Cancer incidence and mortality, 1973-1995

A report card for the U. S.

Authors

  • Phyllis A. Wingo Ph.D., M.S.,

    Corresponding author
    1. Epidemiology and Surveillance Research Department, American Cancer Society, Atlanta, Georgia
    • American Cancer Society, 1599 Clifton Road, N.E., Atlanta, GA 30329-4251
    Search for more papers by this author
  • Lynn A. G. Ries M.S.,

    1. Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, Maryland
    Search for more papers by this author
  • Harry M. Rosenberg Ph.D.,

    1. Division of Vital Statistics, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, Maryland
    Search for more papers by this author
  • Daniel S. Miller M.D., M.P.H.,

    1. Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, Georgia
    Search for more papers by this author
  • Brenda K. Edwards Ph.D.

    1. Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, Maryland
    Search for more papers by this author

Abstract

BACKGROUND

The American Cancer Society, the National Cancer Institute (NCI), and the Centers for Disease Control and Prevention including the National Center for Health Statistics (NCHS) agreed to produce together an annual "Report Card" to the nation on progress related to cancer prevention and control in the U.S.

METHODS

This report provides average annual percent changes in incidence and mortality during 1973-1990 and 1990-1995, plus age-adjusted cancer incidence and death rates for whites, blacks, Asians and Pacific Islanders, and Hispanics. Information on newly diagnosed cancer cases is based on data collected by NCI, and information on cancer deaths is based on underlying causes of death as reported to NCHS.

RESULTS

For all sites combined, cancer incidence rates decreased on average 0.7% per year during 1990-1995 (P > 0.05), in contrast to an increasing trend in earlier years. Among the ten leading cancer incidence sites, a similar reversal in trends was apparent for the cancers of the lung, prostate, colon/rectum, urinary bladder, and leukemia; female breast cancer incidence rates increased significantly during 1973-1990 but were level during 1990-1995. Cancer death rates for all sites combined decreased on average 0.5% per year during 1990-1995 (P < 0.05) after significantly increasing 0.4% per year during 1973-1990. Death rates for the four major cancers (lung, female breast, prostate, and colon/rectum) decreased significantly during 1990-1995.

CONCLUSIONS

These apparent successes are encouraging and signal the need to maximize cancer control efforts in the future so that even greater in-roads in reducing the cancer burden in the population are achieved. Cancer 1998;82:1197-207. © 1998 American Cancer Society.

At a press conference on the 25th anniversary of the National Cancer Act, the American Cancer Society (ACS), the National Cancer Institute (NCI), and the Centers for Disease Control and Prevention (CDC) including the National Center for Health Statistics (NCHS) reported the first sustained decline in cancer mortality since national recordkeeping was instituted in the 1930s.1-5 These organizations agreed to continue to collaborate, monitor cancer statistics, and produce an annual "Report Card" to the nation on progress related to cancer prevention and control in the U. S. The purpose of the current report is to provide an update on the continuing decline in cancer mortality in the U. S. and to present information regarding cancer incidence not reported previously. In addition, the report includes cancer incidence and mortality rates and short term trends in four populations: whites, blacks, Asians and Pacific Islanders, and Hispanics.

MATERIALS AND METHODS

Cancer Cases

Information on newly diagnosed cancer cases occurring in the U. S. is based on data collected by the NCI's Surveillance, Epidemiology, and End Results (SEER) program.5 Briefly, the SEER program collects cancer incidence data from nine population-based registries, including five states (Connecticut, Hawaii, Iowa, New Mexico, and Utah) and four standard metropolitan statistical areas (Atlanta, Detroit, San Francisco-Oakland, and Seattle-Puget Sound), representing an estimated 9.5% of the U.S. population. Estimates of cancer rates and trends for the total U. S. frequently are based on SEER data. For this analysis, we use cancer cases diagnosed during 1973-19955; the second edition of the International Classification for Diseases for Oncology (ICD-02) groupings for the specific cancer sites included in this report have been published previously.5, 6

Cancer Deaths

Information on cancer deaths in the U. S. is based on causes of death reported by certifying physicians on death certificates filed in the states. The mortality information is processed and consolidated into a national data base by the NCHS (reference 7 and unpublished data). The underlying cause of death is selected for tabulation following the procedures specified by the World Health Organization in the relevant Manual of the International Classification of Diseases, Injuries, and Causes of Death. For the period 1973-1978, the eighth revision (ICDA-8) was used, and for 1979-1995 the ninth revision (ICD-9) was used.9 For this analysis, ICDA-8 codes on individual records were converted to ICD-9 codes by applying a conversion algorithm used by the NCI. To ensure comparability between the ICDA-8 and ICD-9 codes, the ICD-9 codes in this report are categorized according to SEER site groupings.5

Cancer Incidence and Death Rates

We use resident population estimates for each year from the U.S. Bureau of the Census to compute age-adjusted cancer incidence and death rates; population data for whites were adjusted slightly for an overcount of whites in Hawaii (reference 10 and unpublished data). Rates are expressed as per 100,000 population, and age-adjustment involves the direct method and the 1970 U.S. standard million population. For cancer sites that pertain only to males or females, rates are based on gender specific data. For cancer incidence rates, the denominators are county level population data for the geographic areas that participate in the SEER program, and for cancer death rates, the denominators are population data for the total U. S., except for Hispanic data. All rates in this report are based on at least 25 cases or deaths. The term "all sites" refers to all cancer sites combined, not just to the aggregate of sites included in each figure; specific site abbreviations include other areas of the nervous system (ONS) and intrahepatic bile duct (IBD).

Average Annual Percent Change

The average annual percent change (APC) is estimated by fitting a regression line to the natural logarithm of the rates using calendar year as a regressor variable, i.e., y = m x + b in which y = ln (rate) and x = calendar year. Then the estimated APC = 100*(e m - 1). Testing the hypothesis that the APC is equal to zero is equivalent to testing the hypothesis that the slope of the line in the above equation is equal to zero (i.e., that the rate is not increasing or decreasing). The hypothesis test statistic uses the t distribution of m/SE m in which SE is the standard error of m and the number of degrees of freedom is equal to the number of calendar years minus two.11 The calculation assumes that rates increase or decrease at a constant rate over time although the validity of this assumption has not been assessed. Differences between incidence and mortality trends for the time period 1973-1990 versus those for 1990-1995 have been tested for statistical significance using a t statistic with 20 degrees of freedom defined as the difference in the regression coefficients divided by the standard error of the difference.11 Statistical significance for all hypothesis tests was assessed using P = 0.05, and no adjustment was made for multiple hypothesis testing.

Race and Ethnicity

Because information regarding Hispanic origin is collected separately from race, persons categorized as Hispanic are not mutually exclusive from whites, blacks, and Asians and Pacific Islanders. Cancer incidence rates for Hispanics are based on data from all the areas that have participated in the SEER program since the mid-1970s. Cancer death rates for Hispanics include cancer deaths that occurred in all states except Connecticut, Louisiana, New Hampshire, and Oklahoma, which were omitted due to the absence of comparable data on Hispanic origin. Data for Native Americans are not shown separately in this analysis but currently are being evaluated for future reports; they are included in the statistics for all races combined.

RESULTS

Incidence

For all sites combined, SEER incidence rates increased (1.2% per year) during 1973-1990 and decreased (-0.7% per year) during 1990-1995 (Fig. 1 (10K)), although the trend for the latter time period was not statistically significant. The change in direction (the difference in trends) between the two time periods was statistically significant. Among the ten leading cancer incidence sites (Fig. 1 (10K)), a similar reversal in trends was apparent for cancer of the lung, prostate, colon/rectum, urinary bladder, and leukemia. However, the differences in the average annual percent change between the two time periods were significant only for cancers of the lung, colon/rectum, and urinary bladder. Although incidence rates for female breast cancer were not decreasing in 1990-1995, they were no longer significantly increasing as in 1973-1990 (1.8% per year). Incidence rates for non-Hodgkin's lymphoma and melanomas increased during both time periods; it is interesting to note that the annual percent increase for non-Hodgkin's lymphoma for 1990-1995 was significantly lower than for the earlier time period, suggesting a slowing in the annual increase in incidence for this disease. Trends in the incidence of cancer of the corpus uteri and uterus, not otherwise specified (NOS) were in the opposite directions of the overall trends; they decreased significantly during 1973-1990 primarily due to changes in prescribing practices for unopposed noncontraceptive estrogens and increased during 1990-1995. The change in direction in trends for the two time periods was statistically significant. During 1990-1995, the largest annual decreases in incidence occurred in persons who were ages 35-44 years and persons who were age ≥ 75 years at diagnosis (Fig. 2 (10K)). Although most age groups demonstrated significant annual increases during 1973-1990, all ages showed declines or nonsignificant increases during the later time period.

Figure 1.

** Incidence rates are age-adjusted to the 1970 U.S. standard million population. # Annual Percent Change (APC) is based on gender specific rates. * The APC is significantly different from zero (P < 0.05). NOS: not otherwise specified.

Figure 2.

** Incidence rates are age-adjusted by 5-year age groups to the 1970 U.S. standard million population. * The Annual Percent Change is significantly different from zero (P<0.05).

During 1990-1995, the four leading cancer sites for all racial and ethnic populations in the U. S. were cancers of the lung and bronchus, prostate, female breast, and colon/rectum (Figs. 3 (8K), 4 (8K), 5 (7K), and 6 (8K)), which together account for approximately 54% of all newly diagnosed cancers.12 Examination of incidence rates for each of these sites by race and ethnicity revealed that, except for female breast cancer, blacks had higher incidence rates than whites, Asians and Pacific Islanders, or Hispanics. Although all four racial and ethnic groups had cancer of the corpus uteri and uterus, NOS and non-Hodgkin's lymphoma in common among the top ten sites, some sites were unique to a specific population. Melanoma and leukemia were among the top ten incidence sites only in whites, cancers of the pancreas and oral cavity/pharynx were among the top ten sites only in blacks, and liver cancer was among the top ten sites only in Asians and Pacific Islanders. Cancer of the uterine cervix and stomach were among the top ten sites for all racial and ethnic groups except whites.

Figure 3.

** Rates are per 100,000 and age-adjusted to the 1970 U.S. standard million population, # Rates are based on gender specific data. NOS: not otherwise specified; ONS: other nervous system; IBD: intrahepatic bile duct.

Figure 4.

** Rates are per 100,000 and age-adjusted to the 1970 U.S. standard million population. # Rates are based on gender specific data. NOS: not otherwise specified; IBD: intrahepatic bile duct; ONS: other nervous system.

Figure 5.

** Rates are per 100,000 and age-adjusted to the 1970 U.S. standard million population. # Rates are based on gender specific data. NOS: not otherwise specified; IBD: intrahepatic bile duct; ONS: other nervous system.

Figure 6.

** Rates are per 100,000 and age-adjusted to the 1970 U.S. standard million population. † Hispanic is not mutually exclusive from whites, blacks, and Asians and Pacific Islanders; death rates exclude deaths that occurred in Connecticut, Louisiana, New Hampshire, and Oklahoma. # Rates are based on gender specific data. NOS: not otherwise specified; IBD: intrahepatic bile duct; ONS: other nervous system.

Incidence trends during 1990-1995 also varied by gender, race, and ethnicity. Between 1990 and 1995, SEER incidence rates for all cancer sites combined decreased for all racial and ethnic groups, except for black males and Asian and Pacific Islander females; the declines were not statistically significant (Figs. 7 (6K) and 8 (6K)). Black males did not experience this overall decline, primarily due to the average annual increase of 3.9% per year in prostate cancer incidence. However, after increases during 1989-1993, prostate cancer incidence rates in black males declined sharply in 1994 and 1995.5 Cancer incidence rates for Asian and Pacific Islander females did not increase or decrease consistently during this time period.

Figure 7.

** Incidence and death rates are age-adjusted to the 1970 U.S. standard million population; Annual Percent Change (APC) is based on gender specific rates. # Asians and Pacific Islanders. @ Hispanic is not mutually exclusive from whites, blacks, and Asians and Pacific Islanders; death rates exclude deaths that occurred in Connecticut, Louisiana, New Hampshire, and Oklahoma. * The APC is significantly different from zero (P<0.05).

Figure 8.

** Incidence and death rates are age-adjusted to the 1970 U.S. standard million population; Annual Percent Change (APC) is based on gender specific rates. # Asians and Pacific Islanders. @ Hispanic is not mutually exclusive from whites, blacks, and Asians and Pacific Islanders. Death rates exclude deaths that occurred in Connecticut, Louisiana, New Hampshire, and Oklahoma. The APC is significantly different from zero (P<0.05).

The pattern of decreasing incidence for the four major cancers varied by gender, race, and ethnicity (Figs. 7 (6K) and 8 (6K)). The incidence of cancers of the colon/rectum decreased in males and females in all racial and ethnic groups; the declines were statistically significant in white males, white females, and Asian and Pacific Islander males. Lung cancer incidence in males also declined in all racial and ethnic groups, whereas lung cancer incidence in females declined only among black (-1.5% per year) and Hispanic females (-3.4% per year). Other exceptions to the overall declining incidence patterns for the major cancers included an increasing prostate cancer incidence in black, Asian and Pacific Islander, and Hispanic males, increasing lung cancer in white and Asian and Pacific Islander females, and increasing breast cancer in black females.

Mortality

After significantly increasing on average 0.4% per year during 1973-1990, cancer death rates for all sites combined decreased significantly an average of 0.5% per year during 1990-1995 (Fig. 9 (10K)); the change in direction and the difference in trends between the two time periods was statistically significant. Among the ten leading cancer mortality sites, a pattern of significantly increasing cancer death rates during 1973-1990, followed by significantly decreasing cancer death rates during 1990-1995, was observed for lung cancer, female breast cancer, and prostate cancer. The mortality pattern for melanoma was similar, except that the trend for 1990-1995 did not achieve statistical significance. Death rates for cancers of the colon/rectum were decreasing significantly in both time periods. Among the other top ten cancer mortality sites, non-Hodgkin's lymphoma death rates were increasing significantly in both time periods, whereas death rates for cancer of the corpus uteri and uterus, NOS, urinary bladder, ovary, and leukemia decreased over both time periods. During 1990-1995, the average annual decreases in cancer death rates primarily were confined to persons who were age < 65 years at the time of death (Fig. 10 (10K)).

Figure 9.

** Death rates are age-adjusted to the 1970 U.S. standard million population. # Annual Percent Change (APC) is based on gender specific rates. * The APC is significantly different from zero (P < 0.05). NOS: not otherwise specified; lym: lymphoma.

Figure 10.

** Death rates are age-adjusted by 5-year age groups to the 1970 U.S. standard million population. * The Annual Percent Change is significantly different from zero (P < 0.05).

The top four causes of cancer death in the U. S. during 1990-1995 for all racial and ethnic groups were the same sites as for incidence (Figs. 3 (8K), 4 (8K), 5 (7K), and 6 (8K)); > 50% of all cancer deaths involved the lung, female breast, prostate, or colon/rectum.12 Examination of cancer death rates for each of these sites by gender, race, and ethnicity revealed that, except for female lung cancer, blacks had higher cancer death rates than whites, Asians and Pacific Islanders, or Hispanics. The female lung cancer death rate in whites (33.6 per 100,000) was only slightly higher than in blacks (32.7 per 100,000). Other mortality sites among the top ten that were common to all four racial and ethnic groups were cancers of the pancreas, stomach, and ovary.

Otherwise, the causes of cancer death among the ten leading sites varied by racial and ethnic group. Deaths due to cancer of the brain and ONS were among the top ten mortality sites only in whites; deaths due to cancer of the esophagus, cervix, and multiple myeloma were among the top ten sites only in blacks; and liver cancer deaths were among the top ten sites only in Asians and Pacific Islanders and Hispanics.

Between 1990 and 1995, the decline per year in U.S. cancer death rates for all sites combined was greater in males (-0.9%) than in females (-0.1%). Although the overall rates in males were higher than in females, the average annual decrease for all sites combined in males was statistically significant and the decrease in females was not, due to the significant annual increases in female lung cancer death rates of 1.6% per year. The declines per year in U.S. cancer death rates for all sites combined were significant in blacks (-0.8%), whites (-0.4%), and Hispanics (-0.6%). Cancer death rates for all sites combined in Asians and Pacific Islanders were unchanged, due to increasing rates in Asian and Pacific Islander females and decreasing rates in males (Figs. 7 (6K) and 8 (6K)).

There were several exceptions to these mortality patterns among the major cancers (Figs. 7 (6K) and 8 (6K)). Among males, the only exception was prostate cancer mortality in Hispanics, which increased at 1.6% per year during 1990-1995. Among females, the exceptions included increasing cancer death rates in Asians and Pacific Islanders for all of the major cancers, although none of these increasing trends was statistically significant. In addition, white and black females had increasing lung cancer death rates (1.7% and 1.0% per year, respectively).

DISCUSSION

Cancer incidence rates in the U. S. have been increasing since the 1930s,5, 13, 14 and appear to have peaked in 1992.5 Incidence rates for all sites combined decreased significantly an average of 2.7% per year during 1992-1995. Because the decline in incidence occurred only for the 3 most recent years of data, the trend for the entire time period 1990-1995, although negative, was not statistically significant.

The downturn in cancer incidence after long term increases varied by site, gender, race, and ethnicity.5 Among the major cancers, the earliest decreases in incidence occurred for male lung cancer in whites and blacks in the mid-1980s, and for cancer of the colon/rectum in whites during the same time period. Decreases in the incidence of some cancers among blacks occurred later than among whites. Among blacks, decreases in the incidence of colon/rectum cancer began in the early 1990s. Decreases in prostate cancer in blacks appeared to begin in the mid-1990s in contrast to the early 1990s for whites.

To have a clear understanding of current successes and future challenges in cancer control, these findings should be interpreted with caution in a context that includes information on risk factors, screening behaviors, treatment protocols, access to health services, and the biology of specific cancer sites. Decreasing incidence and mortality from lung cancer most likely result from decreased smoking rates over the previous 30 years.15 However, smoking patterns among women lag behind those of men. The decrease in lung cancer incidence among black females provides hope that a downturn in mortality may soon follow. Although we are observing some successes related to the occurrence of lung cancer, particularly among males, these results derive from events that occurred in the distant past, and the cancer control and research programs that produced these results need to continue. Of concern, declines in adult tobacco use have slowed,15 and tobacco use in youth is increasing again.16

Changes in prostate cancer incidence rates may be due to the interaction of multiple factors that warrant further investigation. A sizable proportion of the recent decrease in prostate cancer incidence rates among white men in part may be related to the following sequence of events: 1) the dissemination of prostate specific antigen (PSA) testing into an unscreened population in the late 1980s and early 1990s; 2) the resultant diagnoses of previously undetected prevalent cancer cases in the community during the same time period; and 3) the consequent deficit of prevalent cases that occurs after widespread screening. Some of these cancers may be clinically significant whereas others may be latent cancers that would not have progressed to be clinically detectable. Once the prevalent cancers in the population have been identified through screening, the incidence rates may fall toward an eventual equilibrium reflecting truly incident cases in the population.

Another proportion of the decrease observed in prostate cancer incidence rates may be due to other factors. An increasing number of diagnoses and treatment protocols are occurring in outpatient or nonhospital settings from which cancer reporting and registration for national statistics may be less complete or delayed. In addition, there have been widely publicized, heated debates regarding the efficacy of screening for prostate cancer in the absence of randomized, controlled trials to demonstrate the impact of screening on mortality. Decreased incidence rates in part may be the result of decreased utilization of PSA screening tests, precipitated by organizations that discouraged their use during the early 1990s.17 Although the average annual percent change from 1990-1995 indicates a continued increase in the incidence of prostate cancer among black men, it does not reveal the declining incidence rates observed for these men in 1994 and 1995.5

The reported decreases in prostate cancer mortality are even more difficult to explain. Regions of the country that have experienced the greatest decreases in mortality from prostate cancer are areas that have lower utilization of PSA screening.17 The influence of treatment such as radiation, brachytherapy, cryotherapy, nerve-sparing radical prostatectomy, and use of antiandrogens on mortality and quality of life similarly are unclear. Further studies to investigate the reasons for the decrease in prostate cancer mortality in recent years are warranted. It is noteworthy that prostate cancer mortality appeared to be increasing in Hispanic males. However, this increase did not achieve statistical significance, and prostate cancer incidence and mortality rates in Hispanic men remain substantially lower than in whites and blacks.

Decreasing mortality from breast cancer in white women most likely is a reflection of the increasingly widespread diffusion of breast cancer screening into routine medical care. The increased utilization of adjuvant therapies also contributes to the decline in breast cancer mortality. Although recent studies indicate that the gap between black and white women in self-reported utilization of mammography screening has narrowed considerably,18 the mortality benefits have not yet been observed in black women. Special efforts are being made by the CDC in partnership with health departments in all 50 states to make comprehensive breast cancer screening available to minority and low income women through the National Breast and Cervical Detection Program.19 Breast cancer incidence in black women continues to increase, and mortality trends in black women are not decreasing.5 Breast cancer is diagnosed at later stages in black women (52% at regional and distant stages) than in white women (41% at regional and distant stages), and the 5-year relative survival rate is lower (57% and 71% for blacks and whites, respectively). 5 Finally, black women with breast cancer may have less favorable tumor characteristics (i.e., biologically more aggressive) than white women.20

The reasons for decreasing incidence and mortality due to cancers of the colon/rectum among all racial and ethnic populations are not well understood. The very low utilization of screening tests (e.g., fecal occult blood testing or sigmoidoscopy,18) suggests that factors other than screening may be involved. Increased polyp removal, advances in treatment protocols (e.g., the increased use of newer surgical techniques and adjuvant therapies), and other factors (e.g., daily use of aspirin21 or estrogen replacement therapy,22 or changes in population dietary patterns23) may be contributing factors. Although there may be minor improvements in diet in the U.S. population, the increasing prevalence of obesity argues against sustained dietary changes.24

During 1973-1990, the incidence of non-Hodgkin's lymphoma increased significantly among younger and older males.5 Given that non-Hodgkin's lymphoma in the presence of human immunodeficiency virus (HIV) infection is an acquired immunodeficiency syndrome (AIDS)-defining illness, the increasing incidence among men age < 55 years in the U. S. most probably reflects the increasing impact of HIV infection. However, the average annual percent increase in non-Hodgkin's lymphoma appears to be slowing, possibly due to the beneficial effects of antiretroviral therapies on the rate of HIV disease progression.25

In this report, we used age-adjusted cancer incidence and cancer death rates. Although widely used as summary indicators, age-adjusted rates have recognized limitations.26 First, they may mask important variations in the deaths or incidence rates among specific age groups. For cancer in particular, declining mortality at younger ages is offset by increases at older ages, resulting in a small net mortality decline for all ages and all sites combined, a pattern reflected in specific sites as well. For example, male lung cancer mortality rates began declining in younger age groups long before the overall age-adjusted rate did. Male lung cancer mortality rates peaked in 1970 for those age < 45 years, in 1978 for those ages 45-54 years, and since 1978 for those ages 55-64 years. This is in contrast to an overall age-adjusted rate that did not peak until 1990. Second, trends and patterns in incidence and mortality may be sensitive in some situations to the choice of the population standard used for age-adjustment, in particular the use of an earlier population, such as 1940 compared with 1970 or the year 2000. Recommendations currently are being formulated for national and state agencies to adopt the year 2000 as a uniform new standard, which will better reflect the current (and older) age structure of the population. A more current standard would result in somewhat attenuated reductions in cancer incidence and mortality.

The incidence data presented in this report were derived from selected geographic areas of the U.S. representing only 9.5% of the U.S. population. The National Program of Cancer Registries funded by the CDC was established in 1994 with an emphasis on improving the completeness, timeliness, and quality of cancer incidence data.27 Publications from the North American Association of Central Cancer Registries currently provide a compilation of data from a number of state registries.28 In the future, data regarding the occurrence of cancer will become available for most states 28 and will improve our ability to monitor cancer incidence trends and guide cancer control efforts throughout the U. S.

The ACS, NCI, and CDC including the NCHS expect to continue monitoring the occurrence of cancer in the U. S. and to collaborate in the development of an annual "Report Card" to the nation. Future assessments may include information on prevalence and trends in areas of cancer control, health behaviors such as tobacco use, diet, and utilization of cancer screening procedures, as well as key successes in treatment, and significant cancer research findings.

Ancillary