Key Words. SEER Cancer Survival Incidence Mortality Prevalence

The Oncologist Cancer Survival and Incidence from the Surveillance, Epidemiology, and End Results (SEER) Program LYNN A. GLOECKLER RIES, MARSHA E. REICHMAN, DENISE RIEDEL LEWIS, BENJAMIN F. HANKEY, BRENDA K. EDWARDS Surveillance Research Program, DCCPS, National Cancer Institute, Bethesda, Maryland, USA Key Words. SEER Cancer Survival Incidence Mortality Prevalence LEARNING OBJECTIVES After completing this course, the reader will be able to: 1. Describe the role and significance of the NCI Surveillance and End Results (SEER) Program. 2. Discuss the evolving changes in cancer mortality statistics in the U.S. 3. Explain some of the important definitions of population science as they relate to cancer. CME Access and take the CME test online and receive one hour of AMA PRA category 1 credit at CME.TheOncologist.com ABSTRACT An overview of data on cancer at all sites combined and on selected, frequently occurring cancers is presented. Descriptive cancer statistics include average annual Surveillance, Epidemiology, and End Results (SEER) Program incidence, U.S. mortality and median age at diagnosis, and death for the period 1996-2. Changes during the time period 1992-2 are summarized by the annual percent change in SEER incidence and U.S. mortality data for this period. Fiveyear relative survival for selected cancers is examined by stage at diagnosis, based on data from 199-1999. In addition, 5-year conditional survival for patients already surviving for 1-3 years after diagnosis is discussed as well as relative survival for other time periods. These measures may be more meaningful for clinical management and prognosis than 5-year relative survival from time of diagnosis. The likelihood of developing cancer during one s lifetime is 1 in 2 for males and 1 in 3 for females, based on 1998-2 data. It is estimated that approximately 9.6 million people in the U.S. who have had a diagnosis of cancer are alive. Five-year relative survival varies greatly by cancer site and stage at diagnosis, and tends to increase with time since diagnosis. The median age at cancer diagnosis is 68 for men and 65 for women. The 5-year relative survival rate for persons diagnosed with cancer is 62.7%, with variation by cancer site and stage at diagnosis. For patients diagnosed with cancers of the prostate, female breast, corpus uteri, and urinary bladder, the relative survival rate at 8 years is over 75%. The Oncologist 23;8:541-552 INTRODUCTION The purpose of this article is to provide a limited statistical overview of cancer at all sites combined and at selected, frequently occurring sites. It also demonstrates the utility of the Surveillance, Epidemiology and End Results (SEER) Program database for the study of the occurrence Correspondence: Marsha E. Reichman, Ph.D., Surveillance Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, 6116 Executive Boulevard, Suite 54, Bethesda, Maryland 2892-8316, USA. Telephone: 31-594-732; Fax: 31-48-477; e-mail: reichmam@mail.nih.gov; web site: http://seer.cancer.gov Received August 7, 23; accepted for publication September 12, 23. AlphaMed Press 183-7159/23/$12./ The Oncologist 23;8:541-552 www.theoncologist.com

Ries, Reichman, Lewis et al. 542 and behavior of newly diagnosed cancer. Aspects of the SEER incidence and survival data and U.S. mortality data are presented. Their interrelationships and the impact of treatment modalities and preventive interventions are discussed in a general sense. In addition to 5-year relative survival from time of diagnosis, 5-year relative survival for those who have already survived for 1, 2, or 3 years after diagnosis and relative survival data for time periods other than 5 years are also presented. In some circumstances, these may be more clinically relevant. MATERIALS AND METHODS The data on newly diagnosed cancer cases used in this analysis were collected from medical records at hospitals and other facilities by population-based cancer registries that participate in the SEER Program based at the National Cancer Institute (NCI) [1]. The SEER Program was established as a direct result of the National Cancer Act of 1971 that mandated the collection, analysis, and dissemination of cancer data for further use in the prevention, diagnosis, and treatment of cancer. Over the last 3 years, the SEER Program has grown and currently includes data from the states of Connecticut, Iowa, New Mexico, Utah, Hawaii, Louisiana, Kentucky, New Jersey, and California and from Alaska Natives in Alaska, in addition to the metropolitan areas of: Detroit, Michigan; San Francisco-Oakland, San Jose-Monterey, and Los Angeles County, California (the latter three metropolitan areas participated in SEER prior to other parts of California); Atlanta, Georgia, and Seattle-Puget Sound, Washington. This covers 26% of the U.S. population and is richly representative of the nation s vast array of racial and ethnic groups including 23% of the nation s African Americans, 4% of the nation s Hispanics, 42% of Native Americans (American Indians and Alaska Natives), and 59% of the Asian/Pacific Islander population, as well as 23% of the nation s Caucasian population. The statistics presented here for incidence are for the group of registries referred to as SEER-12 Areas including Connecticut, Iowa, New Mexico, Utah, Hawaii, Detroit, San Francisco- Oakland, San Jose-Monterey, Los Angeles, Atlanta, Seattle- Puget Sound, and Alaska Natives. Survival data are presented for SEER-9 Areas including registries in Connecticut, Iowa, Hawaii, New Mexico, Utah, Atlanta, Detroit, San Francisco- Oakland, and Seattle-Puget Sound [2]. Survival data are restricted to SEER-9 Areas since these registries have been operating for enough time to provide long-term survival data. Data are actively collected by SEER cancer registries and reported to the NCI on all cancers of residents of the geographic area of the registry. Cases are ascertained from records of hospitals, private laboratories, radiotherapy units, nursing homes, and other health service units that provide diagnostic or treatment services, and from death certificates of residents when cancer is listed as a cause of death. Data collected on each cancer include patient demographics, primary cancer site/type, morphology, diagnosis confirmation, extent of disease, first course of treatment, and active patient follow-up for vital status including cause of death. Cancers are coded according to the International Classification of Diseases for Oncology Second Edition (ICD-O-2) [3]. For 1992-2 diagnoses, site and histology were coded by ICD- O-2 criteria, and all cases before 1992 were machine converted to ICD-O-2 coding. Stage at cancer diagnosis was recorded by extent of disease that can be computer converted into tumor/node/metastasis (TNM) Third Edition stages [4]. The cancer incidence rate is the number of new cancers of a specific site/type, or of all sites combined, occurring in a specified population over a specific time period. It is usually expressed as the number of cancers per 1, population at risk per year. For cancer sites occurring in only one sex, the population at risk is the sex-specific population (e.g., males for prostate cancer). The death rate is the number of deaths with cancer as the underlying cause of death occurring in a specified population over a specific time period, usually expressed per 1, population at risk per year. Numerators for incidence rates are derived from SEER Program data. The SEER Program obtains information on all deaths occurring in the U.S. from the National Center for Health Statistics (NCHS) on an annual basis. Data are available for reporting incidence and mortality approximately 2 years after the end of the calendar year. Thus, in 22, incidence and death data were available through 2. This 2-year period for reporting incidence data is needed for adequate time to complete the first course of treatment and for ascertainment of information on the majority of cases, especially those diagnosed in outpatient settings. Updating of cancer registry data from previous years is an ongoing process resulting in the most accurate statistics possible. While information on 98% of most cancers is collected within 2-3 years, time periods for a few sites may be longer [5]. This may result in adjustments to incidence rates over time due to the updated numerators. In this report, we present the observed data. Modeling to account for these reporting delays may result in somewhat higher predicted rates. Denominator data for both incidence and death rates are obtained from the Bureau of the Census. Census data are available from the 199 and 2 Decennial Census data collections. Population data for years between 199 and 2 were initially estimated based on data from 199, with estimates being updated after data from the 2 Census became available. Data presented here are based on updated intercensal estimates for the period 1991-1999 [6]. When intercensal corrections are made, cancer incidence and death rates may be altered due to changes in the denominators.

543 SEER Program Cancer Survival and Incidence Estimates of racial/ethnic populations include the use of methods to bridge differences in collection of these data between the 199 and 2 Censuses [7, 8]. The 2 Census was the first census where respondents could choose multiple racial/ethnic groups rather than a single group. Methods were developed by the NCHS and the Bureau of the Census to provide estimates of 2 data on race/ethnicity that are consistent with the race/ethnicity groups enumerated in the 199 Census [9]. Estimates of the racial/ethnic populations of Hawaii are the results of local surveys performed periodically by the Hawaii Department of Health [8, 1]. Incidence and death rates were age adjusted where indicated using the 2 U.S. standard population and age-specific rates based on 5-year age groups. The annual percentage change (APC) is a summary statistic that indicates the trend over a defined time period. It is obtained by fitting a regression line through the log of data points for the time period using weighted least squares. The slope of the line is tested for significant increases or decreases. These methods are described in greater detail in the SEER Cancer Statistics Review, 1975-2 [11]. Relative survival is the observed survival (the proportion of cancer patients surviving for a specified time period) adjusted for expected mortality [12]. While the observed survival is calculated taking into account those dying of the given cancer as well as of all other causes, relative survival provides an estimate of the likelihood that cancer patients will not die from causes associated with their cancer. Since the relative survival rate estimates only the effect of the cancer, it is always larger than the observed survival. The relative survival rate measures the survival of the patient cohort compared with the component of the general population having the same characteristics as the patient cohort with respect to age, race, sex, and calendar period. Generally, this means that the relative survival rate measures the effect of the cancer alone, because it is usually the only factor that makes the patient cohort different from the general population. However, sometimes the patients in a cohort may have some other factor that places them at a greater risk for dying compared with the general population. For example, there is a higher percentage of smokers among lung cancer patients than among the general population. Smokers tend to be at greater risks for other diseases, such as heart disease. Relative survival cannot separate the risk of death from lung cancer from the risk of dying of noncancer causes due to smoking. Therefore, the relative survival rate for lung cancer is an underestimate of the effect of lung cancer alone. Complete prevalence for cancer cases is provided as of January 1, 2. In this instance, prevalence is defined as the number of people in a population who are alive on a certain date and who previously had a diagnosis of cancer. Complete prevalence is estimated by applying the completeness index method described elsewhere [13]. Calculation of U.S. prevalence as of January 1, 2 takes into account adjustments for racial differences between the SEER Program data and the total U.S. population and updated population estimates resulting from the 2 Census data. Lifetime probabilities of developing cancer were determined by applying age-specific cancer rates from 1998-2 from SEER-12 Areas to a hypothetical population of 1,, live births. This population is considered to be at risk of developing cancer or of dying from noncancerrelated causes before developing cancer. Modeling techniques described elsewhere [14] were then used to calculate lifetime probabilities. It is estimated that, in 23, 9, children under the age of 15 will be diagnosed with cancer and 1,5 will die of cancer. Unlike adult cancers, pediatric and adolescent cancers are often best described by a combination of histologic type and primary site [15]. Rates for individual pediatric/adolescent cancers also vary widely by age. For these reasons, it was not possible to adequately discuss both adult and childhood/pediatric cancers here. Consequently, only data on adult cancers are presented. Data on childhood/adolescent cancers were reviewed recently in another publication [16]. Cancers included in this report are generally those for which the estimated number of new cases in the U.S. in the year 23 exceeds 15, [17]. RESULTS AND DISCUSSION The distribution of age at diagnosis by primary cancer site is shown in Table 1 for the time period 1996-2. More than 4% of cases of bone and joint cancers, cancer of the cervix uteri, Hodgkin lymphoma, Kaposi s sarcoma, acute lymphocytic leukemia, cancer of the testis, and thyroid cancer were diagnosed in patients younger than 45 years old. In addition to these, the majority of cases (>5%) of cancers at the following sites were diagnosed before age 65: brain and other nervous system, female breast, anus, anal canal and anorectum, corpus uteri and uterus, NOS, eye and orbit, skin (melanoma), oral cavity and pharynx, ovary, and soft tissue including heart. Table 2 shows incidence rates and median ages at diagnosis for the more frequently occurring cancers in males and females. The incidence rate in females for all sites combined is approximately 75% that of males. Among males, the cancer with the highest incidence rate is prostate, 17 per 1, males, followed by lung, with a rate slightly less than half that of prostate, and colon. The three most frequently occurring cancers among women are breast, with an incidence rate of 135 per 1, women, lung, and colon. For cancers that are common in both males and females, the

Ries, Reichman, Lewis et al. 544 Table 1. Age distribution at diagnosis by cancer site for the period 1996-2 Age (in years) at diagnosis (%) Cancer site/type <45 45-54 55-64 65-74 75+ All sites 11 13 19 28 3 Bladder, urinary 3 8 16 3 42 Bones and joints 59 12 1 9 1 Brain and other nervous system 35 14 15 18 18 Breast, female 13 22 21 22 22 Cervix uteri 43 23 14 11 1 Colon/rectum 4 9 17 28 42 anus, anal canal, and anorectum 15 21 19 22 24 colon excluding rectum 4 8 15 28 45 rectum and rectosigmoid junction 6 13 2 28 33 Corpus uteri and uterus, NOS 8 18 25 26 24 Esophagus 3 12 22 32 31 Eye and orbit 29 13 15 21 22 Gallbladder 3 9 15 28 44 Hodgkin lymphoma 65 12 8 8 7 Kaposi s sarcoma 69 16 5 3 8 Kidney and renal pelvis 1 15 22 27 26 Larynx 5 15 27 33 21 Leukemia 23 1 13 22 33 acute lymphocytic leukemia 8 6 4 5 5 acute myeloid leukemia 21 1 13 23 33 chronic lymphocytic leukemia 3 1 18 29 41 Liver and intrahepatic bile duct 7 16 2 27 3 Lung and bronchus 2 9 21 35 33 Melanoma of the skin 26 19 17 19 19 Mesothelioma 3 8 17 31 41 Myeloma 4 11 18 29 38 Non-Hodgkin lymphoma 16 14 16 24 3 Nose, nasal cavity, and middle ear 14 14 18 23 3 Oral cavity and pharynx 11 19 22 25 23 Ovary 18 19 19 21 22 Pancreas 3 9 17 29 42 Penis 8 14 19 25 35 Pleura 16 8 1 2 45 Prostate 7 25 39 29 Small intestine 9 15 2 26 31 Soft tissue including heart 35 14 14 16 21 Stomach 6 1 16 27 41 Testis 84 12 3 1 1 Thyroid 47 21 14 1 8 Ureter 2 5 14 33 46 Vagina 1 14 18 23 34 Vulva 12 14 13 22 38 Source: NCI s SEER program (SEER-12 areas). incidence rates tend to be higher in males. For example, the incidence rate of cancer of the colon/rectum in females is about 72% that found in males. An exception is thyroid cancer where the incidence rate is higher in females. Cancer occurs primarily at older ages with the median age at diagnosis being 68 for males and 65 for females. The median age at diagnosis among more commonly occurring cancers was 65 or under among both men and women for cancer of the brain and other central nervous system (CNS) cancers, melanoma, cancers of the oral cavity and pharynx, and thyroid cancer. In addition, the median age at diagnosis was 65 or under among women for cancers of the breast, corpus uteri and uterus, NOS, and ovary. Among men, this was also the case for cancers of the kidney and renal pelvis and the liver and intrahepatic bile duct and for leukemia and non-hodgkin lymphomas.

545 SEER Program Cancer Survival and Incidence Table 2. Incidence rates and median ages at diagnosis by selected cancer sites for the period 1996-2 Males Females Cancer site/type Incidence rate Median age at diagnosis Incidence rate Median age at diagnosis All sites 555.8 68 417.9 65 Brain and other CNS 7.7 53 5.3 54 Breast 135. 62 Colon/rectum 64.2 7 46.7 74 colon 44.9 71 34.9 74 rectum 19.3 68 11.8 71 Corpus uteri and uterus, NOS 24.1 65 Kidney and renal pelvis 15.4 64 7.7 67 Leukemia 15.8 65 9.3 67 Liver and intrahepatic bile duct 9. 65 3.4 72 Lung and bronchus 8.8 7 49.6 7 Melanoma 21. 6 13.5 53 Non-Hodgkin lymphoma 23.5 64 15.6 69 Oral cavity and pharynx 16.3 62 6.5 65 Ovary 16.8 59 Pancreas 12.6 71 9.8 75 Prostate 17.1 69 Stomach 13. 71 6.2 74 Thyroid 3.6 51 9.9 45 Urinary bladder 36. 72 9.1 72 Note: Incidence rates are per 1, population and are age adjusted to the 2 U.S. standard million population by 5-year age groups. = not applicable. Source: NCI s SEER program (SEER-12 areas). Cancer incidence rates vary by race/ethnicity, as shown in Table 3. For all cancer sites combined, among males, African Americans have the highest rates, while among females, Caucasians have the highest rates. For all cancer sites combined and for colorectal and lung cancers, males have higher incidence rates than females for each racial/ Table 3. Incidence rates (1996-2) and trends (APC) (1992-2) by race/ethnicity and gender for selected cancer sites American Cancer All African Indian/ Asian/ site/type races Caucasian American Alaska Native Pacific Islander Hispanic Rate APC Rate APC Rate APC Rate APC Rate APC Rate APC All sites male and female 472.3 479.8 521.7 239.6 341.7 352.1 male 555.8 555.9 696.8 259. 392. 419.3 female 417.9 431.8 46.3 229.2 36.9 312.2 Colon/rectum male and female 54.2 53.9 62.6 34.7 46.9 4. male 64.2 64.1 72.4 37.5 57.2 49.8 female 46.7 46.2 56.2 32.6 38.8 32.9 Lung and bronchus male and female 62.6 63.2 81.2 33.1 43.2 33.2 male 8.8 79.4 12.4 45.6 62.1 46.1 female 49.6 51.9 54.8 23.4 28.4 24.4 Breast female 135. 14.8 121.7 58. 97.2 89.8 Prostate male 17.1 164.3 272.1 53.6 1. 137.2 Rates are per 1, and are age adjusted to the 2 U.S. standard. = APC shows a statistically significant increase. = APC shows a statistically significant decrease. Hispanic data excludes cases from Detroit, Hawaii, and the Alaska native registry. Source: NCI s SEER program (SEER-12 areas).

Ries, Reichman, Lewis et al. 546 ethnic group observed. For both colorectal and lung cancers, African Americans have higher incidence rates than other racial/ethnic groups among both males and females, although rates for lung cancer in females are similar among African Americans and Caucasians. For breast cancer, Caucasian females have the highest incidence rates. For prostate cancer, African-American males have the highest incidence rates, 1.7 times that of Caucasians. While cancer incidence rates for the American Indian/Alaska Native (AI/AN) population of the U.S. overall are often found to be lower than those for other race/ethnic groups, these rates vary widely geographically [18], and rates for some SEER Program registries are significantly higher than rates for the U.S. population as a whole. This is particularly true for rates of cancers of the lung and colon/rectum reported by the Alaska Native Tumor Registry [19]. Trends shown as the APC in incidence rates are also provided in Table 3. The only statistically significant increases in APC for incidence rates during the period 1992-2 were found for breast cancer among Caucasians, Asian/Pacific Islanders, and Hispanics. Statistically significant decreases in APC were observed for all cancer sites combined among males of all race/ethnic groups observed and for males and females combined, with the exception of Hispanics. For females, however, the APC remained essentially stable over this time period, with the exception of breast cancer. Decreases in APC were observed for lung cancer in Caucasian, African-American, and Hispanic males and when males and females were considered together, for all racial/ethnic groups except Asian/Pacific Islanders. Decreasing APCs were observed in prostate cancer for all racial/ethnic groups shown except Asian/Pacific Islanders. The pattern of cancer death rates (Table 4) is similar to that for incidence in that racial/ethnic groups with higher incidence rates tend to have higher death rates. One anomaly is breast cancer, where African-American females have a higher rate than Caucasian females for mortality but a lower rate for incidence. Trends show that the only statistically significant increases in APC in death rates for the period 1992-2 occurred for lung cancer among Caucasian and African-American females. Significant decreases in death rates were observed for Caucasians and African Americans of both sexes for cancer at all sites combined and for colorectal cancer, as well as for female breast cancer and for prostate and lung cancer among men. Decreases in APC for all sites for males and females combined were observed for all groups shown except the AI/AN population. Roughly 1 in 2 males and 1 in 3 females will be diagnosed with some form of cancer during their lifetimes (Table 5). The estimate for the total U.S. population of the prevalence of individuals alive on January 1, 2 who had a previous diagnosis of cancer at any time in the past is about Table 4. Death rates (1996-2) and trends (APC) (1992-2) by race/ethnicity and gender for selected cancer sites American Cancer All African Indian/ Asian/ site/type races Caucasian American Alaska Native Pacific Islander Hispanic Rate APC Rate APC Rate APC Rate APC Rate APC Rate APC All sites male and female 22.3 199.1 257.1 138. 124.5 137.9 male 255.5 249.5 356.2 172.3 154.8 176.7 female 168.3 166.9 198.6 115.8 12. 112.4 Colon/rectum male and female 21.2 2.7 28.5 14.7 13.1 14.3 male 25.8 25.3 34.6 18.5 15.8 18.4 female 18. 17.5 24.6 12.1 11. 11.4 Lung and bronchus male and female 56.8 56.7 66.4 37.2 28.5 25.8 male 79.5 78.1 17. 52.9 4.9 4.7 female 4.7 41.5 4. 26.2 19.1 15.1 Breast female 27.7 27.2 35.9 14.9 12.5 17.9 Prostate male 32.9 3.2 73. 21.9 13.9 24.1 Rates are per 1, and are age adjusted to the 2 U.S. standard. = APC shows a statistically significant increase. = APC shows a statistically significant decrease. Hispanic data excludes cases from Connecticut, New Hampshire, New York, and Oklahoma. Source: the National Center for Health Statistics.

547 SEER Program Cancer Survival and Incidence Table 5. Lifetime probability (LTP) of developing cancer (1998-2) and complete prevalence (PREV) by gender as of January 1, 2 Males Females Cancer Site/Type LTP PREV LTP PREV All sites 1 in 2 4,241,699 1 in 3 5,313,613 Brain and other CNS 1 in 151 52,73 1 in 191 44,979 Breast 1 in 7 2,197,54 Colon and rectum 1 in 17 499,18 1 in 18 542,481 Corpus uteri and uterus, NOS 1 in 38 556,64 Kidney and renal pelvis 1 in 69 119,642 1 in 116 82,431 Liver and intrahepatic bile duct a 1 in 117 8,711 1 in 246 5,6 Lung and bronchus 1 in 13 174,547 1 in 17 167,91 Melanoma 1 in 55 267,432 1 in 82 283,428 Non-Hodgkin lymphoma 1 in 48 162,16 1 in 57 152,294 Oral cavity and pharynx 1 in 72 142,935 1 in 151 78,473 Ovary 1 in 59 22,949 Pancreas 1 in 82 11,364 1 in 83 12,17 Prostate 1 in 6 1,637,28 Stomach 1 in 81 33,95 1 in 132 25,622 Thyroid 1 in 333 65,22 1 in 119 227,533 Urinary bladder 1 in 29 353,533 1 in 91 125,63 Source: NCI s SEER Program Prevalence is estimated for the total U.S. as described in Materials and Methods. LTP is based on SEER Program data. a For liver and intrahepatic bile duct, completeness index was approximated using historical data from the Connecticut tumor registry. = not applicable. 9.6 million. This estimate includes individuals who were recently diagnosed with cancer, those who have survived many years after diagnosis, those with active cancer, and those who are cancer free. The cancers with the highest prevalence are those that have both high incidence rates and are associated with relatively good survival. For example, there are over 2 million women alive today who had a previous diagnosis of breast cancer, and more than 1.6 million men are alive today who have had a previous diagnosis of prostate cancer. On the other hand, lung cancer has a high incidence rate but a generally poor prognosis. Therefore, there are fewer lung cancer survivors than survivors of some of the less frequently diagnosed cancers, like melanoma. The prevalences of lung cancer survivors are more similar for males and females than expected, given the much higher incidence rate of lung cancer among men, but are also a reflection of the better survival rates and longer life expectancies of women. Other cancer sites shown in Table 5 with high lethalities are the pancreas, stomach, and liver. Relative 5-year survival, shown in Table 6, varies considerably with stage and cancer site. Lung cancer relative survival is low in comparison with the other cancers shown in Table 6. For patients diagnosed with stage I lung cancer, the 5-year relative survival rate is 56%, compared with only 2% for those diagnosed with stage IV disease. However, only 12.2% of lung cancers are diagnosed at stage I. Therefore, the overall relative survival rate is low, only 15% at 5 years. For many cancers, the overall relative survival rate is high, over 9%, when the cancer is diagnosed early (stage I). For most cancers, patients have a low relative survival if the cancer has metastasized before diagnosis. Cancers that are associated with good prognoses, or high relative survival rates for all stages combined, tend to have been diagnosed more frequently at earlier stages than those with overall poor prognoses. For cancer sites shown in Table 6, survival is similar for males and females except for urinary bladder cancer, where males have higher 5-year relative survival rates for all stages, and for stage I lung cancer, where females have a higher 5-year relative survival rate. The clinical relevance of using only 5-year relative survival is somewhat limited. It does not indicate the extent to which prognosis may change as patients have survived several years after diagnosis. Table 7 presents 5-year relative survival rates conditional on patients surviving 1 year, 2 years, and 3 years after diagnosis. That is, it presents the probability of surviving 5 additional years given that the patient has already survived 1 year, 2 years, or 3 years from the time of diagnosis. For some cancers, patients who have survived for 3 years after diagnosis have a very good chance of surviving for an additional 5 years even though the 5-year relative survival rate from the time of diagnosis may be poor. For example, for lung cancer, the 5-year relative survival rate from diagnosis is only 15%, but the 5-year survival rate rises to about 75% for patients who have already survived 3 years

Ries, Reichman, Lewis et al. 548 Table 6. Five-year relative survival rates from SEER Program data for selected cancer sites by gender, 199-1999 All stage Stage I Stage II Stage III Stage IV Unknown Cancer Site/Type 5-year 5-year 5-year 5-year 5-year 5-year Male and female Colon and rectum 62% 95% 82% 57% 7% 61% colon 62% 96% 84% 58% 6% 58% rectum 62% 92% 73% 56% 8% 65% Corpus uteri and uterus, NOS 84% 98% 82% 62% 28% 71% Female breast 86% 1% 85% 58% 19% 8% Lung 15% 56% 32% 9% 2% 16% Ovary 53% 94% 78% 45% 19% 49% Prostate 96% 1% 1% 1% 53% 1% Urinary bladder 82% 97% 65% 56% 22% 77% Male Colon and rectum 63% 95% 82% 57% 6% 65% colon 63% 96% 84% 58% 6% 64% rectum 61% 91% 73% 55% 7% 65% Lung 13% 53% 31% 9% 2% 13% Prostate 96% 1% 1% 1% 53% 1% Urinary bladder 84% 98% 68% 58% 26% 79% Female Colon and rectum 62% 95% 82% 58% 7% 58% colon 62% 96% 84% 58% 7% 52% rectum 63% 93% 73% 57% 8% 65% Corpus uteri and uterus, NOS 84% 98% 82% 62% 28% 71% Breast 86% 1% 85% 58% 19% 8% Lung 17% 6% 33% 1% 2% 19% Ovary 53% 94% 78% 45% 19% 49% Urinary bladder 75% 94% 57% 52% 15% 7% since diagnoses. Similarly, for colorectal cancer, the 5-year relative survival rate from time of diagnosis is 62%, but it rises to 91% for patients who have already survived 3 years. Observed survival is also presented in Table 7. The observed 5-year survival rate provides an estimate of the percent of those diagnosed with cancer who survive 5 years after diagnosis and reflects the effects of both noncancer and cancer-related causes of death. For example, the 5-year relative survival rate for prostate cancer is 96%, whereas the observed survival rate is only 74%. This effect becomes more pronounced with increasing age at diagnosis due to the greater probability of deaths from other causes. For cancer of the lung and bronchus, 5-year relative and 5-year observed survival rates are similar among males and females. In the case of cancers that are rapidly fatal, relative and observed survival rates may be very similar regardless of age. The most frequently quoted survival statistics are for 5 years. However, data in Figure 1 make it clear that, for some cancer sites, survival for time periods other than 5 years may be equal or more important milestones. For cancers of the lung and bronchus, the relative survival curve flattens greatly at 3 years. There is a similar effect, although somewhat less pronounced, for cancers of the corpus uteri and uterus, NOS, and of the colon/rectum. Changes in relative survival over time become clearer when examined by stage at diagnosis. Figure 2 shows data for cancers of the colon/rectum (A), lung (B), female breast (C), and ovary (D). In the cases of colorectal and lung cancers diagnosed at stage IV, there are clear changes in the relative survival curves at about 3 years. This is also the case for lung cancers diagnosed at stage III. In the cases of stages I, II, and III colorectal cancer, relative survival rates at 8 years are more than 5%. In the cases of stage IV breast and ovarian cancers, relative survival declines more gradually than for either stage IV colorectal or lung cancers. However, relative survival does not level off in the same fashion. The relative survival rates for stage I ovarian and breast cancers are over 9% at 8 years.

549 SEER Program Cancer Survival and Incidence Table 7. Five-year relative, conditional, and observed survival rates from SEER Program data for selected cancer sites by gender, 199-1999 5-year relative 5-year relative 5-year relative 5-year conditional on 1 year conditional on 2 years conditional on 3 years 5-year Cancer Site/Type relative after diagnosis after diagnosis after diagnosis observed Male and female Colon and rectum 62% 75% 84% 91% 49% colon 62% 77% 86% 92% 48% rectum 62% 72% 8% 87% 5% Corpus uteri and uterus, NOS 84% 9% 94% 97% 74% Female breast 86% 88% 92% 95% 77% Lung 15% 35% 58% 75% 12% Ovary 53% 66% 76% 86% 47% Prostate 96% 97% 98% 99% 74% Urinary bladder 82% 89% 94% 96% 64% Male Colon and rectum 62% 74% 83% 9% 48% colon 63% 77% 85% 91% 48% rectum 61% 7% 79% 86% 49% Lung 13% 33% 56% 74% 1% Prostate 96% 97% 98% 99% 74% Urinary bladder 84% 9% 94% 97% 65% Female Colon and rectum 62% 76% 85% 92% 5% colon 62% 77% 86% 93% 49% rectum 63% 73% 81% 89% 52% Corpus uteri and uterus, NOS 84% 9% 94% 97% 74% Breast 86% 88% 92% 95% 77% Lung 17% 37% 6% 76% 14% Ovary 53% 66% 76% 86% 47% Urinary bladder 75% 87% 92% 96% 61% Percent 1 75 5 25 1 2 3 4 5 6 7 8 Survival (years) Prostate Female breast Corpus uteri and uterus, NOS Urinary bladder Colon and rectum Ovary Lung and bronchus CONCLUSIONS For those areas covered by the SEER Program, a decline in cancer incidence began in 1992 and then stabilized in 1995 [2]. While the APCs shown here for cancer at all sites combined show an overall decrease for the period 1992-2, examination of a shorter period, 1995-2, no longer shows a decrease in incidence [11]. Thus, time trends for a longer period, while providing greater stability, lack the capacity to point out short-term changes in trends. Shorter term incidence trends may provide evidence of changing circumstances and point to the need for rapid increased vigilance and/or actions in terms of issues such as population compliance with screening recommendations or with preventive interventions, such as smoking cessation, or for new approaches possibly including chemoprevention. Incidence rates for cancers at all sites combined reflect a mixture of data at individual cancer sites and for different subgroups of the population. Decreases in rates for Figure 1. Survival by year from diagnosis by cancer site for cases diagnosed from 199-1999.

Ries, Reichman, Lewis et al. 55 A 1 B 1 75 75 Percent 5 Percent 5 25 25 Stage I C 1 1 2 3 4 5 6 7 8 Survival (years) D 1 1 2 3 4 5 6 7 8 Survival (years) Stage II Stage III Stage IV Unknown 75 75 Percent 5 Percent 5 25 25 1 2 3 4 5 6 7 8 Survival (years) 1 2 3 4 5 6 7 8 Survival (years) Figure 2. Survival by year from diagnosis by stage for cancer cases diagnosed from 199-1999 for colon/rectal cancer (A), lung and bronchial cancer (B), female breast cancer (C), and ovarian cancer (D). some cancer sites or for some population groups may be offset by increases in others. Incidence rates are more sensitive to temporal effects of screening than are mortality statistics. This was observed, for example, in the 197s with the effects of mammography on increases in breast cancer incidence [21]. Screening is affected by the sensitivity of the test, the prevalence of undetected disease in a population, and the frequency of screening in the target population as a whole and in various groups. The stabilization of incidence rates that was seen between 1995 and 2 compared with 1992-1994 may indicate that increased benefit from various preventive/diagnostic strategies will require additional efforts in dissemination or higher population compliance with recommended screening recommendations or that new approaches/interventions are needed for further decreases in incidence rates to be achieved. Alternatively, there may be changes in causative factors that need to be explored and addressed. Survival statistics, which provide information on prognosis, are influenced by many factors including stage at diagnosis, timely treatment, and available treatment options. While

551 SEER Program Cancer Survival and Incidence there are several ways of calculating relative survival statistics, in general, relative survival for a given time period is measured by examining individuals receiving a cancer diagnosis in that time period. It thus has cohort effects and can be greatly influenced by changes in treatment patterns such as the introduction of new, more effective chemotherapy regimens. Mortality, on the other hand, measures deaths due to cancer in a given time period, irrespective of when patients were diagnosed. Death rates, then, are measuring deaths for a group of individuals who may have had very different treatment options available. For this reason, death rate statistics are much less sensitive than survival rate statistics to the introduction of treatment advances. This can lead to a situation where survival for a given time period is increasing while little change in mortality is observed. Recent reports have documented the decline in cancer mortality that began in the early 199s [22, 23]. The decline in cancer mortality for the total U.S. population was statistically significant in 1994 through 1998. In the period 1998-2 the rates leveled off. This may be partially due to a change in the classification of deaths due to cancer. Mortality reflects a mixture of cancer diagnoses made at various points in the past, with differences in available diagnostic practices, treatment options, and levels of supportive care. Much of the overall mortality decline has been due to those cancers for which interventions have been introduced into the general population. Site-specific interventions include prevention programs such as that for smoking cessation, screening programs such as mammography, pap smears, and colorectal cancer screening and treatment advances. Thus, a significant part of the overall decline in cancer mortality appears to be due to the success of various national initiatives sponsored by the NCI. During the diagnostic work-up of a cancer patient, a physician may be interested in the short- and long-term prognoses experienced by similar populations of cancer patients. While 5 years is often referred to as a milestone in survival, for some cancer sites, other time periods may have greater prognostic significance. Relative survival also changes with time since diagnosis. In counseling patients, relative survival for time periods other than 5 years and relative survival conditional on survival for a specific time period since diagnosis can be calculated, based on currently available data, for population groups by age, stage at diagnosis, gender, and possibly other factors such as race/ethnicity and histologic type. Observed survival, taking into account all causes of death, is always less than relative survival. The magnitude of this difference may vary with age and comorbidities for population groups. In the cases of cancers associated with high lethalities there may be little difference between observed and relative survival rates. The NCI has recently begun an initiative to increase research on rare cancers of high lethality [24]. This includes cancer sites such as the pancreas, esophagus, and liver. The group designated as AI/AN is composed of a number of different tribes in a variety of geographic areas. Despite reported cancer incidence rates for this combined group that are lower than for the U.S. population as a whole, rates for this group in several, individual SEER Program registries are higher than overall SEER Program rates and than rates for the U.S. population as a whole [18]. Reasons for this include regional differences among population segments, different levels of case reporting and identification of race/ethnicity, and variations in cancer risk factors. While the AI/AN population in Alaska can be enumerated more easily due to geographic definition, it is particularly difficult to enumerate the AI/AN population in most areas of the U.S. To a lesser extent, similar issues are applicable to the Hispanic and to Asian/Pacific Island groups as well. These issues point to the continued need for more detailed information on subgroups of the U.S. population. Cancer statistics are a foundation of our ability to measure progress against cancer. A high level of continuous vigilance is necessary to use these statistics to detect changes in risk factors, including environmental and lifestyle effects. In addition, there continue to be new initiatives and research on prevention and treatment that may impact cancer incidence, survival, and mortality. To this end, the SEER Program cancer statistics can be used to monitor changes and to point to areas that need closer attention from the oncology and public health communities. ACKNOWLEDGMENTS The authors wish to thank Ms. Danielle Harkins of Information Management Systems, Inc., for technical assistance. REFERENCES 1 Hankey BF, Ries LA, Edwards BK. The surveillance, epidemiology, and end results program: a national resource. Cancer Epidemiol Biomarkers Prev 1999;8:1117-1121. 2 http://www.seer.cancer.gov/about, accessed October 27, 23. 3 Percy C, Van Holten V, Muir C, eds. International Classification of Diseases for Oncology, Second Edition. Geneva: World Health Organization, 199. 4 Beahrs OH, Henson DE, Hutter RVP et al., eds. Manual for Staging of Cancer, Third Edition. American Joint Commission on Cancer. Philadelphia: J.B. Lippincott Co., 1988. 5 Clegg LX, Li FP, Hankey BF et al. Cancer survival among US whites and minorities: a SEER (surveillance, epidemiology and end results) program population-based study. Arch Intern Med 22;162:1985-1993.

Ries, Reichman, Lewis et al. 552 6 Boscoe FP, Miller BA. Population estimation error and its impact on 1991-1999 cancer rates. Professional Geographer (in press). 7 National Center for Health Statistics. Division of Data Services. U.S. Census Populations with Bridged Race Categories. Bridged-Race Population Estimates for July 1, 2-July 1, 22. Available at: http://www.cdc.gov/nchs/about/major/dvs/ popbridge/popbridge.htm, accessed 7/17/23. 8 National Cancer Institute. Surveillance, Epidemiology, and End Results Program. U.S. Population Data 1969-2. Available at: http://seer.cancer.gov/popdata/, accessed 7/17/3. 9 National Cancer Institute. NCI Fact Sheet: How Changes in U.S. Census Counts Affect NCI Cancer Rates, posted 4/15/3. Available at: http://www.cancer.gov/newscenter/pressreleases/ Census2, accessed 7/17/3. 1 National Cancer Institute. Surveillance Epidemiology and End Results Program. Population Estimates Used in NCI s SEER*Stat Software. Available at: http://seer.cancer.gov/pop data/methods.pdf, accessed 7/17/3. 11 Ries LAG, Eisner MP, Kosary CL et al., eds. SEER Cancer Statistics Review, 1975-2. Bethesda, MD: National Cancer Institute, 23. http://seer.cancer.gov/csr/1975_2, accessed 7/17/3. 12 Ederer F, Axtell LM, Cutler SJ. The relative survival rate: a statistical methodology. J Natl Cancer Inst Monogr 1961;6:11-121. 13 Capocaccia R, De Angelis R. Estimating the completeness of prevalence based on cancer registry data. Stat Med 1997;16:425-44. 14 Feuer EJ, Wun LM, Boring CC et al. The lifetime risk of developing breast cancer. J Natl Cancer Inst 1993;85:892-897. 15 International Classification of Childhood Cancer 1996. International Association of Cancer Registries (IARC) Technical Report No. 29, Lyon 1996. 16 Ries LAG, Smith MA, Gurney JG et al., eds. Cancer Incidence and Survival among Children and Adolescents: United States SEER Program 1975-1995. NIH Pub. No. 99-4649. Bethesda, MD: National Cancer Institute, 1999. 17 American Cancer Society. Cancer Facts and Figures 23. Atlanta, GA: American Cancer Society, 23:4. 18 Swan J, Edwards BK. Cancer rates among American Indians and Alaska Natives: is there a national perspective? Cancer 23;98:1262-1272. 19 Lanier AP, Kelly JJ, McEvoy T et al. Alaska Native Cancer Update 1987-1999. Anchorage, AK: Alaska Native Tribal Health Consortium, 22. Refer to the chapter titled Rates: Alaska Natives vs. U.S. 2 Hankey BF, Ries LA, Kosary CL et al. Partitioning linear trends in age-adjusted rates. Cancer Causes Control 2;11:31-35. 21 Miller BA, Feuer EJ, Hankey BF. Recent incidence trends for breast cancer in women and the relevance of early detection: an update. CA Cancer J Clin 1993;43:27-41. 22 Ries LAG, Eisner MP, Kosary CL et al., eds. SEER Cancer Statistics Review, 1973-1999. Bethesda, MD: National Cancer Institute, 22. Available at http://www.seer.cancer.gov/csr/ 1973_1999/, accessed October 27, 23. 23 Edwards BK, Howe HL, Ries LAG et al. Annual report to the nation on the status of cancer, 1973-1999, featuring implications of age and aging on the U.S. cancer burden. Cancer 22;94:2766-2972. 24 http://www.cancer.gov/researchfunding/announcements, accessed October 27, 23.