more than 20 percent in Brazil, Colom- bia, Dominican Republic, Ecuador, EI Salvador, Hon- duras, Panama, and Peru (PAHO 1990b). The diverse reporting standards from various countries necessi- tated several country-specific decisions. In Brazil, for example, the most populous country in Latin America and the Caribbean, the estimated underregistration is approximately 25 percent. The level of underreport- ing differs between areas, although it tends to be worse in the poorer, northern part of the country. The number of reported deaths was used for the whole country, although it is an underestimate. In Paraguay, mortality information is published for only a portion of the country, and the information may not be repre- sentative of the remainder of the country. However, the areas not covered by the mortality registry are geographically defined and include about 40 percent of the population. Thus, reasonably reliable disease rates can be determined for a portion of Paraguay but not for the country as a whole. For this country, data from the well-defined reporting areas only were used; for other countries, similar decision rules were used. Prez~alrircc mrd Mortality 81 Estimate overall mortality For each country, evaluate vital registration and use the portion of the data that provides an accurate population-based mortality estimate. For the 10 jurisdictions without mortality data, use United Nations population schedules and apply mortality rates from countries with similar socio- demographic configurations. Do not correct for underreporting. Exclude and do not correct for ill-defined causes. (Resultant population and mortality estimates are reported in Table 25.) Estimate cause-specific mortality Identify the major smoking-associated disease groups (coronary heart disease; cerebrovascular disease; lung cancer; oral, laryngeal, and esophageal cancer; bladder cancer; and chronic obstructive pulmonary disease [COPDI). Use cause-specific mortality data for countries for which such data are available. For the 10 jurisdictions without such data, use data from four countries representative of the de- mographic and socioeconomic spectrum of the Americas (Guatemala, Colombia, Argentina, and the United States). (Resultant cause-specific mortality estimates are in Table 26.) Estimate relative risk and attributable risk Use U.S. estimates for relative risk since country- specific relative risk is generally not available. Determine the smoking-attributable fraction (SAF) for the United States by using the attributable-risk calculation. Data Quality One measure of the quality of mortality information is the proportion of deaths assigned to the rubric "symp- toms, signs, and illdefined conditions" (Manual oflnter- national Statistical Classification of Diseases, Injuries, and Causes of Death, Ninth Revision [ZCD-91). Currently, the percentage of mortality ascribed to ill-defined Table 22. Method used for calculating smoking-attributable mortality in the Americas Adjust estimates by using an index related to lung cancer Use an index of the maturity of the epidemic that relates the lung cancer rate for each country to that of the United States. For each country, determine an adjusted SAF for each disease by multiplying the index by the U.S. SAF for each disease (Table 32). For each country, multiply the adjusted SAF for each disease by the number of deaths from the disease to obtain smoking-attributable mortality (SAM) (approximately 375,000). Adjust the estimate further Calculate SAM for the United States alone by using this method and compare the result with the official value reported for 1985 (U.S. Department of Health and Human Services 1989). For each cause, calculate the difference between the result from this method and from the official method. Apply these upward adjustments to the cause-specific SAMs: increase COPD by 230R, increase cancers by 10.4% (using the difference in lung cancer esti- mates), and increase other diseases and causes by 16.4% (see footnotes to Table 33). Calculate the adjusted estimate of SAM in the Americas (526,000). causes is greater than 10 percent for 16 of the 39 jurisdictions submitting mortality information (PAHO 1990a). In this analysis, ill-defined causes were excluded from calculations of proportions or rates. Because a decedent may not have received health care or the certifying physician may not have been the physician who treated the decedent, diagnostic 82 Prevalence and MortaZity imprecision may occur. More serious distortion may result because the certifying physician did not have the diagnostic tools necessary for accurately determin- ing the cause of death.* Furthermore, managers of health services may not be willing or able to ensure accurate recording or conduct the diagnostic tests that would yield an accurate diagnosis, especially for the elderly. As a result, assessments of mortality levels and trends are often made by considering disease- specific rates for middle-aged rather than elderly pop- ulations (Doll and Peto 1981). Coding Since 1979, PAHO's participating member states have classified cause of death by using the ICD-9 coding scheme. To store these data, PAHO developed a grouping of causes of death based on, but not iden- tical to, the Basic Tabulation List of the ICD-9. The PAHO grouping is also similar, but not identical, to the groupings used by WHO and CDC. The difference in grouping, which has a variable effect on disease classification, does not affect deaths categorized as due to the following conditions: Com~ition ICD-9 code Coronary heart disease 410414 Cerebrovascular disease 430-438 Lung cancer 162 Cancers of the lip, oral cavity, or pharynx 140-149 Cancer of the esophagus 150 Cancer of the larynx 161 Cancer of the bladder 188 However, PAHO grouped cancers of the pan- creas (ICD-9 157) and kidney (ICD-9 189) with other cancers. Chronic obstructive pulmonary disease (COPD), when coded as ICD-9 490-492 and 496, can- not be isolated in the PAHO grouping. The relevant PAHO categories are "bronchitis (chronic and unspec- ified), emphysema, and asthma" fICD-9 490-493). Thus, unlike the grouping for COPD used in the cal- culation of smoking-attributable mortality (SAM) in the United States (CDC 19911, the PAHO grouping includes ICD-9 493 (asthma) and excludes ICD-9 496. Life Expectancy and Mortality Trends in Life Expectancy and Overall Mortality For all countries and subregions of the Americas, the overall trend is an increase in life expectancy at birth (Table 23). Over the last 35 years in Latin Amer- ica and the Caribbean, the average life expectancy at Table 23. Life expectancy* at birth for persons born during selected periods, by region and count6 Region and country Year of birth 1950-55 1970-75 1985-90 51.8 61.2 66.6 40.4 46.7 53.1 37.6 48.5 54.7 43.9 55.5 61.4 42.1 54.0 62.0 45.3 58.8 62.2 42.3 54.7 63.3 42.3 54.0 64.0 51.0 59.8 64.9 48.4 58.9 65.4 2000 Latin America Bolivia Haiti Peru Guatemala El Salvador Nicaragua Honduras Brazil Ecuador Dominican Republic Paraguay Colombia Mexico Venezuela Argentina Chile Uruguay Costa Rica Cuba 69.7 60.5 59.4 67.9 68.1 68.8 69.3 68.2 68.0 68.2 46.0 59.9 65.9 69.7 62.6 65.6 66.9 67.9 50.6 61.6 68.2 70.7 50.8 62.6 68.9 72.1 55.2 66.2 69.7 71.3 62.7 67.3 70.6 72.3 53.8 63.6 71.5 72.7 66.3 68.8 72.0 73.0 57.3 68.1 74.7 75.8 59.5 71.0 75.2 76.3 Caribbean 56.4 67.1 72.4 74.7 North America 69.1 72.2 76.1 78.1 United States 69.0 71.3 75.4 77.6 Canada 69.1 73.1 76.7 78.5 Source: Centro Latinoamericano de Demografia (1990); Pan American Health Organization (1990a). *Estimates through 1985 are based on actual data. After 1985, estimates are projections based on trends and on comparisons with data from similar countries. birth has increased by about 15 years-from 51.8 to 66.6 years. In North America, the increase was seven years-from 69.1 to 76.1 years, reflecting a slower increase as life expectancy at birth reaches age 75 to 80. Among Latin American and Caribbean coun- tries, the current differences in life expectancy at birth are great-ranging from 53.1 and 54.7 years in Bolivia and Haiti, respectively, to 75.2 years in Cuba. Over the last 35 years, the range has diminished somewhat. 2 Historically, the lack of appropriate diagnostic tools had a major impact on the number of deaths assigned to lung cancer. When diagnostic radiology was introduced in England in the 192Os, the rate of certified lung cancer deaths increased threefold (Pet0 7986). Prmahce mzd Mortality 83 During 1950 to 1955, the range was 28.7 years; today it is 22.1 years, and by the year 2000, it is expected to decrease to 16.9 years. Few Latin American and Ca- ribbean countries are at the low end of the range. Cur- rently, only about 3 percent of the population of Latin America and the Caribbean lives in countries with a life expectancy at birth lower than 55 years, while 86 percent lives incountries with a life expectancy at birth of 65 years or more. All countries except Bolivia and Haiti are expected to achieve a life expectancy at birth of 65 years or more by the year 2000 (PAHO 1990a). Differences in the rate of increase are also evident among countries. For example, although life expec- tancy at birth in Chile and Uruguay is now similar, it increased three times more in Chile than in Uruguay over the last 35 years. In general, the increase in life expectancy at birth was slower in the 1970s and 1980s than in the 1950s. The current life expectancy at birth in Latin America and the Caribbean is equivalent to that in the United States around 1945 to 1950-before many major advances in chronic disease prevention and treatment occurred (PAHO 1990a). Based on the cur- rent rate of improvement, the life expectancy at birth in Latin America and the Caribbean should reach that currently found in the United States by about the first quarter of the next century (Centro Latinoamericano de Demografia 1990). The range of population and mortality parame- ters is illustrated by data for four countries (Guate- mala, Colombia, Argentina, and the United States) that represent the broad spectrum of variation within the Americas (Table 24). This variation highlights the diverse potential effects of smoking on a population. For example, for all deaths in women (excluding deaths from ill-defined causes), the fraction of deaths in women aged 35 or older ranges from 34 percent in Guatemala to 95 percent in the United States. Since most SAM occurs among persons 35 years old or older, it is this group whose health is most affected by a tobacco habit. Estimates of Mortality The PAHO Technical Information System con- tains national mortality data suitable for this analysis for all but 10 jurisdictions in the Americas: Antigua and Barbuda, Bermuda, Bolivia, Guadaloupe, Gre- nada, Haiti, Montserrat, Netherlands Antilles, Nicara- gua, and Saint Pierre and Miquelon. To determine the number of deaths in the Amer- icas attributable to smoking, the number of deaths for these 10 jurisdictions had to be estimated. The United Nations (1989) population estimates for these jurisdic- tions were used for this calculation. Crude population mortality rates and other major mortality parameters were applied by using data for countries in the PAHO Technical Information System believed to be similar with respect to life expectancy, geographic region, per capita gross national product, tobacco consump- tion rates, and other factors. These estimates were used along with others obtained by standard means (Table 25). These nonstandard estimates are sensitive to the choice of country used to model the mortality struc- ture. In general, these are underestimates of actual mortality because of underreporting and because Table 24. Mortality from defined causes,* selected countries, c. 1985 Persons aged 235 years Country Sex Population+ Total Mortality+ Mortality+ Percentage of total mortality Guatemala M 3,914 F 3,826 Colombia M 14,103 F 14,007 Argentina M 15,049 F 15,283 United States M 116,160 F 122,571 Source: Pan American Health Organization (1990b). *Excludes ill-defined causes; see text. +Number, in thousands. 32 11 34 27 9 34 74 45 61 55 39 70 129 110 85 103 89 86 1,080 987 91 975 930 95 84 Prevalence and Mortality Table 25. Mortality from defined causes,* regions of the Americas, c. 1985 Region Latin America Sex M F Andean Area Southern ConeS Brazil M Central Americas Mexico M 39,744 224 F 39,631 171 Latin Caribbean M 12,934 101 F 12,801 87 32 41 26 42 134 60 112 66 63 62 52 60 Caribbean M 3,510 21 17 78 F 3,571 18 15 82 North America" M 128,768 1,179 F 135,410 1,055 1,078 92 1,006 95 1,831 77 1,614 82 3,444 80 All regions of the Americas M 329,301 2,368 F 335,868 1,965 Total 665,169 4,333 Population+ Total Mortality+ 197,023 1,168 196,887 892 40,177 207 39,705 166 24,377 190 24,785 153 Mortality+ 736 592 109 95 159 131 Percentage of total mortality 63 66 53 57 84 86 67,601 367 239 65 67,963 254 177 70 12,190 78 12,002 62 Source: Pan American Health Organization (1990b) *Excludes ill-defined causes; see text. +Number, in thousands. SIncludes Falkland Islands. SExcludes Belize. IlIncludes Bermuda and St. Pierre and Miquelon. mortality from ill-defined causes has been excluded to a high of 92 to 95 percent in North America. Most (as discussed earlier). The resultant estimates of of the population of Latin America lives in countries smoking-related mortality are conservative. where this fraction is between 60 and 70 percent. Total, Cause-Specific, and Age-Specific Mortality The composite of reported and estimated mor- tality indicates that approximately 4,300,OOO deaths occur in the Americas each year; about half of these deaths (2,060,OOO) occur in Latin America (Table 25). In the Americas overall, about 80 percent of deaths occur among persons aged 35 or older; in Latin Amer- ica, the fraction is about 64 percent. The fraction of deaths occurring among persons aged 35 or older varies from a low of about 41 percent in Central America The greatest absolute increase in life expectancy at birth is generally associated with improvements in mortality rates for children. In Latin America, a gra- dient of economic development is associated with increased life expectancy. In general, the death rate for children is lower in more highly developed coun- tries, but the death rate for older persons is similar in various economic settings. For example, in Argentina, the mortality rate per 1,000 persons under five years of age is 7.9, and for persons aged 65 or older, it is 65.8. In Guatemala, the rate for persons under five years of Persons aged 235 years Prevalence and Mortality 85 age is 21.4, but for persons aged 65 or older, it is 67.5 (PAHO 1990a). The gradient of economic development is also reflected in the cause-of-death mortality structure. Among men aged 45 to 64, mortality from heart dis- ease, expressed as a percentage of total mortality, is 11 percent in Guatemala, 27 percent in Colombia, and 37 percent in the United States. But some similarities are emerging. For both the 45 to 64 and the 65 or older age groups, the three leading causes of death for each sex are the same in both Colombia and the United States. For the oldest age group, the leading cause of death- diseases of the heart-is also the same in Guatemala (PAHO 1990a). This pattern-increasing similarity of causes of death-is likely to intensify. As life expectancy im- proves, the epidemiologic profile of a country changes. Countries with a lower life expectancy tend to have a younger population, and the greatest mor- tality is in the younger age groups. In these countries, deaths are primarily due to infections (such as acute respiratory infections and diarrhea), malnutrition, and conditions originating in the perinatal period. As these diseases are controlled and life expectancy in- creases, deaths from chronic diseases-in particular, cardiovascular diseases and cancer-become the dominant health problem (PAHO 1990a). Mortality from Smoking-Related Diseases Estimates of Cause-Specific Mortality The major diseases associated with tobacco smoking include coronary heart disease, cerebrovas- cular disease, COPD, and cancers of the lung, lip, oral cavity, pharynx, larynx, esophagus, pancreas, blad- der, and kidney. In the United States, each of these causes (considering cancers of the lip, oral cavity, and pharynx as a single group) contributes at least 2,000 deaths to the total number of deaths attributable to smoking (USDHHS 1989). The four countries for which population data were assessed (Table 24) and the six smoking-related conditions (Table 26) were the focus of this analysis of the effect of smoking on countries of the Americas. Cancers of the lip, oral cavity, pharynx, larynx, and esophagus were grouped because of the similar smoking- attributable risk for these conditions (USDHHS 1989). Cancers of the kidney and pancreas were excluded because they cannot be specifically identified in the PAHO Technical Information System. The four countries Table 26. Deaths from six major causes as a percentage of all deaths from defined causes,* for persons aged 35 or older, selected countries, c. 1985 Corm-;ry Cormr~ry Cerebro- Cerebro- Oral,+ Chronic vascular vascular laryngeal, obstructive Country disease disease disease disease Lung and esopha- Bladder pulmona-ry All and sext (aged 35-64) (aged 265) (aged 3.564) (aged 265) cancer geal cancer cancer diseases categories Guatemala Men 2.2 3.6 1.2 2.5 0.1 0.3 0.1 1.2 11.2 Women 1.6 3.2 1.8 3.8 0.4 0.1 0.0 1.1 12.0 Colombia Men 6.3 10.1 3.4 6.8 2.1 1.6 0.3 1.9 32.5 Women 4.7 10.2 4.7 9.5 1.3 1.0 0.2 1.8 33.4 Argentina Men 4.8 8.1 3.6 7.0 5.6 2.5 0.9 1.2 33.7 Women 1.6 8.6 2.7 9.9 1.1 0.8 0.2 0.9 25.8 United States Men 7.6 21.3 1.0 5.0 8.5 1.5 0.7 1.3 46.9 Women 2.8 24.1 1.0 8.9 4.2 0.6 0.3 0.9 42.8 Source: Pan American Health Organization (1990b). *Codes from Manual uflntm~ational Stntistical Classification of Diseases, [Injuries, ad Causes of Death, Ninth Revision: coronary heart disease, 410-414; cerebrovascular disease,430438; lung cancer, 162; cancers of lip, oral cavity, and pharynx, 140-149; cancer of the esophagus, 150; cancer of the larynx, 161; cancer of the bladder, 188. `Cancer of the lip, oral cavity, and pharvnx. SThe denominator for each row is the total number of deaths from defined causes, by country and sex. %ee text for a description of this rubric. 86 Prevalence and Mortality chosen represented four different points on the spec- trum of mortality rates. Guatemala was chosen, even though its lung cancer rate is low, because it reports nationwide mortality statistics and has one of the low- est levels of life expectancy in Latin America. For persons aged 35 or older, the distribution of deaths from the six major causes was expressed as a percentage of all deaths from defined causes (Table 26). Because SAM from coronary heart disease and cerebrovascular disease differs significantly between persons aged 35 to 64 and persons aged 65 or over (USDHHS 1989), estimates for both these age groups are presented. For all six smoking-related illnesses and age sub- categories taken together (Table 26, last column), the proportion of deaths caused in persons aged 35 or older differed among the countries. In Guatemala, these diseases accounted for slightly over IO percent of adult deaths. In Argentina and Colombia, they accounted for 25 to 33 percent of deaths, while in the United States, they contributed approximately 45 per- cent of deaths. To estimate the number of deaths from smoking- related conditions for subregions of the Americas (Table 27), both the reported mortality data (Table 25) and synthetic mortality estimates for the 10 jurisdic- tions without data were used. For these jurisdictions, the mortality distribution patterns from the four se- lected countries (Table 24) were applied, as described. Substantially more deaths in North America than in Latin America and the Caribbean were attrib- uted to coronary heart disease, lung cancer, and blad- der cancer. The number of deaths was similar in North America and in Latin America and the Carib- bean for cerebrovascular disease, COPD, and oral can- cer. Using these estimates, 81 percent of lung cancer deaths in the Americas occur in North America. When accounting for underreporting, the proportion is prob- ably closer to 75 percent. (Using a different approach, other researchers have estimated that North America accounts for 77 percent of lung cancer deaths [Parkin, Laara, Muir 19881). Because lung cancer is a strong indicator of all smoking-attributable diseases, a rough approximation suggests that the number of deaths in Latin America and the Caribbean attributable to smoking will be about one-third to one-fourth of the number in North America. Estimates of Relative Risk Due to Smoking Relative risk is defined as Y = &1)/d(O), where d(l) and d(O) are the incidence of a particular disease for exposed and unexposed cohorts, respectively. For current smokers, the relative risk for a disease estimates the increase in disease incidence associated with a history of smoking. This risk varies widely among population groups due to differences in smoking- related factors, such as person-years of smoking con- tributed by heavy smokers, age at initiation, and ciga- rette product smoked. For example, among current smokers in a population, the relative risk for lung cancer would be expected to be relatively low if a sizable proportion of the population recently began to smoke heavily. If, however, heavy smoking has been common since World War II, the risk would be rela- tively high. The main reason for this effect is that the exposure category defined by "current smokers" is based on current rather than past smoking habit, but lung cancer rates primarily depend on smoking pat- terns of 20 or more years ago. For many of the smoking-related causes of death, few country-specific estimates of relative risk are available for Latin American and Caribbean popula- tions, and most have focused on cancer. For current cigarette smokers in the United States, aged 35 or older, the estimated relative risk for lung cancer is 22.4 for men and 11.9 for women (USDHHS 1989). In Cuba, the relative risk is 14.1 for men and 7.3 for women. Dark tobacco is the variety of tobacco most commonly smoked in Cuba and many other areas of Latin America. In Cuba, dark tobacco is associated with a higher relative risk for lung cancer than light tobacco is: for men, 14.3 and 11.3, respectively, and for women, 8.6 and 4.6, respectively (Joly, Lubin, Car- aballoso 1983). In Colombia, the relative risk for lung cancer among current smokers was 10.3 in one case- control study of 102 persons with lung cancer, 74 percent of whom were men (Restrepo et al. 1989). The study in Colombia also reported relative risk for cancer of the bladder, larynx, and oral cavity/ hy- popharynx of 3.7, 37.9, and 11.2, respectively. In La Plata, Argentina, where the rate of bladder cancer is high, a relative risk of 7.2 for bladder cancer was found for men who were current smokers (Iscovich et ~31. 1987). In a study of 232 cases of cancer in Brazil (87 percent of patients were men), the relative risk for cancer of the tongue, gum, floor of the mouth ~1~1 other parts of the oral cavity was 9.3 for current snlok- ers of manufactured cigarettes (Franc0 et al. 1989). 11~ a 1966 case-control study of male cigarette smokers and nonsmokers in Puerto Rico, the relati\,e risk \VII~ 1.5 for esophageal cancer, 1.1 for cancer of the scroll cavity, and 2.7 for cancer of the pharynx (Martine' 1969). In Montevideo, Uruguay, the relati\,e risk for laryngeal cancer was 35.4 for male smoker5 Of d&irk tobacco and 14.7 for male smokers of light tobacco (IX Stefani et al, 1987). For comparison, for U.S. mc'il \\`llo Table 27. Deaths (in thousands) from six major causes, * for persons aged 35 or older, selected regions - of the Americas, c. 1985 _.~. ~~__~ Coronary Coronary Cerebro- Cerebro- heart heart vascular vascular Re ion Fi disease disease disease disease Lung an sex (aged 35-W (aged 265) (aged 3S64) (aged 265) cancer Oral,+ laryngeal, and esopha- geal cancer Bladder cancer Chronic obstructive p;;ye;;;q " Latin America Men 38.1 Women 16.7 Andean Area Men 5.5 Women 3.1 Southern Cones Men 7.3 Women 2.2 Brazil Men 16.6 Women 7.2 Central America" Men 1.0 Women 0.5 Mexico Men 4.2 Women 1.9 Latin Caribbean Men 3.5 Women 1.8 Caribbean Men 0.8 Women 0.4 North America' Men 82.2 Women 27.8 All regions of the Americas Men 121.0 Women 44.9 Total 165.9 59.7 28.5 49.8 22.4 14.1 3.0 15.6 53.2 22.6 55.5 6.8 4.0 1.0 11.3 8.7 3.2 6.1 2.1 1.3 0.3 2.0 7.6 3.3 7.2 1.0 0.6 0.1 1.8 13.8 5.5 11.9 8.2 3.9 1.2 2.9 12.6 3.7 14.4 1.4 1.1 0.3 1.7 19.3 15.7 21.0 6.1 6.2 0.9 4.8 17.5 11.5 21.7 1.9 1.4 0.3 2.7 2.0 0.5 1.3 0.3 0.2 0.05 0.6 1.6 0.6 1.5 0.2 0.01 0.02 0.5 7.1 2.3 5.6 2.8 1.1 0.3 4.5 6.3 2.2 6.8 1.3 0.4 0.1 3.8 8.9 1.4 3.9 2.5 1.4 0.3 0.9 7.5 1.3 3.9 0.9 0.4 0.1 0.7 1.2 0.7 1.9 0.4 0.3 0.1 0.3 1.1 0.6 2.2 0.1 0.1 0.03 0.2 230.3 11.2 54.7 92.0 16.4 7.5 14.2 242.2 9.6 89.8 41.7 6.1 3.4 9.2 291.2 40.4 106.3 114.4 30.8 10.6 30.1 296.5 32.8 147.5 48.5 10.1 4.4 20.6 587.7 73.2 253.8 162.9 40.9 15.0 50.7 Source: Pan American Health Organization (1990b). *Codes from Manual of Intertrational Statistical CIassificatim of Diseases, I?rjuries, fl?~d Cawes (?f Dmth, Ninth Revision: coronary heart disease, 410414; cerebrovascular disease, 43&438; lung cancer, 162; cancers of lip, oral cavity, and pharynx, 14G149; cancer of the esophagus, 150; cancer of the larynx, 161; cancer of the bladder, 188. `Cancer of the lip, oral cavity, and pharynx. % ee text for a description of this rubric. _Includes Falkland Islands. "Excludes Belize. `Includes Bermuda and St. Pierre and Miquelon. 88 Prevalence and Mortality are current smokers, the relative risk for cancer of the bladder is 2.9, cancer of the esophagus 7.6, cancer of the larynx 10.5, and cancer of the lip, oral cavity, and pharynx 27.5 (USDHHS 1989). Two case-control studies were conducted to in- vestigate the factors associated with esophageal can- cer in Uruguay, which has one of the highest rates of esophageal cancer in the world. In one study of 226 cases, the relative risk was 6.5 for ever smokers (82 percent were men) (Vassallo et al. 1985). In the other study of 199 cases, the relative risk was 5.7 for current male smokers (De Stefani et al. 1990). In bordering southern Brazil, which also has a high rate of esopha- geal cancer, the relative risk was 8.4 for male smokers (Victora et al. 1987). For countries for which relative risk estimates were lacking, relative risks were derived from U.S. data and used in the following computations of SAM (USDHHS 1989, 1990). Small differences in relative risk estimates are unlikely to have a large overall effect on SAM because of the structure of the formula for calculating attributable risk (see below). Smoking-Attributable Mortality Estimates of Smoking-Attributable Mortality Worldwide Interest in attempting to quantify the extent of the health hazard caused by tobacco led to develop- ment of smoking-attributable fractions (SAFs). These values estimate the proportion of cases of a specific disease in a population that can beattributed to smoking. SAF = P(u-ll--- 1 +p (r-l) in which p is the proportion of the population that has ever smoked and Y is the risk for ever smokers relative to never smokers. The SAF calculated for each disease of interest is multiplied by the number of deaths for that disease, and the result is the SAM for that disease. The sum of SAM values for all diseases associated with tobacco use gives the total number of deaths attributable to smoking. The SAF can be refined to account for differences in smoking status (never, current, or former smoker) and for age and sex subgroups. Smoking prevalence and relative risk can be estimated for each of these subgroups. SAFs have been calculated for 10 selected causes of death in the United States (Table 28). Recent studies have estimated the number of deaths attributable to smoking in the United States (Table 29). The estimates by Rice and colleagues, Table 28. Smoking-attributable fraction for 10 selected causes of death, United States, 1985 Men Women ~~~ ~~~ L?) ~ (7c'c) 45 41 21 12 51 55 24 6 90 79 92 61 81 87 78 75 29 34 47 37 48 12 84 79 -. Cause of death Coronary heart disease (aged 35-64) Coronary heart disease (aged 565) Cerebrovascular disease (aged 35-64) Cerebrovascular disease (aged 265) Cancer of the lung Cancer of the lip, oral cavity, and pharynx Cancer of the larynx Cancer of the esophagus Cancer of the pancreas Cancer of the bladder Cancer of the kidney Chronic obstructive pulmonary disease Source: U.S. Department of Health and (1989). Human Services CDC, and USDHHS all considered smoking status, age, and sex. The estimates vary for several reasons: the diseases included, the specific methodology used, the target year, and the source of the smoking preva- lence data and the relative risk estimates. The most recent (1988) estimate for the United States (434,000 smoking-attributable deaths) is discussed in Chapter 4, "Economic Costs of the Health Effects of Smoking." The 1985 estimate is used here to maintain consistency with data available for Latin America and the Caribbean. SAM has been estimated for many European countries (Table 30), and the current worldwide esti- mate is 3 million smoking-attributable deaths per year. The methodology described earlier for calculat- ing SAM can be used for countries for which reliable information is available on smoking prevalence and on the risk for major tobacco-associated diseases among ever smokers relative to never smokers. Un- fortunately, few countries in Latin America have such data; an alternative methodology for calculating SAM is described below. Lung Cancer Mortality as an Index of Prior Smoking in a Population Numerous attempts have been made to describe the relationship between smoking habits and mortality Prezahce nud Mortality 89 Table 29. Smoking-attributable mortality smoking habits of the population, as expressed by the in the United States risk of dying from lung cancer. Reference Rice et al., 1986 U.S. Office of Technology Assessment, 1985 Centers for Disease Control, 1987b Year Estimate 1980 270,000 1982 314,000 1985 320,000 for R(US,N-Sj < R(C) < RKIS) U.S. Department of Health and Human Services, 1989 1985* 390,000 in which R(C) is the lung cancer mortality rate for a country in the Americas, R(US) is the lung cancer rate for the United States (Table 31), and RtUS,N-S) is the lung cancer rate for never smokers in the United States (12.7 for men and 11.1 for women). When R(C) is greater than RfUS), the index is arbitrarily set to 1. The index has the following properties: o It equals 0 for the few countries in Latin America and the Caribbean with a lung cancer rate below that of never smokers in the United States. o It equals 1 for countries that have a lung cancer rate higher than that of the United States (although there were none). *The 1985 estimate (rather than the 1988 estimate of 434,000 reported in Chapter 4, Table 1) is used here to maintain consistency with the demographic and vital data available for Latin America and the Caribbean. from lung cancer in a population. Many of these attempts have not been entirely successful, primarily due to the lack of key information. Current lung cancer mortality rates reflect smoking habits of 20 to 40 years ago. Reliable data on lung cancer incidence and mor- tality are available for many industrialized countries, but only limited information is available on previous smoking habits. Furthermore, the relationship between smoking and lung cancer is affected by many factors. Duration of smoking is the factor most strongly corre- lated with risk for lung cancer. For example, when duration of regular tobacco use is doubled from 15 to 30 years, lung cancer incidence increases about 20-fold Pete 1986). Other factors that affect lung cancer risk include number of cigarettes smoked per day, age at initiation, tar yield of tobacco products, use of filters, blend of tobaccos, and depth of inhalation. Many of these factors vary over time, not only for a national population but for individuals within a population. Only in recent years have surveys in a few industrial- ized countries collected data on these factors in some detail. Thus, data are unavailable for building an optimal model of smoking habits and lung cancer risk. Nevertheless, tobacco consumption is highly correlated with lung cancer; the SAF has been calcu- lated at over 90 percent for countries that have popu- lations with a long history of high prevalence of heavy smokers (Table 31). This strong association suggests that lung cancer mortality can be used as a surrogate to measure the impact of smoking on a population. The following index (0 uses lung cancer mortal- ity rates for the population aged 55-64. This index, a measure of smoking maturity in a population, con- tains population risk factor information related to the o It falls between 0 and 1 for countries with a lung cancer rate between the U.S. rate for never smokers and the overall U.S. rate, and the value increases as the country's rate approaches that of the United States. This index can be used to develop estimates of SAM for countries in Latin America and the Carib- bean. For a given country, the lung cancer rate and index are calculated, and this lung cancer index is used to adjust all diseases. The index is multiplied bv the disease-specific SAF for the United States to obtain an adjusted disease-specific SAF for a specific country. The number of deaths from a specific cause is then multiplied by the adjusted SAF to obtain the SAM. Thus, the index adjusts the SAF downward-to a level appropriate for the extent of lung cancer in the population. The index is nonlinear; large changes in the upper range of lung cancer rates have only a small effect on the SAF. But changes in the lower range, closer to the rate for never smokers, have a proportionately larger effect on the SAF. In Table 31, the SAF is given with and without the index adjustment. The index uniformly offers a more conservative estimate of SAF. Because of the potential for diagnosis of lung cancer to be more inadequate in some elderly popula- tions than in younger populations, and because of the need to choose a relatively stable measure of smoking habits, the lung cancer rate for persons aged 55 to 64 was used in creating the index. If older age groups are used, significant diagnostic misclassification occurs, and the relationship to smoking is more tenuous. The low rates for younger age groups render the rate estimates 90 Preztalence and Mortality Table 30. Estimated number of deaths due to tobacco use in 27 countries of the World Health Organization (WHO) European Region* Country Austria Belgium Bulgaria Czechoslovakia Denmark Finland France German Democratic Republic Germany, Federal Republic of Greece Hungary Iceland Ireland Israel Italy Luxembourg Malta Netherlands Norway Poland Portugal Romania Spain Sweden Switzerland United Kingdom Yugoslavia Total for region Total worldwide Year 1985 1984 1984 1984 1985 1984 1984 1984 1985 1984 1985 1984 1983 1984 1981 1985 1985 1985 1984 1985 1985 1984 1980 1985 1985 1984 1982 1991 Male Female 5,527 3,354 8,905 2,664 6,129 3,215 14,693 7,363 5,531 3,311 4,094 1,900 25,751 10,102 12,393 6,178 49,572 26,433 5,305 1,718 10,742 5,541 115 78 2,754 1,449 1,416 859 39,489 15,324 298 121 115 54 12,140. 3,892 3,046 1,553 23,858 7,337 3,656 1,778 12,178 7,907 14,492 5,738 7,104 4,339 4,299 1,610 60,764 33,916 9,103 3,732 343,469 161,466 Total 8,881 11,569 9,344 22,056 8,842 5,994 35,853 18,571 76,005 7,023 16,283 193 4,203 2,275 54,813 419 169 16,032 4,599 31,195 5,434 20,085 20,230 11,443 5,909 94,680 12,835 504,935 3,000,000 Source: WHO (1988, 1991 [for worldwide estimate]). Represents about 6OYG of the regional population. Tobacco is held responsible for about 90% of all deaths from lung cancer, 75% of bronchitis/emphysema deaths, and 257c of all deaths from ischemic heart disease. The estimate for each country is based on the most current data provided to WHO by the countries themselves. unstable. Further, the use of a single, well-defined group at risk has the virtue of simplicity-data di- rectly available to a country are used, and adjustment that might be necessary for cross-country compari- sons is avoided. Estimates of Smoking-Attributable Mortality in the Americas Unadjusted Estimates Before adjustment, approximately 375,000 deaths in the Americas were attributable to smc.king around 1985 (Table 32). These were distributed by disease as follows: Disease Total SAM Coronary heart disease 144,200 Cerebrovascular disease 46,800 Lung cancer 128,600 Oral, laryngeal, and esophageal cancer 23,200 Bladder cancer 5,700 COPD 27,300 In Latin America and the Caribbean, an interme- diate estimate of 64,000 smoking-attributable deaths was obtained, and most of these deaths were from coronary heart disease (about 18,500), cerebrovascular disease (about 17,000), and lung cancer (about 13,000). The largest contribution to SAM in Latin America was made by Brazil, followed closely by the Southern Cone subregion. Prezlalence and Mortality 91 Table 31. Smoking-attributable fraction (SAF) and adjusted SAF for lung cancer mortality, selected industrialized countries, 1978-1981 Index of Crude lung Country Sex cancer rate* SAF smoking maturity A%Fd Canada M 142.8 .90 .92 .85 F 34.0 .71 .77 .60 England and Wales M 228.5 .94 1 .oo .92 F 63.3 .80 1.00 .78 Japan M 64.8 .83 .58 .53 F 21.0 .58 .50 .39 Sweden M 85.0 .83 .69 .63 F 28.0 .57 .66 .51 United State& M 166.7 .92 1 .oo .92 F 50.0 .78 1.00 .78 Source: Adapted from International Agency for Research on Cancer (1986). :For persons aged 35 or older. The calculation actually uses the rate for persons aged 5544 years. Difference between SAF and adjusted SAF .05 .ll .02 .02 .30 .19 .20 .06 .oo .oo :Calculated by multiplying the SAF for the United States by the country-specific index of smoking maturity; see text. +Total population. SAM was calculated as the percentage of deaths exclusion of deaths attributed to ill-defined causes. for persons aged 35 or older (last column of Table 32). SAM was adjusted for the first four of these factors as For the Latin American subregions, the proportion follows. For the United States, the estimate of SAM was highest for men in the Southern Cone and lowest was calculated and compared with that made for 1985 for men in Central America. In the Southern Cone, the KJSDHHS 1989). The latter estimate, which provided difference in the rate for men and women reflects a a benchmark, was 37.2 percent larger than the estimate large historical difference in the rate of tobacco con- computed in this analysis. The percent difference be- sumption (see "Prevalence of Smoking" earlier in this tween these two estimates was used to alter upward chapter). The lung cancer mortality rate for women in the estimates for the other countries in the Americas. Peru was less than that for U.S. women who were The adjustments were made by cause (Table 33, see never smokers. The index was zero, and by this footnotes), since the degree of underestimate varied method, no deaths were attributable to smoking. with the condition. Adjusted Estimates The estimates of SAM (Table 32) are under- estimates for several reasons: (1) COPD was un- dercounted due to differences in cause-of-death groupings; (2) cancers of the kidney and pancreas were omitted; (3) the SAF for cancers of the oral cavity, esophagus, and larynx is an underestimate (the three cancers were grouped, and the smallest SAF for the three was used); (4) other categories of disease or death were omitted, including other types of cardiovascular and respiratory disease, cervical cancer, infant deaths due to maternal smoking during pregnancy and post- natal exposure to environmental tobacco smoke, lung cancer deaths due to passive smoking, and deaths from smoking-related fires; and (5) an undercount of deaths due to both underregistration of cases and the After adjustment, an estimated 526,000 annual deaths in the Americas are attributable to tobacco use; about 100,000 of these are in Latin America and the Caribbean. About two-thirds of these deaths occur in Brazil and the Southern Cone. The estimated 36,000 deaths for Bermuda, Canada, and St. Pierre and Miquelon correspond closely with estimates derived by using several different methods and previously reported for Canada alone (Collishaw, Tostowaryk, Wigle 1988; PAHO 1992). As discussed below, the 100,000 annual deaths in Latin America and the Car- ibbean, estimated from data for the mid-1980s, is conservative. If the current U.S. SAF is applied and if Latin American and Caribbean countries follow a tra- jectory similar to that of North America, over 1 million smoking-attributable deaths per year will occur in Latin America and the Caribbean by the year 2030. 92 Prevalence and Mortality A Comment on the Methodology The attribution of mortality requires an empiri- cal approach. In the method used here, which varies somewhat in detail, but not in fundamental approach, from other methods (WHO 19891, at least five basic empirical decisions were made. First, the analysis excluded mortality data for which cause of death was inadequately specified, and no attempt was made to adjust for the underreporting of deaths. Second, syn- thetic estimates of mortality structure were used for countries with little or no data. Third, a proration was used to adjust for causes of death that could not be analyzed by using PAHO data. Fourth, an empirical index was developed to adjust for the many factors that influence the risk that smoking imposes on a population. Fifth, the SAM calculation made for the United States (USDHHS 1989) was used as a bench- mark for adjusting the estimates derived in this anal- ysis. Each of these decisions influenced the final estimate; in addition, some specific features of the index and factors related to attributable risk in general also had an influence. The net effect of the empirical decisions is diffi- cult to assess, but the first decision-no correction for underreporting and no proration for ill-defined causes-probably dominates and results in a sizable underestimate. The order of magnitude of the under- estimate can be approximated by comparing the esti- mate of total mortality in Latin America derived for this analysis (2,060,OOO [Table 251) with an estimate, derived by using regression methods, that attempted to account for underreporting (3,197,OOO [Hakulinen et al. 19861). Based on this difference in overall mor- tality of about 55 percent, the number of smoking- attributable deaths might be as high as 155,000. The more conservative estimate of 100,000 smoking- attributable deaths was deemed more appropriate be- cause it directly relates to the data with which ministries of health in Latin American and Caribbean countries actually work. In addition, the conservative method allows a simple, uniform decision rule to be used by all countries of the region in making their own computations. Finally, this approach allows for in- creasingly credible estimates of SAM to be made as better mortality data become available and the esti- mates are gradually refined. The index of smoking maturation is based on a comparison of lung cancer rates. Although accurate information is more readily available for lung cancer than for other conditions, it may not be the optimal condition for use in calculating the index. Although tar levels affect the risk for lung cancer, they appar- ently do not affect the risk for cardiovascular disease and COPD (USDHHS 1981). Further, the lag between increased consumption of tobacco and a rise in lung cancer mortality may not be representative of the lag for other diseases. In addition, use of the 55 to 64 age group for calculating the index underestimates the population's exposure to smoking in most Latin American and Caribbean countries because peak tobacco-consumption rates have not yet been reached. Because the index is empirical, there is no clear methodologic justification for the square root transfor- mation. Many transformations are available; the square root was used because of properties appropri- ate to the analysis. Specifically, taking the square root of numbers less than one produces a nonlinear effect: it increases all numbers that are less than one, but it has a greater effect on numbers close to zero than on numbers close to one. Thus, upward revision is pro- portionately greater for countries with low rates than for countries with high rates. This choice modulates, to some extent, the conservative nature of the index. On the other hand, no deaths were attributed to smok- ing in countries with lung cancer rates less than those for U.S. never smokers. Since smoking is not uni- formly distributed in such countries, rates may be higher for some subgroups, and at least some deaths should have been attributed to smoking. Finally, this methodology is weakened by a lack of information on multiple risk factors. The SAF may be higher or lower when risks other than smoking play a significant role in disease causation. Because smok- ing is the dominant risk factor for lung cancer, this effect is probably negligible. In cardiovascular disease, however, smoking interacts with hypertension, hyper- cholesterolemia, physical inactivity, obesity, diabetes, and possibly other risk factors as well, and this effect may be considerable. Thus, the empirical choices and the specifics of the analysis may have differing effects, but the final estimate of 526,000 annual deaths attributable to smoking in the Americas is almost certainly conserva- tive. This estimate-perhaps best viewed as the first point on a continuum of such estimates-provides an order of magnitude for the number of smoking-related deaths in the Americas. If, as suggested in the first half of this chapter, the prevalence of cigarette smoking is increasing in some areas, accurate assessment of SAM is of consid- erable importance. As noted, the lack of some critical data diminishes the precision of the estimates and fosters a greater reliance on empirical decisions. As data systems develop, individual countries will be better able to apply these methods for calculating SAM for their own populations. Table 32. Smoking-attributable mortality* for men and women in the Americas, c. 1985 Region and country Men Latin America Andean Area Colombia Peru Venezuela Southern Cone' Argentina Chile Brazil Central America** Mexico Latin Caribbean Cuba Caribbean North America++ Canada United States All regions of the Americas Women Latin America Andean Area Colombia Peru Venezuela Southern Cone' Argentina Chile Brazil Central America** Mexico Latin Caribbean Cuba Caribbean North America++ Canada United States All regions of the Americas Lung cancer mortality rateS Index of smoking maturity --__ - - .302 .140 .444 - .829 .566 .456 - ,376 - .716 (aged ~65) SAfl " (aged 265) (aged ~65) SAM" -___ SAF SAM SAF SAM - - 34.1 19.5 55.6 - 155.5 80.6 57.8 - 44.3 - 119.8 - 8,426 - 6,432 - 7,090 - 785 - 557 - 462 .136 386 ,063 287 .154 237 .063 24 .029 25 .072 24 ,200 354 .093 228 .226 167 - 2,583 - 2,245 - 2,151 .373 1,983 .174 1,156 .423 1,659 .255 290 .119 344 .288 261 .205 3,411 .096 1,844 .233 3,652 - 71 - 89 - 36 ,169 708 ,079 559 .192 435 - 867 - 1,138 - 355 ,322 711 ,150 974 .365 298 - - - 128 - 108 - 129 - - 209.0 .975 219.0 1.000 - 36,907 - 48,251 - 5,696 .439 3,376 .205 4,044 .497 449 .450 33,526 .210 44,204 .510 5,243 - - 45,460 - 54,791 - 12,914 - - 16.1 7.8 23.7 - 16.6 19.5 15.0 - 16.4 - 42.2 - - ,267 - ,409 - ,279 ,338 ,240 .2; - .632 - - ,710 - ,167 - .114 ,139 .098 - .112 - ,259 1,848 346 198 - 1,837 - 234 ,032 126 - - ,049 108 - 403 ,033 259 .041 119 ,029 505 28 ,033 207 - 459 ,076 405 - - 149 236 162 63 706 20 209 331 299 - ,147 - ,225 - .153 ,186 ,132 - ,151 - ,347 857 101 59 - 42 223 149 58 313 10 113 98 85 - - 75.1 90.1 - - .901 1 .ooo - - - .369 ,410 30 11,315 768 10,547 - 27 - - 28,854 - .108 1,919 .496 .120 26,934 .550 29 5,343 386 4,957 - 13,194 - 30,718 - 6,229 Coronary heart disease Coronary heart disease Cerebro- vascular disease Source: Pan American Health Organization (1990b). "Mortality from defined causes for persons aged `Cancer of the lip, oral cavity, and pharynx. 35 or older, in thousands. tThe lung cancer rate for U.S. never smokers used for the index calculation was 15.5 per 100,000 men aged 5.564 and 10.4 per 100,000 women aged 55-64. 94 Prevalence and Mortality - - ,073 .034 ,107 ,199 ,136 ,109 - ,090 ,172 5,959 - 11,549 - 5,946 - 845 - 5,819 52,066 ,071 418 - 579 - 305 - 36 - 441 3,584 ,033 222 .272 258 .236 164 ,142 17 .254 212 1,784 ,040 26 .126 33 ,109 12 ,066 2 ,118 38 184 .009 143 ,400 263 .346 116 ,209 15 ,373 148 1,434 .063 2,182 - 5,952 - 2,446 - 468 - 1,727 19,754 .124 1,527 .746 4,580 ,647 1,782 .390 368 ,697 916 14,370 ,131 315 .509 492 ,441 260 .266 35 475 553 2,549 .083 2,302 .410 2,522 .356 2,197 .214 191 ,383 1,827 17,945 .075 56 - 75 - 33 - 5 - 67 431 .013 503 .339 938 .293 322 .177 47 .316 1,409 4,920 .037 498 - 1,483 - 645 - 98 - 348 3,432 .086 399 ,644 1,334 ,558 462 ,336 86 .601 252 4,517 ,146 - 198 - 214 - 131 - 18 - 105 1,030 ,062 - 13,098 - 82,569 - 12,781 - 3,503 - 11,892 214,696 .199 .234 1,162 .878 7,249 .761 1,254 .458 367 ,819 1,245 19,147 .211 .240 11,932 ,900 75,310 ,780 11,523 .470 3,135 ,840 10,645 195,519 .198 - 19,255 - 94,331 - 18,859 - 4,366 - 17,817 267,792 ,146 - 857 101 59 - 1,567 210 102 - 673 97 64 - - ,016 - ,025 - .017 .020 ,014 - .016 - .038 - .211 - .323 - .220 .267 .190 - .217 - .499 - ,163 - ,249 - ,170 ,206 .147 - ,167 - .385 - .099 - .151 - ,103 ,125 .089 - ,102 - .234 103 11 7 - 2,257 12,247 .021 289 1,716 .018 151 977 .025 - 42 223 149 58 313 10 113 98 85 108 317 220 87 360 2% 389 352 33 181 116 55 207 10 74 104 89 - 5 30 22 6 30 1 10 20 17 - ,211 - .323 - .220 ,267 ,190 - .217 - .499 - - 138 739 386 2,308 170 1,473 203 727 517 4,156 35 140 827 2,052 203 1,877 167 1,671 - .039 .018 .016 ,029 ,023 ,005 ,018 .036 ,069 - 29 5,343 386 4,957 - - .054 .060 - .712 .790 26 32,706 2,242 30,463 - - .550 ,610 11 3,657 312 3,345 - 6,229 - 34,299 - 4,342 - - .333 ,370 - 3 1,253 105 1,148 - - .712 .790 33 218 ,015 7,170 95,562 ,095 532 6,631 .088 6,638 88,928 ,096 1,359 - 9,460 108,027 .067 Cerebro- vascular disease (aged 265) SAF SAM Lung cancer SAF SAM SAF SAM SAF SAM SAF SAM Oral,+ laryngeal, and esophageal cancer `Smoking-attributable fraction. `iSmoking-attributablemortality. `Includes Falkland Islands. *`Excludes Belize. `+lncludes Bermuda and St. Pierre and Miquelon. Bladder cancer Chronic obstructive pulmonary disease Total SAM Total SAM t~:~~:~l mortality Table 33. Adjusted estimates of smoking-attributable mortality (SAM) in the Americas, c. 1985 - Chronic obstructive Total Region and Total p;;;e;;:+q Other diseases5 country SAM* Cancer4 and causes adjusted SAM Latin America 64,300 18,600 1,400 13,800 98,100 Andean Area 5,300 1,700 10 1,200 Southern Cone/l 8,200 22,100 4,900 700 4,500 32,100 Brazil 22,100 5,400 300 4,600 Central America1 32,400 600 200 10 100 900 Mexico 7,000 5,100 100 2,000 14,200 Latin Caribbean 7,300 1,300 200 1,400 10,200 Caribbean 1,200 300 30 300 1,900 North America 310,300 43,800 12,000 60,000 426,100 United States 284,400 39,800 11,000 55,000 390,200 Other** 25,800 4,100 1,000 5,100 36,000 All regions of the Americas 375,800 62,700 13,400 74,100 526,000 Source: Pan American Health Organization (1990b). Adjustments were based on 1985 estimates for the United States; U.S. Department of Health and Human Services (1989). Percentages used for upward adjustment for chronic obstructive pulmo- nary disease and other diseases and causes were specific to those diagnostic rubrics. Upward adjustment for cancers was based on lung cancer. *Total for men and women from Table 32. +230% adjustment to compensate for undercounting. *10.4% increase added to adjust for omission of cancers of the kidney and pancreas and for underestimates of smoking- attributable fraction for cancers of oral cavity, esophagus, and larynx. 516.4% increase added to adjust for exclusion of cervical cancer, other types of cardiovascular and respiratory diseases, deaths among newborns due to smoking by the mother, lung cancer deaths due to passive smoking, and deaths from smoking-related fires. IlIncludes Falkland Islands. IExcludes Belize. **Includes Bermuda, Canada, and St. Pierre and Miquelon. 96 Prevalence and Mortality Conclusions 1. Certain sociodemographic phenomena-such as change in population structure, increasing urban- ization, increased availability of education, and entry of women into the labor force-have in- creased the susceptibility of the population of Latin America and the Caribbean to smoking. 2. The lack of systematic surveillance information about the prevalence of smoking in most areas of Latin America and the Caribbean hinders com- prehensive control efforts. Available information reflects a variety of survey methods, analytic schemes, and reporting formats. 3. Available data indicate that the median preva- lence of smoking in Latin America and the Carib- bean is 37 percent for men and 20 percent for women. Variation among countries is consider- able, however, and smoking prevalence is 50 per- cent or more in some populations but less than 10 percent in others. In general, prevalence is highest in the urban areas of the more developed coun- tries and is higher among men than among women. 4. The initiation of smoking (as measured by the prevalence of smoking among persons 20 to 24 years of age) exceeds 30 percent in selected urban areas. Although systematic time series are not available, the data suggest that more recent co- horts (especially of women) in the urban areas of more developed countries are adopting tobacco use at a higher rate than did their predecessors. The smoking epidemic in Latin America and the Caribbean is not yet of long duration or high intensity, and the mortality burden imposed by smoking is smaller than that for North America. By 1985, an estimated minimum of 526,000 smoking- attributable deaths were occurring each year in all the countries of the Americas; 100,000 of these deaths occurred in Latin American and the Carib- bean countries. The estimate of 526,000 deaths annually is conser- vative and is best viewed as the first point on a continuum of such estimates. However, it pro- vides an order of magnitude for the number of smoking-attributable deaths in the Americas. The time lag between the onset of smoking and the onset of smoking-attributable disease is forebod- ing. In North America, a high prevalence of smok- ing, now declining, has been followed by an increasing burden of smoking-attributable mor- bidity and mortality. In Latin America and the Caribbean, rising prevalence portends a major burden of smoking-attributable disease. Prezvlrrm~ md Mortality 97 References ANDERSON, J.E. Smoking during pregnancy and while using oral contraceptives. Data from seven surveys in west- em hemisphere populations. Paper presented at the Inter- national Conference on Smoking and Reproductive Health, San Francisco, California, October 15-17,1985. ANKER, R., HEIN, C. Empleo de la mujer fuera de la agricultura en paises de1 tercer mundo. In: Desigualdades Entre Hombres y Mujeres en 10s Mercados de Trabajo Urban0 de1 Tercer Mundo. Geneva: Oficina International de1 Trabajo, 1987. ASOCIACION DEMOGRAEICA COSTARRICENSE AND CENTERS FOR DISEASE CONTROL. Encuesta Nacionat de Fecundidad y Salud, Costa Rica 1986. San Jose, Costa Rica, 1987. BECERRA, J.E., SMITH, J.C. Maternal smoking and low birthweight in the reproductive history of women in Puerto Rico, 1982. American Iournal of Public Health 781268-272, March 1988. BRUNNER, J.J. Educacidn Superior en America Latina: Cambios y Desafios. Chile: Fondo de Cultura Economica, 1990. CENTERS FOR DISEASE CONTROL. Brazil Demographic and HealthSurvey, 1986. Columbia, Maryland: Institute for Resource Development/Macro International Systems, Inc., 1986. Data tapes. CENTERS FOR DISEASE CONTROL. Guatemala Demo- graphic and Health Survey, 1987. Columbia, Maryland: Institute for Resource Development/Macro International Systems, Inc., 1987a. Data tapes. CENTERS FOR DISEASE CONTROL. Smoking-attributable mortality and years of potential life lost-United States, 1984. Morbidity and Mortality Week/y Reporf 36t42L693-697, October 30,1987b. CENTERS FOR DISEASE CONTROL. Rio de Janeiro, Curitiba, Recife Young Adult Reproductive Health Survey, 1989-1990. Rio de Janeiro: Sociedade Civil Bern Estar Fa- miliar no Brasil, 1990a. Data tapes. CENTERS FOR DISEASE CONTROL. Costa Rica Young Adult Reproductive Health Survey, 1990. San Jose, Costa Rica: Programa Salud Reproductiva, 1990b. Data tapes. CENTERS FOR DISEASE CONTROL. Smoking-attrib- utable mortality and years of potential life lost-United States, 1988. Morbidity and Mortality Weekly Report 40(4):62- 71, February 1,199l. CENTRO LATINOAMERICANO DE DEMOGRAFIA. Bolefin Demografico/Demographic Bulletin. Santiago, Chile: CELADE, Year XXIII, No. 45, January 1990. COLLISHAW, N.E., TOSTOWARYK, W., WIGLE, D.T. Mortality attributable to tobacco use in Canada. Canadian Journa/ of Public Health 79(3):166-169, May-June 1988. DE OLIVEIRA, O., ROBERTS, B. Los antecedentes de la crisis urbana: Urbanizacibn y transformacibn ocupacional en America Latina: 1940-1980. In: Lombardi, M., Veiga, D. teds.) Las Ci&drs en Co@icto: Una Perspectiza Lntinoameric- ana. Montevideo, Uruguay: Centro de Informaciones y Estudios de1 Uruguay, Ediciones de la Banda Oriental, 1989. DE STEFANI, E., CORREA, I'., OREGGIA, F., LEIVA, J,, RIVERO, S., FERNANDEZ, G., DENEO-PELLEGRINI, H., ZAVALA, D., FONTHAM, E. Risk factors for laryngeal cancer. Cancer 60(12):3087-3091, December 15,1987. DE STEFANI, E., MUNOZ, N., ESTEVE, J., VASALLO, A., VICTORA, C.G., TEUCHMANN, S. Mate drinking, alcohol, tobacco, diet and esophageal cancer in Uruguay. Cancer Research 50:426-431, January 15,199O. DOLL, R., PETO, R. The causes of cancer. Quantitative estimates of avoidable risks of cancer in the United States today. Jocmal of the National Cmcer Institute 66(6):1191- 1308, June 1981. ECONOMIC COMMISSION FOR LATIN AMERICA AND THE CARIBBEAN. Transformacidn Ocupacional y Crisis Social en America Latina. Santiago, Chile: Naciones Unidas, 1989. ERNSTER, V.L. Selling cigarettes: How women are used. In: Forbes, W.F., Frecker, R.C., Nostbakken, D. (eds.) Pro- ceedings of the Fifth World Conference on Smoking and Health, Volume II. Ottawa, Canada: Canadian Council on Smoking and Health, 1983. ESCOBEDO, L.G., ANDA, R.F., SMITH P.F., REMINGTON, P.L., MAST, E.E. So&demographic characteristics of ciga- rette smoking initiation in the United States. Implications for smoking prevention policy. /ournal offhe American Med- ical Association 264(12):1550-1555, September 26,199O. ESCOBEDO, L.G., REMINGTON, P.L., ANDA, R.F. Long- term secular trends in initiation of cigarette smoking among Hispanics in the United States. Public Health Reports 104(6):583-587, November-December 1989. FRANCO, E.L., KOWALSKI, L.P., OLIVEIRA, B.V., CURADO, MI'., PEREIRA, R.N., SILVA, M.E., FAVA, A.S., TORLONI, H. Risk factors for oral cancer in Brazil: A case- control study. Internafionaf Jounznl of Cancer 43:992-1ooO,l989. GALLUP ORGANIZATION. The incidence of smoking in Central and Latin America. Conducted for the American Cancer Society. Princeton, New Jersey: Gallup Organiza- tion, April 1988. HAKULINEN, T., HANSLUWKA, H., LOPEZ, A.D., NAKADA, T. Global and regional mortality patterns by cause of death in 1980. tnternational Journal of Eyidemiotogy 15:226-233, June 1986. 98 Prevalence and Mortality HARRIS, J.E. Cigarette smoking among successive birth cohorts of men and women in the United States during 1900-80. ]ournul offhe Nutional CancerInstitute 71(3):473-479, 1983. HAYNES, S.G., HARVEY, C., MONTES, H., NICKENS, H., COHEN, B.H. Patterns of cigarette smoking among Hispanics in the United States: Results from Hispanic Health and Nutrition Examination Survey, 1982-1984. American Jourd of Public Heu[th 80(supplement):47-53, December 1990. INTERNATIONAL AGENCY FOR RESEARCH ON CAN- CER. Tobacco Smoking. IARC Monographs on the Evalua- tion of the Carcinogenic Risk of Chemicals to Humans, Volume 38. Lyon, France: International Agency for Re- search on Cancer, 1986. ISCOVICH, J., CASTELLETTO, R., ESTEVE, J., MUNOZ, N., COLANZI, R.,CORONEL, A., DEAMEZOLA, I., TASSI, V., ARSLAN, A. Tobacco smoking, occupational exposure and bladder cancer in Argentina. Zderdiorzol Jourd of Cancrr 40(6):734-740, December 15,1987. JAMISON, D.T., MOSLEY, W.H. Disease control priorities in developing countries: Health policy responses to epide- miological change. American \ourno/ of Pnblic Health 81(1):15-22, January 1991. JOLY, D.J. Enczlesta Sobre lus Caracteristicas de1 HLbito de Fumar en AmPrica Latina. Washington, DC: Organizacibn Panamericana de la Salud, Oficina Sanitaria Panamericana, Oficina Regional de la Organizacidn Mundial de la Salud. Publicaci6n Cientifica No. 337,1977. JOLY, O.G., LUBIN, J.H., CARABALLOSO, M. Dark to- bacco and lung cancer in Cuba. Jourd ofthe National Cancer Iwtitute 70(6):1033-1039, June 1983. MARTINEZ, I. Factors associated with cancer of the esoph- agus, mouth and pharynx in Puerto Rico. Journal of the Nutionul Cancer lnstifllle 42(6):1069-1094, June 1969. MCFARLANE, C., WARREN, C.W. 1989 jamuicu Contrucep- tiue Prezlalence Surzley: Final Report. Kingston, Jamaica: National Family Planning Board, December 1989. MERRICK, T.W. Population pressures in Latin America. Populufion Bulletin 41(3), July 1986. NATIONAL FAMILY PLANNING BOARD AND CEN- TERS FOR DISEASE CONTROL. Young Adulf Reproductive Health Survey,Jumuicu 1987: Final Report. Kingston, Jamaica: National Family Planning Board, March 1988. OMRAN, A.R. The epidemiologic transition: A theory of the epidemiology of population change. Milbunk Memorial Fwd Qluzrterly 49(4):509-538, October 1971. PALLONI, A. Mortality in Latin America: Emerging pat- terns. Poplrlation and Dmelopment Rezliezo 7(4):623449, De- cember 1981. PAN AMERICAN HEALTH ORGANIZATION. Health Conditions in the Americas, 1990 Edition. Washington, DC: Pan American Health Organization, Pan American Sanitary Bureau, Regional Office of the World Health Organization. Scientific Publication No. 524, 1990a. PAN AMERICAN HEALTH ORGANIZATION. Mortality Database. Washington, DC: Pan American Health Organi- zation, Health Situation and Trend Assessment Program, 199Ob. Database. PAN AMERICAN HEALTH ORGANIZATION. Tobacco or Heulth: Status in the Americas. Washington, DC: Pan Amer- ican Health Organization. Scientific Publication No. 536, 1992. PARKIN, D.M., LAARA, E., MUIR, C.S. Estimates of the worldwide frequency of sixteen major cancers in 1980. ln- ternationa/JournulofCuncer 41(2):184-197, February 15,1988. PETO, R. Influence of dose and duration of smoking on lung cancer rates. In: Zaridze, D., Peto, R. teds.) Tobacco: A Major International Health Huxard. Lyon, France: International Agency for Research on Cancer. IARC Scientific Publication No. 74,1986. PORTES, A. Lotin American C/ass Structures: Their Com- posifionand Change Dzcring theLust Decades. Washington, DC: The Johns Hopkins University, School of Advanced Interna- tional Studies, 1984. RESTREPO, H.E., CORREA, I'., HAENSZEL, W., BRINTON, L.A., FRANCO, A. A case-control study of tobacco-related cancers in Colombia. &&tin of the Pun American Health Oqanization 23(4):40-13, 1989. RICE, D.P., HODGSON, T.A., SINSHEIMER, I'., BROWNER, W., KOPSTEIN, A.N. The economic costs of the health effects of smoking, 1984. The MiIbunk Quarterly 64(4):489-547,1986. SAKAMOTO, C.P., FREIRE, H.S., MORRIS, L. lnvestigqrio Sobre S&de Reproductizla do jovem nu Cidade de Srio Puulo: Relutdrio Final Atlanta, Georgia: Centers for Disease Con- trol, 1991. SCHOENBORN, C.A. Smoking and Other Tobacco Use: United States 1987. U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Center for Health Statistics. DHHS Publication No. (PHS) 89-1597, Vital and Health Statistics, Series 10, No. 169, 1989. SMITH, J.C., WARREN, C.W., GARCIA-NUNEZ, J. The U.S.-Mexico Border: Contraceptive Use and Maternal Health Cure in Perspective, 1979. El Paso, Texas: United States- Mexico Border Health Association, 1983. THE WORLD BANK. World Development Report 2992: The Challenge of Development. New York: Oxford University Press, 1991. Prrz~alencr and Mortality 99 TUKEY, J.W. Exploratory Data Analysis. Reading, Massa- chusetts: Addison-Wesley Publishing Company, 1977. UNITED NATIONS. World Population Prospects, 1988. De- partment of International Economic and Social Affairs. New York: United Nations, 1989. UNITED NATIONS. World Population Prospects, 1990. De- partment of International Economic and Social Affairs. New York: United Nations, 1991. UNITED NATIONS EDUCATIONAL, SCIENTIFIC, AND CULTURAL ORGANIZATION. Statistical Yeurbook/An- nuuire Sfutistique/Anuurio Estadziistico 7989. Paris: Unesco, 1989. U.S. DEPARTMENT OF HEALTH AND HUMAN SER- VICES The Health Consequences of Smoking. The Changing Cigarette: A Report of the Surgeon General. U.S. Department of Health and Human Services, Public Health Service, Office of the Assistant Secretary for Health, Office on Smoking and Health. DHHS Publication No. (PHS) Sl-50156,198l. U.S. DEPARTMENT OF HEALTH AND HUMAN SER- VICES. Reducing the Heulth Consequences of Smoking: 25 Years ofProgress. A Report ofthe Surgeon General. U.S. Department of Health and Human Services, Public Health Service, Cen- ters for Disease Control, Center for Chronic Disease Preven- tion and Health Promotion, Office on Smoking and Health. DHHS Publication No. (CDC) 89-8411,1989. U.S. DEPARTMENT OF HEALTH AND HUMAN SER- VICES. The Health Benefits of Smoking Cessation. A Report of the Surgeon General. U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, Center for Chronic Disease Prevention and Health Promo- tion, Office on Smoking and Health. DHHS Publication No. (CDC) 90-84161990. U.S. OFFICE OF TECHNOLOGY ASSESSMENT. Smoking- Related Deaths and Financial Costs. Health Program, Office of Technology Assessment, U.S. Congress, September 1985. UNIVERSIDADE FEDERAL DA BAHIA AND CENTERS FOR DISEASE CONTROL. Salide e Educucuo Sexual do Jovem: Urn Estudo em Salvador. Salvador, Brazil. February 1989. VALENZUELA, M.S., HEROLD, J.M., MORRIS, L. Encuesta de Salad Reprodzzctizla en Adnltos loz~enes: Grun San&go-I988 Informe Final. Atlanta, Georgia: Centers for Disease Control, July 1989. VASSALLO, A., CORREA, I'., DE STEFANI, E., CENDAN, M., ZAVALA, D., CHEN, V., CARZOGLIO, J., DENEO- PELLEGRINI, H. Esophageal cancer in Uruguay: A case- control study. Journal of the Nationul Cancer Institute 75(6):1005-1009, December 1985. VICTORA, C.G., MUNOZ, N., DAY, N.E., BARCELOS, L.B., PECCIN, D.A., BRAGA, N.M. Hot beverages and oesopha- geal cancer in southern Brazil: A case-control study. Inter- national \ournal of Cancer 39:710-716, 1987. WARNER, K.E. Health and economic implications of a tobacco-free society. Journal of the American Medical Associn- tion 258(15):2080-2086, October 16, 1987. WILKIE, J.W., OCHOA, E. feds.) Statistical Abstract of Latin America, Volume 27. Los Angeles, California: UCLA Latin American Center Publications, 1989. WILKIE, R.W. Latin American Population and Urbanization Arrulysis: Maps und Stutistics, 1950-2982. Los Angeles, Cali- fornia: UCLA Latin American Center Publications, 1984. WORLD HEALTH ORGANIZATION. Manna/ of Interna- tional Statistical Classification of Diseases, Injuries, and Causes of Death. Based on Recommendations of the Ninth Revision Conference, 1973. Geneva: World Health Organization, 1977. WORLD HEALTH ORGANIZATION. A 5 Yeur Action Plan: Smoke free Europe. World Health Organization Regional Of- fice for Europe. Copenhagen, Denmark: World Health Or- ganization, 1988. WORLD HEALTH ORGANIZATION. Report of a WHO consultation on statistical aspects of tobacco-related mortal- ity. Report prepared by World Health Organization, Ge- neva, WHO/TOH/CLH/90.2, October 9-13,1989. WORLD HEALTH ORGANIZATION. Tobacco-attributable mortality: global estimates and projections. Tobacco Alert: 4-5,7, January 1991. 200 Prevalence and Mortality