Lung cancer deaths in the United States from 1979 to 1992: an analysis using multiple-cause mortality data

Transcription

1 O International Epidemiological Assodauon 1998 Printed in Great Britain International Journal of Epidemiology 1998:27: Lung cancer deaths in the United States from 1979 to 1992: an analysis using multiple-cause mortality data David M Mannino, a Earl Ford, b Gary A Giovino c and Michael Thun d Background Methods Results We sought to describe trends in the presence of lung cancer at the time of death in the United States from 1979 to We analysed death certificate reports in the Multiple-Cause Mortality Files compiled by the National Center for Health Statistics, searching for any mention of lung cancer, lung cancer as the underlying cause of death, and comorbid conditions. Of the decedents in the study period, (6.5%) had a diagnosis of lung cancer listed on their death certificates; of these decedents, (91.7%) had lung cancer listed as the underlying cause of death. Decedents with lung cancer listed as being present but not the underlying cause of death were more likely to be male (relative risk [RR] 1.16, 95% confidence interval [CI] : ), and older (RR 4.61, 95% CI : for decedents older than 85 compared to those aged less than 44), but less likely to be black than white (RR 0.88, 95% CI : ). The mortality rate, age-adjusted to the 1980 population, increased 23.0%, from 47.9 per in 1979 to 58.9 per in Over the study period, black men had the highest mortality rates ( per ), followed by white men ( per ), men of other races (37.4-^6.7 per ), white women ( per ), black women ( per ), and women of other races ( per ). Age-adjusted, state specific rates varied threefold, from 30.4 per in Utah to 93.9 per in Nevada. Conclusions We conclude that the underlying cause of death data base, which captures almost 92% of decedents with lung cancer present, accurately tracks lung cancer mortality trends in the US. Mortality rates of lung cancer, which are decreasing among men, continue to increase among women. Keywords Lung cancer, mortality, epidemiology Accepted 15 August 1997 Lung cancer, which was rare prior to the twentieth century, is now the leading cause of cancer mortality among both men and women in the United States. 1 ' 2 The increases in the prevalence * Air Pollution and Respiratory Health Branch, Division of Environmental Hazards and Health Effects. National Center for Environmental Health (NCEH), Centers for Disease Control and Prevention (CDC), Atlanta, Georgia, USA. b Chronic Disease Prevention Branch, Division of Nutrition, National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP), CDC. Atlanta, Georgia, USA. c Office of Smoking and Health, NCCDPHP, CDC, Atlanta, Georgia, USA. Department of Epidemiology and Surveillance Research. American Cancer Sodety. Atlanta, Georgia, USA. Reprint requests: David M Mannlno, 4770 Buford Highway, M/S F-39, Atlanta, Georgia 30341, USA. of lung cancer have been linked to the use of tobacco products, predominantly cigarettes.^ Tobacco smoke causes 85-90% of all lung cancer cases and represents the most preventable risk factor for this malignancy. 7 ' 8 We used the Multiple-Cause Mortality Files (MCMF) to analyse national mortality trends of lung cancer for With this data base we searched for death certificates on which lung cancer was listed as being present but was not necessarily classified as the underlying cause of death (UCD). We have previously determined that only 50% of decedents with pulmonary fibrosis, another chronic pulmonary disease, had that diagnosis listed as the UCD, 9 and sought to determine whether a similar Dnding applied to lung cancer. In addition, we searched for comorbid diseases that were listed as being present with lung cancer. 10 "

2 160 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 1 Age stratification of people who died with lung cancer and from all causes combined from 1979 through 1992, the proportion per deaths involving lung cancer within each age stratum, and the number and percentage of deaths in each age stratum in which lung cancer was listed as the underlying cause of death (UCD). From the Multiple-Cause Mortality Files, National Center for Health Statistics Age (years) < >84 None listed Total All decedents No. of decedents with lung cancer Proportion per decedents Lung cancer as the UCD , Methods The National Center for Health Statistics (NCHS) annually compiles data from all death certificates filed in the US, using the vital records from the individual states. These data contain the "International Classification of Diseases (ICD) codes for the underlying cause of death and up to 20 conditions listed on the death certificates. The data also include demographic and geographic information on the decedents. The International Classification of Diseases, 9th Revision (ICD-9) was implemented in 1979 and was in effect throughout the 14-year period used in this study. 13 The conditions are recorded in two places: on the entity axis, which contains the conditions exactly as reported on the death certificates, and on the record axis, which is edited by a computerized algorithm known as the translation of axes. The automated classification of medical entities algorithm determines the underlying cause of death (UCD) from the conditions and their positions as listed on the death certificates. Quality assurance of the data is maintained by trained nosologists who code conditions at the state level and, in turn, by nosologists at NCHS who periodically review data from a 10% sample of the submitted death certificates. The result of this process is the MCMF. We searched the record axis portion of the MCMF for records containing ICD codes (lung cancer). Within this group of death records, we searched for the UCD and comorbid conditions: obstructive pulmonary disease (ICD-9 codes ); pneumoconioses, including coal worker's pneumoconiosis (ICD-9 code 500), asbestosis (ICD-9 code 501), silicosis (ICD-9 code 503), other inorganic dust pneumoconiosis (ICD-9 code 503), unspecified pneumoconiosis (ICD-9 code 505); and tobacco use disorder (ICD ). We used x 2 analyses to determine the significance of male/female differences in the presence of obstructive lung diseases and pneumoconiosis among decedents with lung cancer. We also searched the MCMF for records containing any mention of obstructive lung diseases or pneumoconioses, and searched for comorbid lung cancer among these subsets, separately. We analysed the data base, which contained all death records with any mention of lung cancer, according to age, race and sex. We used logistic regression models (SAS Institute, Cary, North Carolina, USA) to determine the effect of sex, age and race on whether lung cancer was listed as being present but not the UCD. We did a similar analysis using the number of conditions (four or more versus three or fewer) listed on the death certificate as the dependent variable. The rates reported in this paper are those determined using any mention of lung cancer on the death certificate, unless we specifically state we are using the UCD. The 1980 and 1990 US census data were used to calculate rates for 1980 and 1990, and the intercensal estimate was used to calculate rates for Linear interpolation was used to estimate population in other intercensal years. For ageadjusted, state-specific rates, the 1980 US population was used as the standard. Results Of the people who died in the US from 1979 to 1992, lung cancer was listed on the death records of (6.5%). Table 1 shows the age distribution of decedents with lung cancer and the proportion of deaths that listed lung cancer in each age stratum. Of the deaths, lung cancer was the UCD for (91.7%) deaths and this proportion remained stable, ranging from 91.1% to 92.4%, through the study period. Decedents who had lung cancer listed as being present but not the UCD were more likely to be male and older, but were less likely to be black or of a race other than black or white (Tables 1 and 2). We found 79.6% of the lung cancer decedents had three or fewer causes of death listed on the death certificate. Decedents with four or more causes of death were more likely to be older, male and black (Table 2). Obstructive lung disease, which is related both to lung cancer and smoking, 10 was listed on (12.2%) of the death certificates that listed lung cancer (Table 3). This proportion increased over the study period from 8.7% in 1979 to 14.7% in Among decedents with lung cancer, males were more likely than were females to have obstructive lung disease listed

3 LUNG CANCER IN THE US Table 2 The relative risk and 95% confidence intervals for sex, racial group, and age strata among people who died with lung cancer present but not the underlying cause of death (UCD) compared to decedents with lung cancer listed as the UCD and for decedents having four or more versus three or fewer causes of death listed on the death certificate. From the Multiple-Cause Mortality Files, , National Center for Health Statistics Risk factor Sex Male Race 3 Black Non-black, non-white Age* 1 a *85 Lung cancer listed but not the UCD RR Compared to white decedents b Compared to decedents aged 44 and younger. 95% CI ( ) (0.87^0.90) ( ) ( ) ( ) ( ) ( ) ( ) *4 causes of death RR % CI ( ) ( ) (0.97-O.99) ( ) ( ) ( ) ( ) ( ) Table 3 Age stratification of people who died with lung cancer and with a diagnosis of obstructive lung disease (OLD) or pneumoconiosis (PCN) from 1979 to 1992 and the percentage of deaths involving OLD or PCN within each age stratum. From the Multiple-Cause Mortality Files, National Center for Health Statistics Age (Years) < ^ >84 None listed Total Deaths with lung cancer N Deaths with OLD present N Deaths with pneumoconiosis N (%T (relative risk [RR] 1.22, 95% confidence interval [CI] : ). Over the study period, decedents had a diagnosis of obstructive lung disease listed on their death certificates. Thus, lung cancer was listed on 10.2% of death certificates listing obstructive lung disease. Pneumoconiosis was listed on 6486 (0.34%) of the death certificates that listed lung cancer. This proportion decreased from 0.41% in 1979 to 0.29% in Among decedents with lung cancer, males were much more likely than females to have pneumoconiosis listed (RR 36.04, 95% CI : 28.92^4.48) Over the study period, decedents had pneumoconiosis mentioned one or more times. Thus, lung cancer was listed on 12.4% of death certificates listing pneumoconiosis. Among the 8767 decedents with asbestosis, lung cancer was listed on 2105 (24.0%) of the death certificates; among the decedents with coal worker's pneumoconiosis, lung cancer was listed on 3388 (10.3%) of the death certificates; among the 4885 decedents with silicosis, lung cancer was listed on 418 (8.6%) of the death certificates; and among the 5549 decedents with unspecified pneumoconiosis, lung cancer was listed on 595 (10.7%) of the death certificates. Tobacco use disorder (ICD ) was listed on (2.3%) of the death certificates that listed lung cancer. This proportion increased from 0.6% in 1979 to 3.9% in The total number of people who died with lung cancer present increased by 47.6%, from in 1979 to in 1992 (Table 4). The age-adjusted mortality rates for any mention of lung cancer are shown in Table 4. The age-adjusted mortality rate for lung cancer as the UCD increased 23.8%, from 44.1 per in 1979 to 54.6 per in The age-adjusted mortality rate for lung cancer as the UCD increased 8.0% among men, from 76.4 per in 1979 to 82.5 per in 1987 before decreasing 2.7% to 80.3 per in 1992, whereas the age-adjusted mortality rate for

4 162 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 4 Total counts of deaths with lung cancer (ICD ), counts among men and among women, and death rates per population, age-adjusted to the 1980 population, by year from 1979 to From the Multiple-Cause Mortality Files, National Center for Health Statistics Year Total Total deaths with lung cancer* Ageadjusted rates Deaths with lung cancer among males Age-adjusted rates among males Deaths with lung cancer among females Age-adjusted rates among females a 140 decedents with no age listed are excluded from this Table. Rates per 100,000 Population 140 Black Males White Males eo 60 Other Males 40 Black Females 20 Other Females Year Figure 1 Age-adjusted mortality rates of people with lung cancer present (ICD-9 codes ) per population, stratified by sex and race, in the US from 1979 to From the Multiple-Cause Mortality Files lung cancer as the UCD increased 75.6% among women, from 20.5 per in 1979 to 36.0 per in Over the study period, black men had the highest mortality rates ( per ), followed by white men ( per ), men of other races ( per ), white women ( per ), black women ( per ), and women of other races ( per , Figure 1). Among both white men and white women the age-specific mortality rates of lung cancer were higher in older age strata (Figures 2 and 3). Similar patterns were seen among blacks and people of other races (data not shown). Within certain age strata, the mortality rate of lung cancer

5 LUNG CANCER IN THE US Rates per 100,000 Population > 85 Figure 2 Age-specific mortality rates among white men with lung cancer (ICD-9 codes ) present, per population, in the US from 1979 to From the Multiple-Cause Mortality Files Rates per 100,000 Population > 85 Figure 3 Age-specific mortality rates among white women with lung cancer (ICD-9 codes ) present, per population, in the US from 1979 to From the Multiple-Cause Mortality Files

6 164 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY decreased over the later years of the study period. This decrease was noted among white men age 35-64, among black men age 35-54, and among white women age 35^4. Annual, age-adjusted, state-specific rates (averaged over the 14-year study period) varied threefold, from 30.4 per in Utah to 93.8 in Nevada (Table 5). Discussion Our analysis of lung cancer mortality in the US differs from previous analyses of this problem in that we used the MCMF while previous analyses used the UCD. 1 ' 14 ' 15 With the MCMF we could search for any mention of lung cancer in the death records. We found that 91.7% of the decedents had lung cancer classified as the UCD, and this proportion was stable through the study period. We determined, however, that males and older people were more likely to have lung cancer noted as being present but not the UCD, whereas blacks and people of non-black, non-white races were less likely to have lung cancer noted as being present but not the UCD. The reasons for these findings are unclear, but may be related, in part, to how completely the physician completed the death certificate or how many other chronic medical conditions the decedent had listed (Table 2). From these data we cannot determine how important lung cancer was in those deaths where it was not the UCD. Possibilities could range from the lung cancer being in remission, either because of surgery or other therapy, and thus not being important, to lung cancer being very important, such as deaths attributed to obstructive lung disease, pulmonary embolism or sepsis. National estimates of the effects of smoking, such as the 'Smoking-attributable mortality, morbidity, and economic costs' (SAMMEC) use the UCD as the basis for the estimates. 16 Using the UCD, however, may miss some decedents in whom lung cancer was an important factor in the death but not listed as the UCD, and this under-ascertainment differs across age, race and gender strata. Overall, though, the UCD captures most decedents with lung cancer and should provide adequate data for following long-term mortality trends. This is an important finding because most national databases include only the UCD, and these databases have been used to track smokingrelated cancers Information on smoking, the most common risk factor for lung cancer, is not routinely listed on death certificates. Although there is an ICD-9 code for tobacco use disorder (ICD ), we found that this code appeared on only 2.3% of the death certificates listing lung cancer; this low percentage clearly represents underreporting. This underreporting of tobacco use disorder is similar to that reported by researchers in Wisconsin, who found that this code (ICD ) was listed on only 131 of 2407 (5.4%) death certificates in which lung cancer was the UCD. 19 This finding can be contrasted with results from a study from Oregon, where physicians have to answer the question on the death certificate 'Did tobacco use contribute to the death?' In that study physicians responded that in 2772 of 3560 (77.9%) lung cancer deaths tobacco use definitely or probably contributed to the death. 20 Occupational exposures are thought to be a considerable cause of lung cancer among some workers, with as many as 40% of lung cancers in certain groups of workers possibly attributable to chemicals or dusts such as asbestos, nickel, arsenic and Table 5 Annual age-adjusted rates per population of deaths with lung cancer (ICD ) from 1979 to 1992 by state. From the Multiple-Cause Mortality Files, National Center for Health Statistics State Utah New Mexico Idaho Wyoming Minnesota South Dakota Nonh Dakota Wisconsin Colorado Iowa Nebraska Kansas Montana Hawaii New York Connecticut Massachusetts Illinois Pennsylvania New Jersey New Hampshire Rhode Island Michigan California Vermont Oklahoma Mississippi Texas Alabama South Carolina Missouri Arizona Oregon Maine Nonh Carolina Ohio Arkansas Washington Indiana Georgia Virginia Florida West Virginia Maryland Louisiana Kentucky Tennessee Delaware District of Columbia Alaska Nevada Age-adjusted mortality rate

7 LUNG CANCER IN THE US chromium. 21 ' 22 We found that occupational lung diseases were mentioned as comorbid factors in only 0.3% of the lung cancer deaths, which is less than the estimated 15% of lung cancers among men and 5% among women thought to be related to occupational exposures. 4 The lower prevalence we found may be related to a reporting bias and lack of exposure or occupational history data. Conversely, we found that obstructive lung diseases were listed on 12.2% of the death certificates, and increased in prevalence over the study period. We found that both obstructive lung diseases and pneumoconiosis were more likely to be reported among male decedents than among female decedents. These findings are probably related to a larger number of males historically working in the mining industry and a larger proportion of males smoking, although a reporting bias cannot be excluded. We were not able to evaluate exposures such as radon, 23 ' 24 environmental tobacco smoke, diesel engine exhaust, other occupational exposures or air pollution 14 in this analysis. We determined that while the total number of deaths with lung cancer increased steadily through the study period, the age-adjusted mortality rate increased through 1990, and decreased slightly in 1991 and The trends in mortality rates, however, varied by sex and age group. Among men, these rates peaked in 1987 and decreased steadily through 1992, whereas among women these rates increased steadily through the study period. While decreases in current smoking occurred from 1965 to 1991 among US adults, the decrease has been greater among men. 25 The prevalence of ever smoking (combined former smoking and current smoking) among men peaked in the late 1960s, whereas this prevalence among women did not peak until the mid to late 1980s. 25 If the trends among women mirror those among men, with a lung cancer peak occurring approximately 20 years after the peak in the rate of ever smoking, we would expect the age-adjusted mortality rate with lung cancer among women to peak between 2005 and 2010, as has been predicted by other researchers. 15 We also detected differences in lung cancer mortality rates between racial groups (Figure 1), and between different states. The highest rates of lung cancer mortality have consistently occurred among black males, who also have had the highest rate of current smoking. 25 The differences between the states are also probably a reflection of current and former smoking rates in these states, as data have shown the highest prevalence of current smoking to be in Nevada, and the lowest to be in Utah, corresponding to the states with the highest and lowest lung cancer mortality rates. 25 In 1986, Horm and Kessler reported decreasing age-specific lung cancer mortality rates in white males age years old and years old. We found a similar trend in our analysis, along with decreases among black men age years old and among white women age years old (Figures 2 and 3). The reasons for this decrease in younger age groups is unclear but is probably related to a lower prevalence of smoking. 27 Brown and Kessler used an age-period-cohort model to predict lung cancer mortality through the year The results of their models, comparing estimated age-adjusted mortality rates for to actual rates for , predicted a 2.8% increase in mortality among men and a 49.5% increase in these rates among women. Our results, comparing actual age-adjusted mortality rates for to rates for showed a 5.1% increase in deaths with lung cancer and a 4.4% increases in deaths with lung cancer as the UCD among men. We detected a 50.9% increase in deaths with lung cancer and a 50.0% increase in deaths with lung cancer as the UCD among women. Mortality rates reflect two underlying components: incidence and case-fatality rates. Lung cancer incidence, as monitored among the Surveillance, Epidemiology, and End Results (SEER) populations, 28 increased over the study period, driven by the continued increase of lung cancer incidence among women. Case-fatality rates of lung cancer, which are related to cell type, cancer stage, and physical status of the patient, did not change appreciably over the study period. 14 Using death certificate data for epidemiological studies has certain limitations. Death certificate data are not independently validated, and some degree of misclassification inevitably occurs. However, insofar as misclassification is likely to be constant over time, the interpretation of time trends should be fairly accurate. After reviewing the SEER data, Gilliland and Samet concluded that misclassifications bias, in which lung cancer was present but not listed on the death certificate, may have affected overall lung cancer rates and lung cancer rates in the elderly. 1 Nevertheless, the death certificates from individuals who were part of the SEER program showed that of those having lung cancer the proportion of lung cancer listed as the underlying cause of death remained fairly constant at 83-84% for people who were younger than 55 years at the time of diagnosis and at 74-76% for individuals who were 85 years or older at the time of diagnosis between 1973 and Because death certificates do not provide histological information we cannot examine mortality rates by histological subtypes of lung cancer, which are changing over time. 29 This type of information could have been helpful in understanding what malignant tumours occurred in the 591 decedents in this study who were younger than 25 and thus very unlikely to have a tobacco-related malignancy. In conclusion, we determined that the underlying cause of death data base, which captures almost 92% of decedents with lung cancer present, accurately tracks lung cancer mortality trends in the US. Because 85-90% of lung cancers are attributed to smoking, future lung cancer mortality trends will continue to be dominated by smoking trends. Additionally, we conclude that mortality rates of lung cancer, which are decreasing among men, continue to increase among women, which is probably related to historical smoking prevalences in these populations. References 1 Gilliland FD, Samet JM. Lung cancer. In: Doll R, Fraumeni JF Jr, Muir CS, (eds). Cancer Surveys- Trends in Cancer Incidence and Mortality. Vol 19/20. Plainview, NY: Cold Spring Harbor Laboratory Press, Devesa SS, Blot WJ, Stone BJ etai. Recent cancer trends in the United States. J Nail Cancer Inst 1995;87: Doll R, Hill AB. Smoking and carcinoma of the lung. Preliminary report. BMJ 1950,11: Doll R, Peto R. The Causes of Cancer. New York, NY: Oxford University Press, U.S. Department of Health and Human Services. The Health Consequences of Smoking. Cancer. A Report of the Surgeon General. U.S. Department of

8 166 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Health and Human Services, Public Health Service, Office of the Assistant Secretary for Health, Office on Smoking and Health, Washington, DC. DHHS Publication No. (PHS) Thun MJ, Day-Lally CA, Calle E et al. Excess mortality among cigarette smokers: changes in a 20-year interval. Am J Public Health 1995; 85: Wu-Williams AH, Samet JM. Lung cancer and cigarette smoking, In. Samet JM (ed.). Epidemiology of Lung Cancer. New York: Marcel Dekker Inc., Doll R, Peto R, Wheailey K, Gray R, Sutherland T. Mortality in relation to smoking: 40 years' observations on male British doctors. Br Med J 1994; 309: Mannino DM, Etzel RA, Parrish GB. Pulmonary fibrosis deaths in the United States, : an analysis of multiple-cause mortality data. Am J Respir Cnt Care Med 1996,153: Skillrud DM, Offord KP, Miller RD. Higher risk of lung cancer in chronic obstructive pulmonary disease. Ann Intern Med 1986; 105: ' Simonato L, Vineis P, Fletcher AC. Estimates of the proportion of lung cancer attributable to occupational exposure. Cardnogenesis 1988;9: Berry G, Newhouse ML, Turk M. Combined effect of asbestos exposure and smoking on mortality from lung cancer in factory workers. Lancet 1972,11: U.S. Public Health Service. International Classification of Diseases, Ninth Revision U.S. Government Printing Office, Washington, DC: DHHS Publication No. (PHS) Beckett WS. Epidemiology and etiology of lung cancer Chn Chest Med 1993;14:1-15. "Brown CC, Kessler LG. Projections of lung cancer mortality in the United States: J Nat! Cancer Inst 1988;80: Nelson DE, Kirkendall RS, Lawton RL et al Surveillance for smokingattributable mortality and years of potential life lost, by state United States, 1990 MMWR CDC Surveillance Summaries 1994:43 (No. SS-l):l-8. l7 Boffetta P, LaVecchia C, Levi F, Lucchini F. Mortality patterns and trends for lung cancer and other tobacco related cancers in the Americas, lnt J Epidemiol 1993,12: LaVecchia C, Boyle P. Trends in the tobacco-related cancer epidemic in Europe. Cancer Detection and Prevention 1993; 17: Pezzino G, Remington PL, Anderson HA et al. Smoking as a contributing cause of death in Wisconsin, United States, Tobacco Control 1994:3: McAnulty JM, Hopkins DD, Grant-Worley JA et al. A comparison of alternative systems for measuring smoking-attributable deaths in Oregon, USA. Tobacco Control 1994:3: Vineis P, Simonalo L. Proportion of lung and bladder cancer in males resulting from occupation: a systematic approach. Arch Environ Health 1991;46: Coultas DB. Other occupational carcinogens. In: Samet JM (ed.). Epidemiology of Lung Cancer. New York: Marcel Dekker Inc., US Environmental Protection Agency. Technical Support Document for the 1992 Citizen's Guide to Radon. Washington, DC: USEPA 400- R , National Institutes of Health. Radon and Lung Cancer Risk: A Joint Analysisofll Underground Miners Studies. NIH Pub. No , Giovino GA, Schooley MW, Zhu BP et al. Centers for Disease Control and Prevention Surveillance for selected tobacco-use behaviors United States, MMWR 1994;43(No. SS-3):l Horm JW, Kessler LG. Falling rates of lung cancer in men in the United States. Lancet 1986:1: Shopland DR. Tobacco use and its contribution to early cancer mortality with a special emphasis on cigarette smoking. Environ Health Perspecl 1995;103(Suppl 8):134-^1. 28 National Cancer Institute. Cancer Statistics Review NIH Pub. No , Travis WD, Lubin J, Ries L, Devesa S. United States lung carcinoma incidence trends: declining for most histologic types among males, increasing among females. Cancer 1996;77: