Prevalence of Prior Cancer Among Persons Newly Diagnosed With Cancer: An Initial Report From the Surveillance, Epidemiology, and End Results Program

Key Points

Question  How many patients diagnosed with incident cancer are cancer survivors?

Findings  In an analysis of 740 990 persons in the Surveillance, Epidemiology, and End Results program newly diagnosed with cancer, approximately 25% of older adults (≥65 years) and 11% of younger adults had a history of prior cancer. Prevalence of prior cancer ranged from 4% to 37% according to age and incident cancer type, with most prior cancers diagnosed in a different cancer site.

Meaning  As the population of cancer survivors continues to grow, understanding the nature and impact of a prior cancer is critical to improving clinical trial accrual, generalizability of results from trials and observational studies, disease outcomes, and patient experience.

Importance  The US cancer survivor population is rapidly growing. Cancer survivors are frequently excluded from cancer clinical trials and observational research.

Objective  To examine prevalence of prior cancer among individuals newly diagnosed with cancer.

Design, Setting, and Participants  Linked observations across the population-based Surveillance, Epidemiology, and End Results (SEER) program of cancer registries (1975-2013) for 740 990 persons newly diagnosed with cancer from January 2009 through December 2013. Prevalence of prior cancer was estimated by age (<65 years vs ≥65 years) and incident cancer type.

Main Outcomes and Measures  Prevalence of prior cancer was derived from SEER sequence numbers, which represent the order of all primary reportable tumors diagnosed in a lifetime. Incident cancers were categorized as: (1) first or only primary; (2) second order or higher primary in the same cancer site; and (3) second order or higher primary in a different cancer site.

Results  Of 765 843 incident cancers diagnosed between 2009 and 2013, 141 021 (18.4%) represented a second order or higher primary cancer. Overall, approximately one-fourth (25.2%) of older (≥65 years) and 11% of younger adults newly diagnosed with cancer had a history of prior cancer. Prevalence of prior cancer ranged from 3.5% to 36.9% according to incident cancer type and age, with most prior cancers diagnosed in a different cancer site.

Conclusions and Relevance  A substantial proportion of patients diagnosed with incident cancer in the United States have survived a prior cancer. These patients may be excluded from clinical trials and underrepresented in observational research, and little is known about their treatment and survivorship needs. Understanding the nature and impact of prior cancer is critical to improving clinical trial accrual and generalizability, disease outcomes, and patient experience.

The number of US cancer survivors is rapidly growing, largely driven by the aging population, expanding cancer screening efforts, and improvements in cancer treatment. Over the past 30 years, the cancer survivor population increased 4-fold to 15.5 million in 2016 and is expected to reach 26.1 million by 2040.¹ Almost half of all survivors have lived 10 years after their initial diagnosis, and two-thirds have survived beyond 5 years.² Survivors have complex health needs,³ including surveillance for recurrence, monitoring treatment-related toxic effects, and managing emerging diagnoses, such as chronic conditions⁴ or new primary cancers.⁵

Cancer survivors are frequently excluded from cancer clinical trials. More than 80% of National Cancer Institute–affiliated lung cancer trials exclude patients with a prior cancer.⁶ Such restrictive eligibility criteria may exclude as many as 25% of patients newly diagnosed with lung cancer from participating trials.⁷ Although considerable scientific progress⁵ has been made understanding risk of developing a future primary cancer among specific groups of cancer survivors, this earlier work does not address how many patients diagnosed with incident cancer have survived a prior cancer. Understanding prevalence of prior cancer among patients with different types of incident cancer has important implications for both treatment and research.

We report prevalence of prior cancer among individuals newly diagnosed with cancer from January 2009 to December 2013 using the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) program of cancer registries. We linked observations across SEER 9 registries (Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Puget Sound, and Utah, all from 1975-2013) to estimate prevalence of prior cancer by incident cancer type and age (<65 years vs ≥65 years). Prevalence of prior cancer was derived from SEER sequence numbers, which represent the order of all primary reportable tumors (ie, not metastatic or recurrent tumors) diagnosed in a lifetime. Sequence number “00” indicates an individual has only 1 primary cancer. For persons with multiple primaries, the sequence number for the first cancer is “01,” “02” for the second, and so forth. With few exceptions, including ovarian and prostate cancers, tumors from different anatomic sites, of different histology, or from separate organs of a pair are considered independent primaries.⁸ The University of Texas Southwestern institutional review board approved this study.

We categorized incident cancers as a: (1) first or only primary; (2) second order or higher primary in the same cancer site (eg, 2 melanomas diagnosed at least 1 year apart); or (3) second order or higher primary in a different cancer site. The eAppendix in the Supplement lists tumors classified as belonging to the same or different site. SEER collects the number but not site of cancers diagnosed outside geographically defined registry areas; therefore, some (range, 1.0%-14.3%) cases are categorized as having a prior cancer of an unknown site (data not shown).

For persons with more than 1 cancer diagnosed in the same year (n = 23 150 [3.1% of total]), we were unable to determine the order of diagnoses within that year; therefore, we randomly selected 1 cancer for analysis. The majority of persons diagnosed with more than 1 cancer in the same year (n = 17 420 [75.2% of those with ≥1 cancer in same year and 2.4% of total]) were diagnosed with 2 cancers of the same site (eg, right and left breast cancer).

There were 765 843 incident cancers diagnosed among 740 990 persons from January 2009 to December 2013, of which 141 021 (18.4%) represented a second order or higher primary cancer. The Table shows the proportion of incident cancers diagnosed as the first or only primary or a second order or higher primary of the same or different site. Prevalence of prior cancer differed by age: 11.0% among ages 20 to 64 years and 25.2% among ages 65 years or older (Table). Prevalence also differed by incident cancer type. Among persons age 20 to 64 years, prior cancer was most prevalent among incident myeloid and monocytic leukemia (24.8%); anus, anal canal and rectum (18.2%); cervix and other female genital organs (eg, vagina, vulva; 15.0%); and lung and other respiratory (14.6%) cancers. Prior cancers in this younger age group generally occurred in a different cancer site, although second order breast, cervical, and other female genital, male genital, and testicular cancers were more often in the same site. For patients 65 years and older, incident cancers with highest prevalence of prior cancer were melanoma (36.9%); myeloid and monocytic leukemia (36.9%); bone and joints (34.0%); and urinary bladder and other urinary organs (32.5%). With the exception of breast cancer melanoma, most prior cancers among the older age group occurred in a different site.

One-quarter of older adults (≥65 years) and more than 10% of younger adults newly diagnosed with cancer have a history of prior cancer. Prevalence of prior cancer ranged from 3.5% to 36.9% according to incident cancer type and age, with most prior cancers diagnosed in a different cancer site.

Prior cancer has important implications for cancer care delivery. Patients may have competing priorities concerning treatment decisions: a new diagnosis may interrupt management, treatment adherence, or outcomes related to a prior cancer. Differences in the prevalence of prior cancer by incident cancer type also highlight underlying or shared risk factors that may be amenable to targeted surveillance. For example, 30% or more of older persons diagnosed with cancers attributable to human papilloma virus (eg, cervical and female genital, anal, oral cavity) or tobacco (eg, lung, esophageal, oral cavity) had a prior cancer. An even larger proportion (36.9%) diagnosed with myeloid leukemia had a prior cancer, which may reflect leukemogenic effects of earlier cancer treatments.

Many cancer clinical trials exclude patients with a prior cancer, a practice that may exclude a substantial proportion of otherwise eligible patients. Excluding patients with a prior cancer likely arises from a long-held belief that a prior cancer diagnosis may interfere with study conduct and/or outcomes. However, this restrictive criterion limits generalizability and trial-generated knowledge to patients with a first or only primary—a slight majority of patients with certain cancer types. This is particularly concerning for older adults with uncommon cancers, where trial accrual is critical, standard therapies may be suboptimal, and prior cancer is prevalent.

Determining the impact of prior cancer exclusion criteria on trial accrual requires disease-specific and protocol-specific details, including stage and timing of prior cancer diagnoses.⁹ In lung cancer, most trials use a 5-year exclusion window,⁶ prior cancers generally occur within that window, and having a prior cancer does not adversely impact survival.^7,10,11 Consequently, including patients with a prior cancer in lung cancer trials could substantially improve accrual without affecting study outcomes. The sizable number of cancers newly diagnosed among cancer survivors highlights the importance of addressing similar questions for other cancer types.

Patients with prior cancer are also frequently excluded from observational research, including treatment and outcome studies using SEER-Medicare,¹² Patterns of Care,¹³ Cancer Care Outcomes Research and Surveillance Consortium,¹⁴ and Veterans Health Administration¹⁵ data. Because observational studies often provide real-world data to complement clinical trials, reconsidering the rationale of this eligibility criterion is important to advancing evidence-based practice.

We could not determine order of multiple cancers diagnosed in the same year because only year of diagnosis is available in SEER data (ie, not month or day). Prior cancers diagnosed outside of registry geographic areas are reflected in sequence number only, and there is no corresponding information on the prior cancer characteristics, including site. However, these limitations pertained to fewer than 5% of the total cancer cases diagnosed in the study period and are unlikely to impact our conclusions.

As the cancer survivor population continues to grow, understanding the nature and impact of a prior cancer is critical to improving trial accrual, generalizability of results from trials and observational studies, disease outcomes, and patient experience.

Corresponding Author: Caitlin C. Murphy, PhD, MPH, Division of Epidemiology, Department of Clinical Sciences, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390 (caitlin.murphy@utsouthwestern.edu).

Accepted for Publication: August 18, 2017.

Published Online: November 22, 2017. doi:10.1001/jamaoncol.2017.3605

Author Contributions: Dr Murphy had full access to all the data in the study and takes responsibility for the data and the accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Murphy, Pruitt.

Obtained funding: Gerber.

Study supervision: Gerber, Pruitt.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by the National Cancer Institute (grant No. R03CA191875 to Drs Gerber and Pruitt and grant No. K24CA201543 Dr Gerber) and the National Center for Advancing Translational Sciences at the National Institutes of Health (grant No. KL2TR001103 to Dr Murphy).

Role of the Funder/Sponsor: The funders/sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.