Trends in Radiation Therapy among Cancer Survivors in the United States, 2000–2030

The number of cancer survivors in the United States has steadily increased over the past several decades, reaching an estimated 15.5 million individuals in 2016 (1). Of these, 10.3 million survivors (two thirds) were alive 5 years or more following their cancer diagnosis. The demographics of cancer survivors are also changing, with 63% of survivors projected to be 65 years or older by 2020 (2). These trends reflect improved cancer survival and underlying shifts in the age distribution of the U.S. population (1, 3). The expanding number and aging of cancer survivors are expected to increase national health expenditures due to a greater comorbidity burden compounded by the long-term demand for medical care associated with complications of cancer therapy (3).

Radiation therapy plays a central role in the management of cancer. Unlike other treatment modalities, the most burdensome toxicities of radiation often manifest years after treatment. Reducing the long-term late effects of radiation is a central research initiative within radiation oncology; however, the epidemiology of the cancer survivors who were treated with radiation is poorly defined. In this study, we use cancer registry and census data to describe and project trends in the population of radiation-treated, 5-year cancer survivors from 2000 projected through 2030.

Cancer incidence data were obtained from the National Cancer Institute's (NCI's) Surveillance, Epidemiology, and End Results (SEER) database. SEER collects data from cancer registries across the United States and includes information about all incident cases of cancer within a defined population-based region. We studied the original 9 SEER registries (Utah, Hawaii, Iowa, Connecticut, New Mexico, and the greater areas of Detroit, San Francisco-Oakland, Atlanta, and Seattle-Puget Sound, representing about 9% of the population), as these contain data from the beginning of the SEER program in 1973 (3). We analyzed cases from 1975 (due to under-reporting in 1973 and 1974) through 2013 (the most recent year available). We categorized cases into the following 21 cancer sites: anus, bladder/kidney, bones and soft tissue, brain and other nervous system, breast, cervix, colon, esophagus, head and neck, leukemia, lung, lymphoma, myeloma, pancreas and hepatobiliary, prostate, rectum, skin excluding basal and squamous, stomach, uterus, testis, and other. We considered only the first diagnosed malignant tumor and excluded nonmalignant tumors. SEER records the delivery of radiation only when administered during the first course of treatment for an incident cancer. The definition of radiation in this study included either external beam radiation (any modality) or brachytherapy and excluded radioisotopes.

Population data were obtained from the U.S. Census intercensal population estimates for each calendar year, sex, and age (single-year), including projections to 2030. Population projections are calculated from the most recent decennial census (2010) and incorporate assumptions about future birth rates, death rates, and international migration patterns (4). All-cause mortality rates were obtained from the Social Security Administration.

We used the Prevalence Incidence Approach Model (PIAMOD) to estimate prevalence using trends in cancer incidence, cancer survival, and all-cause mortality (3, 5). The PIAMOD approach uses incidence and relative cancer survival rates obtained from registry data to create projected estimates of incidence and survival beyond the range of known years. In this study, we used SEER data between 1975 and 2013 to calculate age-, sex-, and cancer site–specific incidence rates. We assumed that incidence rates from 2014 to 2030 would follow incidence trends from 2004 to 2013 (the 10 most recent years of data; ref. 3). For each age-, sex-, and site-specific group, we fit a linear equation to the yearly log-transformed incidence rates from SEER between 2004 and 2013 with the following equation

where I_observed(y) represents the observed incidence in year y and α and β are regression coefficients. We determined the predicted incidence for years 2014–2030 with the following equation

where I_predicted(y) is the predicted incidence in year y, I_average is the average incidence over the years 2011–2013, and β is the regression coefficient from Equation (A). This approach to calculating predicted incidence has the advantage of incorporating historic incidence trends while using the most recent 3-year average incidence as an anchor point for the prediction.

Observed survival was determined from the SEER-9 data, and expected survival was calculated from all-cause mortality data from the Social Security Administration. To allow for survival projections after 2013, we fit a mixture survival model to the existing data (5). The mixture survival model is a parametric model that assumes a fraction of the cancer cohort will be “cured” of cancer and experience the same risk of death as the non-cancer population. For the mixture survival model, we assumed that among those who die of cancer, their survival time would follow a Weibull distribution. The relative survival S_relative(t,y) at time t for patients diagnosed with cancer in year y was defined as

where P(y) represents the fraction of cured patients in year y, represented by

and S_cancer(t,y) represents cancer-specific survival at time t in year y per the Weibull distribution (3, 6)

In the above equations, y₀ represents an arbitrary reference year (1975) and parameters a, b, c, d, and e represent regression parameters from the model. The above mixture survival model was fit for each disease site, sex, and age group (0–44, 45–54, 55–64, 65–74, and 75–84 years old; ref. 3). The above relative survival equations require closed age ranges; therefore, patients older than 85 years were assumed to have the same relative survival as the 75–84 age group.

Total and projected prevalence were determined by projecting the above age-, sex-, and site-specific incidence and relative survival estimates onto U.S. census data. Our primary analysis as presented focused on radiation-treated survivors alive a minimum of 5 years, although secondary analyses evaluated different lengths of minimum survivorship ranging from 1 to 10 years (Supplementary Table S1). Our projection of survivor numbers assume that incidence and survival trends would continue based on historical trends, per Equations (B) and (C). We tested 3 other scenarios in a sensitivity analysis where we assumed (i) constant incidence, (ii) constant relative survival, and (iii) both constant incidence and constant relative survival. In these sensitivity analyses, we assumed that the “constant incidence” scenario would use the average incidence from 2011 to 2013 [I_average from Equation (A)] for years 2014 through 2030. In the “constant relative survival” scenario, we used the relative survival from the year 2013 (last known year) for years 2014 through 2030. The sensitivity analyses are presented in Supplementary Table S2. All data analysis was performed with SAS v9.4 (SAS Institute).

In 2016, there were an estimated 10.5 million 5-year cancer survivors, of whom 3.05 million received radiation therapy (Fig. 1A). The number of radiation-treated survivors is projected to reach 3.38 million by 2020 and 4.17 million by 2030. In 2016, radiation-treated survivors constituted 29% of all cancer survivors (Fig. 1B). The fraction of all cancer survivors who received radiation increased from 24% in 2000 to a projected maximum of 29% in 2020. After 2020, the fraction of radiation-treated survivors is projected to slightly decline to 28% by 2030.

The absolute number of radiation-treated survivors increased from 2000 to 2030 for most cancer sites (Table 1, Fig. 2, and Supplementary Table S2) with the largest absolute increases coming from breast and prostate cancer. In 2016, there were an estimated 1.25 million radiation-treated breast cancer survivors and 713,000 radiation-treated prostate cancer survivors. The number of radiation-treated breast cancer survivors increased steadily from 2000, and by 2030, this number is projected to reach 2.01 million patients. The fraction of breast cancer survivors treated with radiation increased from 29% in 2000 to 54% in 2016, reaching 60% by 2030. The number of radiation-treated prostate cancer survivors increased from 266,000 in 2000 to a projected maximum of 721,000 in 2020, after which the number is projected to decline to 627,000 in 2030.

In lymphoma, cervical, and brain cancer, the absolute number of radiation-treated survivors is projected to increase but the fraction of cancer survivors treated with radiation is projected to decrease. For example, in 2000, there were 135,000 radiation-treated lymphoma survivors which constituted 42% of all lymphoma survivors. By 2030, the number of radiation-treated lymphoma survivors is projected to reach 234,000, but the fraction of survivors treated with radiation will decrease to 25%. With head and neck, rectal, and lung cancer, the absolute number and fraction of cancer survivors treated with radiation are both projected to increase. In head and neck cancer, there were 105,000 radiation-treated survivors (47% of all head and neck cancer survivors) in 2000, and by 2030, this increases to 242,000 and 56%.

The age distribution of radiation-treated cancer survivors has shifted steadily toward older ages (Table 2 and Fig. 3). In 2016, an estimated 1.59 million radiation-treated survivors were older than 70 years, representing 52% of all radiation-treated survivors. This number increases to 2.70 million by 2030, constituting 65% of all radiation-treated survivors. The absolute number of radiation-treated survivors younger than 70 years is projected to remain relatively stable over the study period (Fig. 3A), although the fraction is projected to decrease substantially (Fig. 3B). Similarly, the relative number of pediatric and young adult radiation-treated survivors (those younger than 40 years) is projected to decrease from 2.8% of all radiation-treated survivors in 2016 to 2.3% in 2030.

The projected sex distribution of radiation-treated cancer survivors shifts toward females for all age groups (Table 2). The total number of female radiation-treated survivors is projected to increase from 1.78 million in 2016 to 2.75 million in 2030. Between 2000 and 2016, the fraction of female radiation-treated survivors remained stable at 58% and then is projected to increase to 66% in 2030.

The purpose of this study was to describe the landscape of 5-year cancer survivors treated with radiation therapy. The number and composition of cancer survivors who receive radiation has changed over the past several decades and this will continue in years to come. In 2016, an estimated 3.05 million 5-year cancer survivors received radiation, which we project will increase to 4.17 million survivors by 2030. The wide array of radiation-treated survivors across multiple disease sites emphasizes the broad use of radiation in cancer therapeutics.

Breast cancer accounted for the greatest number of radiation-treated cancer survivors, with 1.25 million survivors in 2016 and a projected steady increase to 2.01 million by 2030. Pivotal trials reported in the 1980s and 1990s established the role of adjuvant radiotherapy after lumpectomy in early-stage breast cancer (7–9), and additional trials demonstrated the benefits of radiation in select subsets of women after mastectomy (10, 11). In our study, the large number of breast cancer survivors arose from the frequent use of radiation in breast cancer coupled to the high incidence and effective treatment for this disease (12). Breast cancer also drove the increased relative proportion of female radiation-treated survivors from 58% in 2000 to 66% in 2030. Our sensitivity analysis demonstrated relative stability in future projections of the number of breast cancer radiation survivors; however, unforeseen changes in breast cancer treatment in the future could influence these estimates.

Prostate cancer comprised the second largest group of radiation-treated cancer survivors in 2016. Unlike in breast cancer, the projected number of radiation-treated prostate cancer survivors is projected to peak in 2020 at 721,000 and decrease thereafter. This decline likely comes from the decreasing incidence of prostate cancer, which mostly reflects changes in prostate-specific antigen (PSA) screening (13, 14). In addition, changes in patterns of treatment including more men choosing active surveillance may reduce the number of men undergoing radiation therapy (15, 16). The landscape of prostate cancer survivorship may continue to evolve with the recently published PROTECT study that found similar cancer-specific survival among active surveillance, surgery, and radiation in early-stage prostate cancer (17, 18).

While breast and prostate cancer together account for the majority of all long-term radiation-treated survivors, other disease sites deserve discussion. Head and neck cancer represents a steadily increasing survivor population, and by 2030, it will represent the third most common group of radiation-treated cancer survivors. The shifting epidemiology in head and neck cancer with increasing numbers of human papilloma virus–related cancers will likely alter the demographics of this survivorship population in the years to come (19, 20).

From a survivorship perspective, one must consider the potential for long-term or delayed toxicity from radiation therapy. Breast cancer historically carried a nontrivial risk of cardiovascular mortality from radiation to the heart, but in recent years, this risk has decreased because of improved efforts to spare the heart from radiation (21, 22). However, sparing the heart is often difficult in other thoracic cancers including Hodgkin disease, lung cancer, and esophageal cancer, and these survivors have an increased risk of cardiovascular mortality (23–25). Radiation directed at the pelvis in prostate, rectal, anal, and gynecologic cancers carries risks of long-term rectal and bladder toxicity, as well as risks of sexual dysfunction (26–29). Radiation therapy to the head and neck may be curative for many patients, but radiation in this region bears a risk of toxicities that have the potential to decrease quality and length of life, including salivary dysfunction, radiation fibrosis, dental disease, taste alterations, and osteoradionecrosis (30). Finally, radiation treatment in any part of the body carries a small but real lifetime risk of a secondary radiation–induced malignancy (31, 32).

The increasing age of radiation-treated cancer survivors reflects the increasing age of overall cancer survivors and the aging of the U.S. population (2). The elderly cancer survivor population stands apart with respect to their healthcare needs. Elderly cancer survivors have more functional physical limitations, higher levels of lost productivity, and poorer overall health outcomes (33–35). This population also has higher healthcare costs compared with individuals without cancer (3).

While the number of elderly cancer survivors treated with radiation is projected to increase, the fraction of young adults and pediatric cancer survivors treated with radiation is projected to decrease, consistent with long-term declines in radiation therapy utilization for many pediatric cancers (36). However, with U.S. population growth and improved survival among pediatric patients with cancer, the absolute number of radiation-treated pediatric cancer survivors is expected to increase through 2030. This indicates a continuing need to manage the long-term effects of radiation therapy in pediatric cancer survivors, including increased risks of secondary malignancy, growth deficiency, and neurocognitive deficits (37). As a result, this growing population will continue with special healthcare needs extending into the foreseeable future (38–40).

Our methodology has several limitations. First, while PIAMOD is a standard method to estimate complete cancer prevalence, it requires estimation of incidence and mortality beyond the interval for which data are available (2, 3). In our sensitivity analyses (presented in the Supplementary Material), we found that survivorship predictions for certain disease sites—particularly prostate cancer—depended on assumptions about continuing future incidence and survival trends. While our method relied on projections from historic trends, any future variability in cancer incidence or survival will impact the projected numbers of radiation-treated cancer survivors. Second, our projections assume no change in current patterns of radiation therapy utilization that might substantially change our projections. For example, our breast cancer projections assume that radiation therapy will play the same role in local control as it does in current practice. Third, recent studies in breast cancer suggest that SEER may underestimate the true rates of radiation delivery (41, 42). In addition, SEER only records radiation during a patient's first course of treatment, which implies that patients receiving radiation for recurrent or metastatic disease beyond their first course of treatment would not be included. However, patients receiving radiation beyond their first treatment course likely have recurrent or metastatic disease and therefore would be less likely to survive for 5 years or more. Finally, our study included only invasive malignant disease; therefore, patients with noninvasive histologies such as breast ductal carcinoma in situ or benign central nervous system tumors would not be included in our survivorship estimates. Together, these limitations suggest that our radiation-treated survivorship estimates may be low compared with the true number of radiation-treated cancer survivors in the United States.

Despite these limitations, this study provides a description of current and future trends in the population of cancer survivors who received radiation. From a survivorship perspective, radiation represents a unique challenge in that long-term side effects may take years to manifest. Increased research and awareness of survivorship issues are needed to care for this growing segment of the population.

No potential conflicts of interest were disclosed.

Conception and design: A.K. Bryant, J.D. Murphy

Development of methodology: A.K. Bryant, J.D. Murphy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Bryant, J.D. Murphy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Bryant, L.K. Mell, J.D. Murphy

Writing, review, and/or revision of the manuscript: A.K. Bryant, M. Banegas, M.E. Martinez, L.K. Mell, J.D. Murphy

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.K. Bryant, L.K. Mell, J.D. Murphy

Study supervision: J.D. Murphy

This study was supported by NIHTL1TR00098 (A.K. Bryant).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.