- Email: [email protected]
Measuring the Projected Public Health Impact of Lung Cancer Through Lifetime and Age-Conditional Risk Estimates
Measuring the Projected Public Health Impact of Lung Cancer Through Lifetime and Age-Conditional Risk Estimates RAY M. MERRILL, PhD, MPH
PURPOSE: Lifetime and age-conditional probability (risk) estimates of developing lung cancer in the United States are presented by age, race, and gender. Effects on the risk estimates of an aging population and changing tobacco use are identified. METHODS: Risk estimates were derived by applying cross-sectional, population-based incidence rates of malignant lung cancer and mortality rates from other causes to a hypothetical cohort. The cohort was aged through a double-decrement life table to determine the expected proportion of the population that would develop the disease across age intervals. Incidence and mortality data were obtained from the Surveillance, Epidemiology, and End Results (SEER) Program and the National Centers for Health Statistics, respectively. RESULTS: Among all cancers, risk estimates of developing lung cancer within 10 years, conditioned on being free of the disease at age 50, 60, or 70, ranked second to prostate cancer for men and second to breast cancer for women. For men, despite higher incidence rates of lung cancer for blacks than whites across most age groups, the risk of developing this disease over the life-span becomes similar, because white men are more likely to live to older ages where lung cancer is common. For women, lung cancer incidence rates are similar between Whites and Blacks, but an older age distribution among white women explains their greater lifetime risk of being diagnosed with the disease. Changes in the age distribution between the mid 1970s and the mid 1990s had little impact on the short-term risk estimates of developing lung cancer for younger ages but had a large influence on long-term risk estimates, particularly for the older age groups. CONCLUSIONS: Declining lung cancer among younger age groups may be attributed to declining tobacco use among the cohorts, but several more years may be required before the trends begin to fall in older age groups, particularly in women. In the meantime, an aging population is contributing to more people being diagnosed with lung cancer. Consequently, the projected risk of developing lung cancer will remain high for several years to come. Ann Epidemiol 2000;10:88–96. 2000 Elsevier Science Inc. All rights reserved. KEY WORDS:
Lung Neoplasms, Tobacco, Lifetime Risk, Age-Conditional Risk, Burden, Life Table.
INTRODUCTION As the average life expectancy in the United States continues to increase (1), a rising proportion of the population is being affected by chronic diseases of the elderly, such as cancer. This has direct public health implications in terms of increasing needs for health care resources, services, and planning (2, 3). However, considerable efforts have been made to prevent cancer by modifying diet, physical activity, cigarette smoking and other risk factors earlier in life. The combined effects of an aging population and prevention efforts are reflected in annually reported estimates of cancer incidence in the US (4). Likewise, reported estimates of From the Department of Health Science, Brigham Young University, Provo, UT. Address reprint requests to: Ray M. Merrill, Ph.D., M.P.H., Brigham Young University, Department of Health Science, 213 Richards Building, Provo, UT 84602. Received January 4, 1999; revised June 15, 1999; accepted June 18, 1999. 2000 Elsevier Science Inc. All rights reserved. 655 Avenue of the Americas, New York, NY 10010
the projected probability (risk) of developing cancer (4, 5) are also influenced by both an aging population and preventive efforts. Lifetime and age-conditional risk estimates, commonly reported statistics in the popular press and scientific literature, provide an indication of the proportion of the population expected to develop disease. Lifetime risk reflects the proportion of the infant population expected to develop the disease over the average life expectancy in the population. In contrast, age-conditional risk is tailored to a person’s particular age and indicates the proportion of people expected to be diagnosed with the disease within a specified time frame among those free of the disease at a given age. These measures combine the effects of both cross-sectional incidence of the disease and mortality from other causes across age groups. If we compute lifetime and age-conditional risk estimates from incidence and mortality data in 1975–77 and then again from data in 1993–95, the estimates are likely to differ 1047-2797/00/$–see front matter PII S1047-2797(99)00051-4
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
Selected Abbreviations and Acronyms SEER ⫽ Surveillance, Epidemiology, and End Results NCI ⫽ National Cancer Institute
because of changes over time in the age distribution of the population and in the incidence and mortality rates upon which these risk measures are based. In this paper, we report lifetime and age-conditional risk estimates of developing lung cancer based on cross-sectional incidence and mortality data from six three-year time periods, beginning in 1975–77. We also perform calculations for each group of years using the observed rates of lung cancer incidence in those years, while holding the mortality rates constant at their 1975–77 levels. This adjustment shows how changing mortality has influenced the lifetime and age-conditional risk estimates of developing lung cancer. Lung cancer represents an important disease for analysis
89
because of its widespread impact, particularly among the elderly, and because it is dominated by a modifiable risk factor, tobacco. Comparisons in lifetime and age-conditional risk estimates will be made between males and females, Whites and Blacks. Such comparisons are of particular interest because of the large differences observed in life expectancy and tobacco use among these subgroups of the population.
MATERIALS AND METHODS Lifetime and age-conditional risk estimates are derived by applying population-based incidence rates of lung cancer and mortality rates from other causes to a hypothetical cohort. This cohort is aged through a double-decrement life table to determine the number developing the disease at each age interval. A detailed description of the methodology is given elsewhere (6, 7). Because the estimates are based on population data, they provide a projection of the average
TABLE 1. Estimated risk (number per 100,000) of developing selected cancers (invasive only) over the next ten years for those who are free from disease at age 40, 50, 60, or 70 by sex and race (White, Black)a Age for white men Cancer All cancers Prostate Lung & bronchus Colon/rectum Bladderb Lymphomas Oral & pharynx Stomach Pancreas Leukemia Esophagus Liver
Age for black men
40
50
60
70
40
50
60
70
1945 115 215 168 108 211 123 40 32 76 26 24
6342 1681 1068 696 400 331 320 131 146 162 104 61
16,255 6508 2990 1767 1159 549 505 295 318 357 227 159
24,855 9983 4342 2840 1829 813 543 517 580 634 314 264
2765 278 488 276 57 322 278 86 68 59 84 85
10,113 3262 2209 914 241 297 655 272 317 152 359 192
21,433 10,505 4268 1651 450 326 722 505 542 290 534 272
31,236 15,405 5337 2807 918 401 430 873 789 408 495 278
Age for white women Cancer All cancers Breast Lung and bronchus Colon/rectum Corpus uteri Ovarian Cervix uteri Bladderb Lymphomas Oral & pharynx Stomach Pancreas Leukemia Esophagus Liver a b
Age for black women
40
50
60
70
40
50
60
70
3097 1516 187 142 187 187 134 34 102 47 18 26 48 4 7
6389 2669 812 489 559 323 134 126 221 117 48 99 107 24 25
10,723 3600 1924 1233 900 475 135 309 412 205 112 254 204 67 54
14,688 4188 2510 2195 1002 538 114 494 686 241 212 483 340 99 111
3137 1504 303 234 104 104 191 9 88 70 39 46 48 31 11
6308 2379 979 773 307 204 244 61 143 170 90 175 45 77 35
9911 2759 1912 1408 653 350 245 161 194 224 190 384 189 187 80
13,145 3298 1972 2279 642 470 244 403 344 154 399 726 246 197 124
Based on 1993–95 incidence rates from SEER and mortality rates from NCHS. Includes both in situ and invasive cancer.
90
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 1. Age-specific lung cancer incidence rates by gender, race, and age at diagnosis.
disease risk expected in the population over the next x years among those currently at risk of developing the disease at a given age. While these measures provide somewhat specific information when conditioned on a particular age, they do not incorporate individualized information as do other risk models (e.g., family history of disease, screening status, risk factor exposure) (8–13). Lung cancer incidence data were obtained from tumor registries from five states (Connecticut, Iowa, New Mexico, Utah, and Hawaii) and four metropolitan areas (Atlanta, Georgia; Detroit, Michigan; Seattle, Washington; and San Francisco, California). These registries are included in the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (NCI), and represent about 10% of the US population (5). Mortality data were obtained from death certificates among people identified in the SEER areas, as recorded by the National Centers for Health Statistics. Rates were derived from combining the respective incidence and mortality data with population estimates from the US Bureau of the Census. The NCI’s computer program for estimating the risk of developing cancer was used to derive the risk estimates (14). Three
years of data (e.g., 1993–95) were used in the computations in order to provide more stable estimates. Although lifetime risk estimates receive much attention in the popular press, short-term, age-conditional risk estimates are less restricted by long-term data extrapolations. They can be used to show the expected relative burden of the disease among select age groups where cancer is common. However, lifetime risk estimates may be effective at encouraging preventive behavior, particularly among young people where the benefits of prevention are most effective. In the current study we consider both short- and long-term risk estimates of developing lung cancer.
RESULTS To provide a context for our analysis, it is wise to first consider the risk of being diagnosed with lung cancer relative to other selected cancers over 10-year intervals for ages 40, 50, 60, and 70 (Table 1). The risk of developing any cancer is greater among men than women, except for women aged 40, primarily due to breast cancer. White men experi-
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
91
FIGURE 2. Number at risk of developing lung cancer over the life span by gender, race, and time period.
ence a higher cancer risk than black men only with respect to bladder cancer, lymphomas, and leukemia. White women tended to show a greater risk than black women for cancers of the breast, corpus uteri, ovary, lymphomas, and leukemia. The 10-year risk of lung cancer ranks first among men aged 40; ranks second to prostate cancer for men aged 50, 60, and 70; and ranks second to breast cancer for women aged 40, 50, 60, and 70. A primary input to computing risk measures are the disease incidence rates. Figure 1 shows the lung cancer incidence rates between whites and blacks by age, gender, and time period (1975–77, 1993–95). For men, incidence rates are lower when based on data from 1993–95, as opposed to 1975–77, for ages through about 59 for Blacks and 64 for Whites. In older ages, the incidence rates tend to be higher when based on the 1993–95 data. Black men generally have higher incidence rates. For women, incidence rates are lower when based on data from 1993–95, compared to 1975–77 for ages through 49. However, the opposite is true for subsequent ages. The incidence rates are similar for both Blacks and Whites, except in 1993–95 across ages 65 to 80.
From about age 50 and older, the incidence rates are much higher when based on the later time period. Figure 2 shows the number at risk of developing lung cancer across age for cohorts of 10 million people. White men and women are more likely to live to older ages where lung cancer becomes common. White men show the largest improvement in life expectancy between 1975–77 and 1993–95. White women show a greater improvement in life expectancy than black women in the older age groups. Despite the fact that black men tend to have higher incidence of lung cancer than white men, the larger proportion of white men living to older ages where lung cancer becomes common will make the actual burden of this disease more equivalent. Similarly, although white women have similar or slightly higher lung cancer incidence rates than black women, because the former are living to older ages, their risk of developing the disease is higher. To illustrate the dynamics of lung cancer risk, we present the cumulative risk of developing this disease over the expected life-span by race, gender, and time period (Figure 3). The cumulative risk of developing lung cancer has a
92
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 3. Cumulative risk estimates of developing lung cancer by age, race, gender, and time period. *Adjusting for 1975–77 mortality.
sigmoid shape, beginning very low in early adulthood. Risk rises throughout life and plateaus in the oldest age groups. In men, the risk remains consistently lower in white men, although the difference begins to decrease around age 70. The greater improvement in life expectancy between 1975–77 and 1993–95 for white men than black men (see Figure 2) is illustrated by a much larger impact on the risk measures for white men. In women, Blacks show higher risk through about age 60, after which the risk of lung cancer becomes much greater for Whites. The difference is more pronounced in the period 1993–95, both because of higher incidence rates in whites and greater improvement in life expectancy, particularly in the older age groups. Age-conditional risk estimates provide both relevant information to people approaching the ages where lung cancer becomes common and an indication of the relative burden of the disease in specific age groups. Figure 4 presents risk estimates of developing lung cancer within 10, 20, 30, and 35⫹ years among people free of the disease at age 50, by race, gender, and time period. Changes in the risk estimates occur because they are based on cross-sectional incidence and mortality data from seven 3-year time periods. The
broken line represents the risk measures adjusting for 1975–77 mortality (i.e., controlling for changing life expectancy). Among those alive without lung cancer at age 50, black men have much higher risk estimates of developing the disease in 10, 20, 30, and 35⫹ years. The greatest differences on an absolute scale occur for 20- and 30-year risk. In contrast, the conditional 10- and 20-year risk estimates are similar for white and black women, but the conditional 30- and 35⫹ year risks are higher for white women. In Figure 5, cumulative risk of developing lung cancer in the next x years is presented for the population without diagnosed lung cancer at ages 50, 60, 70, and 80 by age, race, gender, and time period (1975–77, 1993–95). The broken lines adjust for 1975–77 mortality. The 10-year risk of developing lung cancer is greatest among those free of the disease at age 70 (except among black women where the age is 60); 20 and 30-year risk of developing lung cancer is greatest among those free of the disease at age 60; and 40 year risk of developing lung cancer is greatest among those free of the disease at age 50. Similar results occur if the mortality rates in 1975–77 were used in deriving the risk estimates in 1993–95. The changing age distribution
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
93
FIGURE 4. Risk of developing lung cancer in 10, 20, 30, or 35⫹ years for those free of the disease at age 50 by gender, race, and time period. *Adjusting for 1975–77 mortality.
had its greatest effect on the risk estimates for people free of lung cancer at age 50, followed by ages 60, 70, and then 80 (except among black women where it appears to have had its greatest effect on those at age 80).
DISCUSSION The prominent role of lung cancer in the US is well known, representing the second most frequently diagnosed cancer and the leading cause of cancer death for both men and women (4). This is particularly disconcerting since most lung cancer cases and 90% of subsequent deaths are attributed to cigarette smoking (15). Hence, any changes in smoking behavior will translate into changes in lung cancer incidence and mortality. Since 1964, cigarette smoking has declined in the US. For ages 18 years and older, the percentage of smokers fell from 42% in 1965 to 25% in 1995, with decreases consistently seen across age, race, and gender (16). A number of events have contributed to this decline (e.g., scientific evidence of a link between cigarette smoking and lung can-
cer, tobacco warning labels, cigarette tax, and the nonsmoker’s rights movement). Accompanying the decline have been lower lung cancer rates among younger age groups (see Figure 1). However, because of the long latency period for this disease (i.e., about 30 years) (17), it may take several more years before the trends in the oldest age groups begin to decline. The competing effects of changing smoking habits and an increasing age distribution has had a profound effect on the estimated lifetime and age-conditional risk estimates of developing lung cancer. For white men, had mortality rates stayed constant from 1975–77 to 1993–95, the estimated risk of developing lung cancer would have remained lower over the entire life-span, not just among age groups below 70, such that by ages 95⫹ the probability dropped from 7.3% to 7.0%. However, because of the older age distribution for white men, the actual lifetime risk of developing lung cancer increased from 7.3% to 8.3%. For black men, had there been no change in life expectancy over the study period, only about half of the observed increased lifetime risk of developing the disease would have occurred: 7.5% to 8.0% rather than 7.5% to 8.5%. For both white and black women,
94
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 5. Cumulative risk of developing lung cancer in the next x years for those free of the disease at select ages by race and gender, 1993–95. *Adjusting for 1975–77 mortality.
the lifetime risk of developing lung cancer increased dramatically, as explained by sharp increases in lung cancer incidence. The increasing life expectancy only explained a relatively small part of this increase. Had there been no change in life expectancy over the study period, the lifetime risk of developing lung cancer would have increased from 2.7% to 5.5% rather than from 2.7% to 5.9% for white women, and from 2.5% to 4.7% rather than from 2.5% to 4.8% for black women. It is interesting to observe that, despite higher lung cancer incidence rates for black men than white men at almost every age, the older age distribution among white men resulted in a fairly similar lifetime risk estimate of developing the disease. In the case of women, although the lung cancer incidence rates are similar between whites and blacks, the older age distribution among white women suggests that they should have a greater probability of being diagnosed with lung cancer over a lifetime, which was the case. In a related study, it was shown that although the incidence rate of developing prostate cancer was higher for black men than for white men across all ages, the lifetime risk of developing
this disease was lower for Blacks than Whites because fewer blacks were living to ages where prostate cancer becomes common (18). Similarly, invasive incidence rates for all cancers combined in the period 1993–95 were higher for Blacks than Whites across all age groups, but a projected 41% of Blacks compared to 45% of Whites are expected to develop an invasive cancer over their lifetime (14). Invasive cancers of the breast, colon, and rectum, as well as Hodgkin’s disease also have higher age-specific incidence rates among Blacks than Whites, but Whites are more likely to develop these diseases over their longer life expectancy (5, 14). Hence, the potential public health impact for certain diseases may be less for blacks than whites despite higher agespecific incidence rates. Notwithstanding declines in cigarette smoking and subsequent declines in lung cancer rates among younger age groups, increasing life expectancy and the fact that lung cancer rates in the older age groups continue to increase indicate that the lifetime risk estimates of developing this disease will continue to increase. Hence, the overall burden of lung cancer is expected to remain high for some time to
AEP Vol. 10, No. 2 February 2000: 88–96
come. Yet, when we condition the risk measures on being free of the disease at age 50, then 10-, 20-, 30-, and ⭓ 35year risk estimates, based on data from more recent years show some decline in the projected burden (Figure 4). These declines begin earlier for 10 years, then 20 years, and so on. Thus, these conditional risk estimates more clearly illustrate the benefits of cigarette smoking cessation than do the lifetime-risk measures. Changes in the age distribution over the study period had little impact on the short-term probability estimates of developing lung cancer for younger ages, but had a noticeable influence on long-term risk estimates, particularly for the older age groups. Additionally, the 10-, 20-, 30-, and 40-year risks of developing lung cancer tended to be greatest among those at ages 70, 60, 60, and 50, respectively. This was the case whether or not the estimates were based on 1975–77 or 1993–95 data. It should be noted that the risk estimates used in this analysis are based on population data. The data is not distinguished by select subgroups of the population that may be at elevated risk of developing lung cancer, such as those with a family history, those exposed to certain environmental factors (e.g., asbestos, radon), or those who are current or past tobacco smokers. Hence, the risk measures presented apply to the average individual in the population, not to those considered to be at high risk of developing lung cancer. However, the estimates do provide tailored information for specific ages and indicate the population disease burden of lung cancer which may be useful for guiding personal health decisions and public health policy actions. Probability measures of developing disease may be influenced by lead time. This is because they are based on disease incidence rates which are susceptible to lead time effects. A lead time effect results when a new screening technique is rapidly diffused into the population, advancing the time of diagnosis prior to when it would have been symptom detected without altering the time of death (19, 20). This will cause an artifactual increase in the incidence rates, but these rates will eventually return to their secular trend. The time the incidence rates remain inflated depends on both how rapidly the screening program is diffused into the population and the length of the lead time between screen detection and symptom detection. The use of short-term probability measures of developing disease may be useful for identifying projected disease burden when lead time affects the estimates, such as for prostate cancer, because they reflect the actual public health burden. Yet screening has been discouraged in the case of lung cancer, because randomized studies have failed to identify its efficacy (21). Without significant changes in screening behavior during the study period, changes observed in the lifetime and ageconditional risk estimates are primarily the result of changing patterns in smoking behavior and increases in life expectancy.
Merrill RISK ESTIMATES OF LUNG CANCER
95
Advances in lung cancer treatment have shown little impact on changing the prognosis for people diagnosed with this disease. Lung cancer is an extremely virulent disease, particularly the small-cell type. The five-year relative survival for all lung carcinomas is 13.9%, whereas for small cell it is 4.6% (22). This explains why incidence and mortality trends for lung cancer are similar (5). Because almost everyone diagnosed with lung cancer dies from the disease, the use of the term burden in this paper to refer to the extent of the population being affected also has reference to the extent of the population facing death. For other diseases where cure is common (e.g., leukemia, prostate cancer, or breast cancer), the type of burden referred to is not as closely related to mortality.
REFERENCES 1. Vital Statistics of the United States, 1994. Mortality. Preprint of v. 11, part A, sec. 6 life tables. Hyattsville, MD: National Center for Health Statistics; 1998. 2. Brown ML, Fintor, L. The economic burden of cancer. In: Greenwald P, Kramer BS, Weed DL, eds. Cancer Prevention and Control. Bethesda, MD: National Cancer Institute; 1995;69–81. 3. Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause 1990–2020: Global burden of disease study. The Lancet. 1997;349:1498–1504. 4. Landis SH, Murray T, Bolden S, Wingo PA. Cancer Statistics, 1998. CA Cancer J Clin. 1998;48:6–29. 5. Ries LA, Kosary CL, Hankey BF, Miller BA, Harras A, Edwards BK, eds. SEER Cancer Statistics Review, 1973–1994. NIH Pub. No. 972789. Bethesda, MD: National Cancer Institute; 1997. 6. Feuer EJ, Wun LM, Boring CC, Flanders WD, Timmel MJ, Tong T. The lifetime risk of developing breast cancer. J Natl Cancer Inst. 1993;85:892–897. 7. Wun LM, Merrill, RM, Feuer EJ. Estimating lifetime and age-conditional probabilities of developing cancer. Lifetime Data Analys. 1998;4:169–18. 8. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst. 1989;81:1879–1886. 9. Reed BW, Wineman J, Bechtel GA. Using a health risk appraisal to determine an Appalachian community’s health care needs. J Cult Divers. 1995;2:131–135. 10. Bechtel GA, Franklin R. Health risk appraisal differences between well elderly and university students. J Community Health Nurs. 1993;10:241–247. 11. Gazmararian JA, Foxman B, Yen LT, Morgenstern H, Edington DW. Comparing the predictive accuracy of health risk appraisal: The Centers for Disease Control versus Carter Center program. Am J Public Health. 1991;81:1296–1301. 12. Fielding JE. Frequency of health risk assessment activities at U.S. work sites. Am J Prev Med. 1989;5:73–81. 13. D’Agostino RB, Wolf PA, Belanger AJ, Kannel WB. Stroke risk profile: Adjustment for antihypertensive medication. The Framingham Study. Stroke. 1994;25:40–43. 14. Feuer EJ, Wun LM. DEVCAN: Probability of DEVeloping CANcer Software. Version 3.4. Bethesda, MD: National Cancer Institute; 1997. 15. Shopland DR, Eyre HG, Pechacek TF. Smoking-attributable cancer mortality in 1991: Is lung cancer now the leading cause of death among smokers in the United States? JNCI. 1991;83:1142–1148.
96
Merrill RISK ESTIMATES OF LUNG CANCER
16. National Center for Health Statistics. Health, United States, 1998. Hyattsville, MD. URL (accessed January 4, 2000): [http://www.cdc. gov/nchswww/datawh/statab/pubd/hus-t62.htm]. 17. Weiss, W. Cigarette smoking and lung cancer trends: A light at the end of the tunnel. CHEST. 1997;111:1414–1416. 18. Merrill RM, Weed DL, Feuer EJ. The lifetime risk of developing prostate cancer in white and black men. Cancer Epidemiol Biomark Prev. 1997;6:763–768. 19. Feuer EJ, Wun LM. How much of the recent rise in breast cancer
AEP Vol. 10, No. 2 February 2000: 88–96
incidence can be explained by increase in mammography utilization? Am J Epidemiol. 1992;136:1423–1436. 20. Wun LP, Feuer EJ, Miller BA. Are increases in mammographic screening still a valid explanation for trends in breast cancer incidence in the United States? Cancer Causes Control. 1995;6:135–144. 21. Strauss GM. Measuring effectiveness of lung cancer screening: From consensus to controversy and back. CHEST. 1997;112:216S–228S. 22. Travis WD, Travis LB, Devesa SS. Lung Cancer. Cancer. 1995;75: 191–202.
PURPOSE: Lifetime and age-conditional probability (risk) estimates of developing lung cancer in the United States are presented by age, race, and gender. Effects on the risk estimates of an aging population and changing tobacco use are identified. METHODS: Risk estimates were derived by applying cross-sectional, population-based incidence rates of malignant lung cancer and mortality rates from other causes to a hypothetical cohort. The cohort was aged through a double-decrement life table to determine the expected proportion of the population that would develop the disease across age intervals. Incidence and mortality data were obtained from the Surveillance, Epidemiology, and End Results (SEER) Program and the National Centers for Health Statistics, respectively. RESULTS: Among all cancers, risk estimates of developing lung cancer within 10 years, conditioned on being free of the disease at age 50, 60, or 70, ranked second to prostate cancer for men and second to breast cancer for women. For men, despite higher incidence rates of lung cancer for blacks than whites across most age groups, the risk of developing this disease over the life-span becomes similar, because white men are more likely to live to older ages where lung cancer is common. For women, lung cancer incidence rates are similar between Whites and Blacks, but an older age distribution among white women explains their greater lifetime risk of being diagnosed with the disease. Changes in the age distribution between the mid 1970s and the mid 1990s had little impact on the short-term risk estimates of developing lung cancer for younger ages but had a large influence on long-term risk estimates, particularly for the older age groups. CONCLUSIONS: Declining lung cancer among younger age groups may be attributed to declining tobacco use among the cohorts, but several more years may be required before the trends begin to fall in older age groups, particularly in women. In the meantime, an aging population is contributing to more people being diagnosed with lung cancer. Consequently, the projected risk of developing lung cancer will remain high for several years to come. Ann Epidemiol 2000;10:88–96. 2000 Elsevier Science Inc. All rights reserved. KEY WORDS:
Lung Neoplasms, Tobacco, Lifetime Risk, Age-Conditional Risk, Burden, Life Table.
INTRODUCTION As the average life expectancy in the United States continues to increase (1), a rising proportion of the population is being affected by chronic diseases of the elderly, such as cancer. This has direct public health implications in terms of increasing needs for health care resources, services, and planning (2, 3). However, considerable efforts have been made to prevent cancer by modifying diet, physical activity, cigarette smoking and other risk factors earlier in life. The combined effects of an aging population and prevention efforts are reflected in annually reported estimates of cancer incidence in the US (4). Likewise, reported estimates of From the Department of Health Science, Brigham Young University, Provo, UT. Address reprint requests to: Ray M. Merrill, Ph.D., M.P.H., Brigham Young University, Department of Health Science, 213 Richards Building, Provo, UT 84602. Received January 4, 1999; revised June 15, 1999; accepted June 18, 1999. 2000 Elsevier Science Inc. All rights reserved. 655 Avenue of the Americas, New York, NY 10010
the projected probability (risk) of developing cancer (4, 5) are also influenced by both an aging population and preventive efforts. Lifetime and age-conditional risk estimates, commonly reported statistics in the popular press and scientific literature, provide an indication of the proportion of the population expected to develop disease. Lifetime risk reflects the proportion of the infant population expected to develop the disease over the average life expectancy in the population. In contrast, age-conditional risk is tailored to a person’s particular age and indicates the proportion of people expected to be diagnosed with the disease within a specified time frame among those free of the disease at a given age. These measures combine the effects of both cross-sectional incidence of the disease and mortality from other causes across age groups. If we compute lifetime and age-conditional risk estimates from incidence and mortality data in 1975–77 and then again from data in 1993–95, the estimates are likely to differ 1047-2797/00/$–see front matter PII S1047-2797(99)00051-4
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
Selected Abbreviations and Acronyms SEER ⫽ Surveillance, Epidemiology, and End Results NCI ⫽ National Cancer Institute
because of changes over time in the age distribution of the population and in the incidence and mortality rates upon which these risk measures are based. In this paper, we report lifetime and age-conditional risk estimates of developing lung cancer based on cross-sectional incidence and mortality data from six three-year time periods, beginning in 1975–77. We also perform calculations for each group of years using the observed rates of lung cancer incidence in those years, while holding the mortality rates constant at their 1975–77 levels. This adjustment shows how changing mortality has influenced the lifetime and age-conditional risk estimates of developing lung cancer. Lung cancer represents an important disease for analysis
89
because of its widespread impact, particularly among the elderly, and because it is dominated by a modifiable risk factor, tobacco. Comparisons in lifetime and age-conditional risk estimates will be made between males and females, Whites and Blacks. Such comparisons are of particular interest because of the large differences observed in life expectancy and tobacco use among these subgroups of the population.
MATERIALS AND METHODS Lifetime and age-conditional risk estimates are derived by applying population-based incidence rates of lung cancer and mortality rates from other causes to a hypothetical cohort. This cohort is aged through a double-decrement life table to determine the number developing the disease at each age interval. A detailed description of the methodology is given elsewhere (6, 7). Because the estimates are based on population data, they provide a projection of the average
TABLE 1. Estimated risk (number per 100,000) of developing selected cancers (invasive only) over the next ten years for those who are free from disease at age 40, 50, 60, or 70 by sex and race (White, Black)a Age for white men Cancer All cancers Prostate Lung & bronchus Colon/rectum Bladderb Lymphomas Oral & pharynx Stomach Pancreas Leukemia Esophagus Liver
Age for black men
40
50
60
70
40
50
60
70
1945 115 215 168 108 211 123 40 32 76 26 24
6342 1681 1068 696 400 331 320 131 146 162 104 61
16,255 6508 2990 1767 1159 549 505 295 318 357 227 159
24,855 9983 4342 2840 1829 813 543 517 580 634 314 264
2765 278 488 276 57 322 278 86 68 59 84 85
10,113 3262 2209 914 241 297 655 272 317 152 359 192
21,433 10,505 4268 1651 450 326 722 505 542 290 534 272
31,236 15,405 5337 2807 918 401 430 873 789 408 495 278
Age for white women Cancer All cancers Breast Lung and bronchus Colon/rectum Corpus uteri Ovarian Cervix uteri Bladderb Lymphomas Oral & pharynx Stomach Pancreas Leukemia Esophagus Liver a b
Age for black women
40
50
60
70
40
50
60
70
3097 1516 187 142 187 187 134 34 102 47 18 26 48 4 7
6389 2669 812 489 559 323 134 126 221 117 48 99 107 24 25
10,723 3600 1924 1233 900 475 135 309 412 205 112 254 204 67 54
14,688 4188 2510 2195 1002 538 114 494 686 241 212 483 340 99 111
3137 1504 303 234 104 104 191 9 88 70 39 46 48 31 11
6308 2379 979 773 307 204 244 61 143 170 90 175 45 77 35
9911 2759 1912 1408 653 350 245 161 194 224 190 384 189 187 80
13,145 3298 1972 2279 642 470 244 403 344 154 399 726 246 197 124
Based on 1993–95 incidence rates from SEER and mortality rates from NCHS. Includes both in situ and invasive cancer.
90
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 1. Age-specific lung cancer incidence rates by gender, race, and age at diagnosis.
disease risk expected in the population over the next x years among those currently at risk of developing the disease at a given age. While these measures provide somewhat specific information when conditioned on a particular age, they do not incorporate individualized information as do other risk models (e.g., family history of disease, screening status, risk factor exposure) (8–13). Lung cancer incidence data were obtained from tumor registries from five states (Connecticut, Iowa, New Mexico, Utah, and Hawaii) and four metropolitan areas (Atlanta, Georgia; Detroit, Michigan; Seattle, Washington; and San Francisco, California). These registries are included in the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (NCI), and represent about 10% of the US population (5). Mortality data were obtained from death certificates among people identified in the SEER areas, as recorded by the National Centers for Health Statistics. Rates were derived from combining the respective incidence and mortality data with population estimates from the US Bureau of the Census. The NCI’s computer program for estimating the risk of developing cancer was used to derive the risk estimates (14). Three
years of data (e.g., 1993–95) were used in the computations in order to provide more stable estimates. Although lifetime risk estimates receive much attention in the popular press, short-term, age-conditional risk estimates are less restricted by long-term data extrapolations. They can be used to show the expected relative burden of the disease among select age groups where cancer is common. However, lifetime risk estimates may be effective at encouraging preventive behavior, particularly among young people where the benefits of prevention are most effective. In the current study we consider both short- and long-term risk estimates of developing lung cancer.
RESULTS To provide a context for our analysis, it is wise to first consider the risk of being diagnosed with lung cancer relative to other selected cancers over 10-year intervals for ages 40, 50, 60, and 70 (Table 1). The risk of developing any cancer is greater among men than women, except for women aged 40, primarily due to breast cancer. White men experi-
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
91
FIGURE 2. Number at risk of developing lung cancer over the life span by gender, race, and time period.
ence a higher cancer risk than black men only with respect to bladder cancer, lymphomas, and leukemia. White women tended to show a greater risk than black women for cancers of the breast, corpus uteri, ovary, lymphomas, and leukemia. The 10-year risk of lung cancer ranks first among men aged 40; ranks second to prostate cancer for men aged 50, 60, and 70; and ranks second to breast cancer for women aged 40, 50, 60, and 70. A primary input to computing risk measures are the disease incidence rates. Figure 1 shows the lung cancer incidence rates between whites and blacks by age, gender, and time period (1975–77, 1993–95). For men, incidence rates are lower when based on data from 1993–95, as opposed to 1975–77, for ages through about 59 for Blacks and 64 for Whites. In older ages, the incidence rates tend to be higher when based on the 1993–95 data. Black men generally have higher incidence rates. For women, incidence rates are lower when based on data from 1993–95, compared to 1975–77 for ages through 49. However, the opposite is true for subsequent ages. The incidence rates are similar for both Blacks and Whites, except in 1993–95 across ages 65 to 80.
From about age 50 and older, the incidence rates are much higher when based on the later time period. Figure 2 shows the number at risk of developing lung cancer across age for cohorts of 10 million people. White men and women are more likely to live to older ages where lung cancer becomes common. White men show the largest improvement in life expectancy between 1975–77 and 1993–95. White women show a greater improvement in life expectancy than black women in the older age groups. Despite the fact that black men tend to have higher incidence of lung cancer than white men, the larger proportion of white men living to older ages where lung cancer becomes common will make the actual burden of this disease more equivalent. Similarly, although white women have similar or slightly higher lung cancer incidence rates than black women, because the former are living to older ages, their risk of developing the disease is higher. To illustrate the dynamics of lung cancer risk, we present the cumulative risk of developing this disease over the expected life-span by race, gender, and time period (Figure 3). The cumulative risk of developing lung cancer has a
92
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 3. Cumulative risk estimates of developing lung cancer by age, race, gender, and time period. *Adjusting for 1975–77 mortality.
sigmoid shape, beginning very low in early adulthood. Risk rises throughout life and plateaus in the oldest age groups. In men, the risk remains consistently lower in white men, although the difference begins to decrease around age 70. The greater improvement in life expectancy between 1975–77 and 1993–95 for white men than black men (see Figure 2) is illustrated by a much larger impact on the risk measures for white men. In women, Blacks show higher risk through about age 60, after which the risk of lung cancer becomes much greater for Whites. The difference is more pronounced in the period 1993–95, both because of higher incidence rates in whites and greater improvement in life expectancy, particularly in the older age groups. Age-conditional risk estimates provide both relevant information to people approaching the ages where lung cancer becomes common and an indication of the relative burden of the disease in specific age groups. Figure 4 presents risk estimates of developing lung cancer within 10, 20, 30, and 35⫹ years among people free of the disease at age 50, by race, gender, and time period. Changes in the risk estimates occur because they are based on cross-sectional incidence and mortality data from seven 3-year time periods. The
broken line represents the risk measures adjusting for 1975–77 mortality (i.e., controlling for changing life expectancy). Among those alive without lung cancer at age 50, black men have much higher risk estimates of developing the disease in 10, 20, 30, and 35⫹ years. The greatest differences on an absolute scale occur for 20- and 30-year risk. In contrast, the conditional 10- and 20-year risk estimates are similar for white and black women, but the conditional 30- and 35⫹ year risks are higher for white women. In Figure 5, cumulative risk of developing lung cancer in the next x years is presented for the population without diagnosed lung cancer at ages 50, 60, 70, and 80 by age, race, gender, and time period (1975–77, 1993–95). The broken lines adjust for 1975–77 mortality. The 10-year risk of developing lung cancer is greatest among those free of the disease at age 70 (except among black women where the age is 60); 20 and 30-year risk of developing lung cancer is greatest among those free of the disease at age 60; and 40 year risk of developing lung cancer is greatest among those free of the disease at age 50. Similar results occur if the mortality rates in 1975–77 were used in deriving the risk estimates in 1993–95. The changing age distribution
AEP Vol. 10, No. 2 February 2000: 88–96
Merrill RISK ESTIMATES OF LUNG CANCER
93
FIGURE 4. Risk of developing lung cancer in 10, 20, 30, or 35⫹ years for those free of the disease at age 50 by gender, race, and time period. *Adjusting for 1975–77 mortality.
had its greatest effect on the risk estimates for people free of lung cancer at age 50, followed by ages 60, 70, and then 80 (except among black women where it appears to have had its greatest effect on those at age 80).
DISCUSSION The prominent role of lung cancer in the US is well known, representing the second most frequently diagnosed cancer and the leading cause of cancer death for both men and women (4). This is particularly disconcerting since most lung cancer cases and 90% of subsequent deaths are attributed to cigarette smoking (15). Hence, any changes in smoking behavior will translate into changes in lung cancer incidence and mortality. Since 1964, cigarette smoking has declined in the US. For ages 18 years and older, the percentage of smokers fell from 42% in 1965 to 25% in 1995, with decreases consistently seen across age, race, and gender (16). A number of events have contributed to this decline (e.g., scientific evidence of a link between cigarette smoking and lung can-
cer, tobacco warning labels, cigarette tax, and the nonsmoker’s rights movement). Accompanying the decline have been lower lung cancer rates among younger age groups (see Figure 1). However, because of the long latency period for this disease (i.e., about 30 years) (17), it may take several more years before the trends in the oldest age groups begin to decline. The competing effects of changing smoking habits and an increasing age distribution has had a profound effect on the estimated lifetime and age-conditional risk estimates of developing lung cancer. For white men, had mortality rates stayed constant from 1975–77 to 1993–95, the estimated risk of developing lung cancer would have remained lower over the entire life-span, not just among age groups below 70, such that by ages 95⫹ the probability dropped from 7.3% to 7.0%. However, because of the older age distribution for white men, the actual lifetime risk of developing lung cancer increased from 7.3% to 8.3%. For black men, had there been no change in life expectancy over the study period, only about half of the observed increased lifetime risk of developing the disease would have occurred: 7.5% to 8.0% rather than 7.5% to 8.5%. For both white and black women,
94
Merrill RISK ESTIMATES OF LUNG CANCER
AEP Vol. 10, No. 2 February 2000: 88–96
FIGURE 5. Cumulative risk of developing lung cancer in the next x years for those free of the disease at select ages by race and gender, 1993–95. *Adjusting for 1975–77 mortality.
the lifetime risk of developing lung cancer increased dramatically, as explained by sharp increases in lung cancer incidence. The increasing life expectancy only explained a relatively small part of this increase. Had there been no change in life expectancy over the study period, the lifetime risk of developing lung cancer would have increased from 2.7% to 5.5% rather than from 2.7% to 5.9% for white women, and from 2.5% to 4.7% rather than from 2.5% to 4.8% for black women. It is interesting to observe that, despite higher lung cancer incidence rates for black men than white men at almost every age, the older age distribution among white men resulted in a fairly similar lifetime risk estimate of developing the disease. In the case of women, although the lung cancer incidence rates are similar between whites and blacks, the older age distribution among white women suggests that they should have a greater probability of being diagnosed with lung cancer over a lifetime, which was the case. In a related study, it was shown that although the incidence rate of developing prostate cancer was higher for black men than for white men across all ages, the lifetime risk of developing
this disease was lower for Blacks than Whites because fewer blacks were living to ages where prostate cancer becomes common (18). Similarly, invasive incidence rates for all cancers combined in the period 1993–95 were higher for Blacks than Whites across all age groups, but a projected 41% of Blacks compared to 45% of Whites are expected to develop an invasive cancer over their lifetime (14). Invasive cancers of the breast, colon, and rectum, as well as Hodgkin’s disease also have higher age-specific incidence rates among Blacks than Whites, but Whites are more likely to develop these diseases over their longer life expectancy (5, 14). Hence, the potential public health impact for certain diseases may be less for blacks than whites despite higher agespecific incidence rates. Notwithstanding declines in cigarette smoking and subsequent declines in lung cancer rates among younger age groups, increasing life expectancy and the fact that lung cancer rates in the older age groups continue to increase indicate that the lifetime risk estimates of developing this disease will continue to increase. Hence, the overall burden of lung cancer is expected to remain high for some time to
AEP Vol. 10, No. 2 February 2000: 88–96
come. Yet, when we condition the risk measures on being free of the disease at age 50, then 10-, 20-, 30-, and ⭓ 35year risk estimates, based on data from more recent years show some decline in the projected burden (Figure 4). These declines begin earlier for 10 years, then 20 years, and so on. Thus, these conditional risk estimates more clearly illustrate the benefits of cigarette smoking cessation than do the lifetime-risk measures. Changes in the age distribution over the study period had little impact on the short-term probability estimates of developing lung cancer for younger ages, but had a noticeable influence on long-term risk estimates, particularly for the older age groups. Additionally, the 10-, 20-, 30-, and 40-year risks of developing lung cancer tended to be greatest among those at ages 70, 60, 60, and 50, respectively. This was the case whether or not the estimates were based on 1975–77 or 1993–95 data. It should be noted that the risk estimates used in this analysis are based on population data. The data is not distinguished by select subgroups of the population that may be at elevated risk of developing lung cancer, such as those with a family history, those exposed to certain environmental factors (e.g., asbestos, radon), or those who are current or past tobacco smokers. Hence, the risk measures presented apply to the average individual in the population, not to those considered to be at high risk of developing lung cancer. However, the estimates do provide tailored information for specific ages and indicate the population disease burden of lung cancer which may be useful for guiding personal health decisions and public health policy actions. Probability measures of developing disease may be influenced by lead time. This is because they are based on disease incidence rates which are susceptible to lead time effects. A lead time effect results when a new screening technique is rapidly diffused into the population, advancing the time of diagnosis prior to when it would have been symptom detected without altering the time of death (19, 20). This will cause an artifactual increase in the incidence rates, but these rates will eventually return to their secular trend. The time the incidence rates remain inflated depends on both how rapidly the screening program is diffused into the population and the length of the lead time between screen detection and symptom detection. The use of short-term probability measures of developing disease may be useful for identifying projected disease burden when lead time affects the estimates, such as for prostate cancer, because they reflect the actual public health burden. Yet screening has been discouraged in the case of lung cancer, because randomized studies have failed to identify its efficacy (21). Without significant changes in screening behavior during the study period, changes observed in the lifetime and ageconditional risk estimates are primarily the result of changing patterns in smoking behavior and increases in life expectancy.
Merrill RISK ESTIMATES OF LUNG CANCER
95
Advances in lung cancer treatment have shown little impact on changing the prognosis for people diagnosed with this disease. Lung cancer is an extremely virulent disease, particularly the small-cell type. The five-year relative survival for all lung carcinomas is 13.9%, whereas for small cell it is 4.6% (22). This explains why incidence and mortality trends for lung cancer are similar (5). Because almost everyone diagnosed with lung cancer dies from the disease, the use of the term burden in this paper to refer to the extent of the population being affected also has reference to the extent of the population facing death. For other diseases where cure is common (e.g., leukemia, prostate cancer, or breast cancer), the type of burden referred to is not as closely related to mortality.
REFERENCES 1. Vital Statistics of the United States, 1994. Mortality. Preprint of v. 11, part A, sec. 6 life tables. Hyattsville, MD: National Center for Health Statistics; 1998. 2. Brown ML, Fintor, L. The economic burden of cancer. In: Greenwald P, Kramer BS, Weed DL, eds. Cancer Prevention and Control. Bethesda, MD: National Cancer Institute; 1995;69–81. 3. Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause 1990–2020: Global burden of disease study. The Lancet. 1997;349:1498–1504. 4. Landis SH, Murray T, Bolden S, Wingo PA. Cancer Statistics, 1998. CA Cancer J Clin. 1998;48:6–29. 5. Ries LA, Kosary CL, Hankey BF, Miller BA, Harras A, Edwards BK, eds. SEER Cancer Statistics Review, 1973–1994. NIH Pub. No. 972789. Bethesda, MD: National Cancer Institute; 1997. 6. Feuer EJ, Wun LM, Boring CC, Flanders WD, Timmel MJ, Tong T. The lifetime risk of developing breast cancer. J Natl Cancer Inst. 1993;85:892–897. 7. Wun LM, Merrill, RM, Feuer EJ. Estimating lifetime and age-conditional probabilities of developing cancer. Lifetime Data Analys. 1998;4:169–18. 8. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst. 1989;81:1879–1886. 9. Reed BW, Wineman J, Bechtel GA. Using a health risk appraisal to determine an Appalachian community’s health care needs. J Cult Divers. 1995;2:131–135. 10. Bechtel GA, Franklin R. Health risk appraisal differences between well elderly and university students. J Community Health Nurs. 1993;10:241–247. 11. Gazmararian JA, Foxman B, Yen LT, Morgenstern H, Edington DW. Comparing the predictive accuracy of health risk appraisal: The Centers for Disease Control versus Carter Center program. Am J Public Health. 1991;81:1296–1301. 12. Fielding JE. Frequency of health risk assessment activities at U.S. work sites. Am J Prev Med. 1989;5:73–81. 13. D’Agostino RB, Wolf PA, Belanger AJ, Kannel WB. Stroke risk profile: Adjustment for antihypertensive medication. The Framingham Study. Stroke. 1994;25:40–43. 14. Feuer EJ, Wun LM. DEVCAN: Probability of DEVeloping CANcer Software. Version 3.4. Bethesda, MD: National Cancer Institute; 1997. 15. Shopland DR, Eyre HG, Pechacek TF. Smoking-attributable cancer mortality in 1991: Is lung cancer now the leading cause of death among smokers in the United States? JNCI. 1991;83:1142–1148.
96
Merrill RISK ESTIMATES OF LUNG CANCER
16. National Center for Health Statistics. Health, United States, 1998. Hyattsville, MD. URL (accessed January 4, 2000): [http://www.cdc. gov/nchswww/datawh/statab/pubd/hus-t62.htm]. 17. Weiss, W. Cigarette smoking and lung cancer trends: A light at the end of the tunnel. CHEST. 1997;111:1414–1416. 18. Merrill RM, Weed DL, Feuer EJ. The lifetime risk of developing prostate cancer in white and black men. Cancer Epidemiol Biomark Prev. 1997;6:763–768. 19. Feuer EJ, Wun LM. How much of the recent rise in breast cancer
AEP Vol. 10, No. 2 February 2000: 88–96
incidence can be explained by increase in mammography utilization? Am J Epidemiol. 1992;136:1423–1436. 20. Wun LP, Feuer EJ, Miller BA. Are increases in mammographic screening still a valid explanation for trends in breast cancer incidence in the United States? Cancer Causes Control. 1995;6:135–144. 21. Strauss GM. Measuring effectiveness of lung cancer screening: From consensus to controversy and back. CHEST. 1997;112:216S–228S. 22. Travis WD, Travis LB, Devesa SS. Lung Cancer. Cancer. 1995;75: 191–202.