Temporal Trends in Treatment and Subsequent Neoplasm Risk Among 5-Year Survivors of Childhood Cancer, 1970-2015

Key Points

Question  Are treatment-era related changes in chemotherapy or radiation therapy doses associated with changes in the risk of subsequent neoplasms over time among survivors of childhood cancer?

Findings  In this longitudinal cohort study of 23 603 survivors of childhood cancer, reductions in therapeutic radiation doses over time were associated with reduced rates of subsequent neoplasms, including subsequent malignancies, nonmelanoma skin cancers, and benign meningiomas.

Meaning  Ongoing efforts to reduce long-term therapeutic toxicity were associated with decreasing subsequent neoplasms among 5-year survivors of childhood cancer.

Importance  Cancer treatments are associated with subsequent neoplasms in survivors of childhood cancer. It is unknown whether temporal changes in therapy are associated with changes in subsequent neoplasm risk.

Objective  To quantify the association between temporal changes in treatment dosing and subsequent neoplasm risk.

Design, Setting, and Participants  Retrospective, multicenter cohort study of 5-year cancer survivors diagnosed before age 21 years from pediatric tertiary hospitals in the United States and Canada between 1970-1999, with follow-up through December 2015.

Exposures  Radiation and chemotherapy dose changes over time.

Main Outcomes and Measures  Subsequent neoplasm 15-year cumulative incidence, cumulative burden, and standardized incidence ratios for subsequent malignancies, compared by treatment decade. Multivariable models assessed relative rates (RRs) of subsequent neoplasms by 5-year increments, adjusting for demographic and clinical characteristics. Mediation analyses assessed whether changes in rates of subsequent neoplasms over time were mediated by treatment variable modifications.

Results  Among 23 603 survivors of childhood cancer (mean age at diagnosis, 7.7 years; 46% female) the most common initial diagnoses were acute lymphoblastic leukemia, Hodgkin lymphoma, and astrocytoma. During a mean follow-up of 20.5 years (374 638 person-years at risk), 1639 survivors experienced 3115 subsequent neoplasms, including 1026 malignancies, 233 benign meningiomas, and 1856 nonmelanoma skin cancers. The most common subsequent malignancies were breast and thyroid cancers. Proportions of individuals receiving radiation decreased (77% for 1970s vs 33% for 1990s), as did median dose (30 Gy [interquartile range, 24-44] for 1970s vs 26 Gy [interquartile range, 18-45] for 1990s). Fifteen-year cumulative incidence of subsequent malignancies decreased by decade of diagnosis (2.1% [95% CI, 1.7%-2.4%] for 1970s, 1.7% [95% CI, 1.5%-2.0%] for 1980s, 1.3% [95% CI, 1.1%-1.5%] for 1990s). Reference absolute rates per 1000 person-years were 1.12 (95% CI, 0.84-1.57) for subsequent malignancies, 0.16 (95% CI, 0.06-0.41) for meningiomas, and 1.71 (95% CI, 0.88-3.33) for nonmelanoma skin cancers for survivors with reference characteristics (no chemotherapy, splenectomy, or radiation therapy; male; attained age 28 years). Standardized incidence ratios declined for subsequent malignancies over treatment decades, with advancing attained age. Relative rates declined with each 5-year increment for subsequent malignancies (RR, 0.87 [95% CI, 0.82-0.93]; P < .001), meningiomas (RR, 0.85 [95% CI, 0.75-0.97]; P = .03), and nonmelanoma skin cancers (RR, 0.75 [95% CI, 0.67-0.84]; P < .001). Radiation dose changes were associated with reduced risk for subsequent malignancies, meningiomas, and nonmelanoma skin cancers.

Conclusions and Relevance  Among survivors of childhood cancer, the risk of subsequent malignancies at 15 years after initial cancer diagnosis remained increased for those diagnosed in the 1990s, although the risk was lower compared with those diagnosed in the 1970s. This lower risk was associated with reduction in therapeutic radiation dose.

As cure rates for childhood cancers have improved, there has been an increasing awareness of the late health consequences of childhood cancer therapies. One outcome associated with significant morbidity and mortality for these survivors is the development of subsequent neoplasms, unique from recurrence of the original childhood malignancy.¹ Survivors with a subsequent neoplasm are more likely to report adverse general and mental health outcomes² and have increased hospitalization rates³ compared with survivors without a subsequent neoplasm. Subsequent malignant neoplasms are the most common nonrelapse cause of late mortality, accounting for approximately one-half of all nonrelapse deaths among 5-year survivors.^1,4

Quiz Ref IDThe Childhood Cancer Survivor Study (CCSS)⁵ and other cohorts of childhood cancer survivors⁶ have reported extensively on the incidence of and risk factors for subsequent neoplasms.^7-9 Therapeutic radiation has been strongly associated with development of subsequent neoplasms¹⁰; however, dose-response relationships have also been identified for specific chemotherapeutic agents, including alkylating agents and epipodophyllotoxins.^11-13 With this knowledge, childhood cancer treatment has been modified over time with the hope of reducing subsequent neoplasm risk, while maintaining or improving 5-year survival.^14,15

The CCSS cohort was recently expanded to include survivors diagnosed and treated across 3 decades (1970-1999). The aim of the present analysis was to assess temporal changes in subsequent neoplasms among individuals diagnosed during this period. We hypothesized that historical modifications in radiotherapy and chemotherapy dosing would be associated with changes in the incidence of subsequent neoplasms among 5-year childhood cancer survivors, based on their decade of diagnosis, throughout the course of follow-up.

The CCSS is a retrospective cohort study with longitudinal follow-up of 5-year survivors of childhood cancer diagnosed at 1 of 27 participating institutions in the United States or Canada between January 1, 1970, and December 31, 1999. Participants were younger than 21 years at initial diagnosis of leukemia, Hodgkin lymphoma, non-Hodgkin lymphoma, central nervous system cancer, Wilms tumor, neuroblastoma, rhabdomyosarcoma, or bone cancer. Human subjects committee approval was granted at participating institutions prior to recruitment. Participants, or parents of children younger than 18 years, provided informed consent. Minor participants were reconsented once they reached 18 years.

Participants completed a baseline and up to 4 follow-up questionnaires. Race/ethnicity data were obtained through self-report. Participants were asked to select white, black, American Indian or Alaska Native, Asian or Pacific Islander, or other, with the option to write in their race. Hispanic ethnicity was ascertained through a yes/no question. Race and ethnicity were included in the analysis for descriptive purposes. The final date of follow-up was December 31, 2015. The study design and methods have been described.^5,16

Quiz Ref IDSubsequent neoplasms were identified via self- or next-of-kin proxy report or death certificate and confirmed by pathology report or, when unavailable, death certificate, medical records, or both. Only subsequent neoplasms occurring 5 years or more after initial cancer diagnosis were evaluated. Subsequent neoplasms were classified into 3 mutually exclusive groups, based on historical experience with frequently occurring neoplasm subtypes: (1) subsequent malignant neoplasms, which include invasive neoplasms classified as International Classification of Diseases for Oncology (ICD-O, third version) behavior code of 3,¹⁷ excluding nonmelanoma skin cancers; (2) benign meningiomas; and (3) nonmelanoma skin cancers (including ICD-O morphology codes 8070, 8071, 8081, 8090, and 8094). Cancer therapies, including surgery, chemotherapy, and radiation, were ascertained through medical record abstraction, as described.^16,18 Cumulative alkylating agent dose was reported as a cyclophosphamide equivalent dose.¹⁹ Maximum radiation treatment dose was calculated for 8 body regions (brain, other head, neck, chest, abdomen, pelvis, arm, and leg) for each patient. For this analysis, we considered any radiation treatment (yes/no) for cumulative incidence and cumulative burden estimates and maximum doses for multivariable models.

Cohort follow-up started at 5 years from diagnosis and ended on death or date of last completed questionnaire. Cumulative burden, assessed using the mean cumulative count,^20,21 and cumulative incidence were estimated using time from initial diagnosis as the time scale, treating death as a competing risk event. Cumulative burden is an estimate of the mean number of subsequent neoplasms per 100 survivors by a given time, in the presence of competing risk events, and accounts for multiple events in individuals, whereas cumulative incidence only accounts for the first event. Cumulative incidence and cumulative burden at 15 years from diagnosis were compared across treatment decades using permutation tests.

For subsequent malignant neoplasms, standardized incidence ratios (SIRs) (ratio of the observed to expected number of events) and absolute excess risk per 1000 person-years were calculated using age-, sex-, and calendar-year–specific US cancer incidence rates from the Surveillance, Epidemiology, and End Results program to determine expected numbers of events.²²

Because comparison by treatment era is subject to confounding by attained age, SIRs were calculated stratifying on 10-year age intervals. Multivariable piecewise-exponential models were used to assess the incidence rate of subsequent neoplasm types, in association with demographic variables and childhood cancer diagnosis, adjusting for attained age, treatment doses, and 5-year treatment eras. Reference absolute rates per 1000 person-years were calculated using the fitted model for survivors with reference characteristics (no chemotherapy, splenectomy, or radiation therapy; male; attained age of 28 years). Multiple subsequent neoplasm occurrences within individual survivors were included and accounted for in the models by modifications of the models using generalized estimating equations. Adjusted relative rates (RRs) and 95% confidence intervals were estimated.

In addition, using mediation analysis methods previously described,^4,23-25 changes in subsequent neoplasm rates in 5-year treatment era increments were estimated with and without adjustment for treatment variables in the same model, to assess whether changes in subsequent neoplasm rates over time were mediated by treatment modifications. Specifically, as shown in the causal diagram that depicts the assumptions of the mediation analysis (eFigure 1 in the Supplement), a multivariable piecewise-exponential model was fit to assess the association of treatment era with subsequent neoplasm rates, adjusting for attained age, sex, age at initial cancer diagnosis, and treatment variables (the full model), followed by removal of treatment variables from the model. Attenuation and the statistical significance of the treatment era regression coefficient, the parameter representing the adjusted log rate ratio of subsequent neoplasms by 5-year treatment increments, by the inclusion of treatment variables in the model, constitute the key step of establishing the mediator role of treatment variables, along with the associations of treatment era and treatment variables and those of treatment variables and subsequent neoplasm rates in the full model. Nonparametric bootstrap was used to test statistical significance of the changes in the regression coefficient associated with the 5-year treatment era with and without adjustment for treatment variables. For analyses examining treatment doses, only individuals with available treatment data were included.

All tests were 2-sided, with P < .05 considered statistically significant. SAS version 9.4 was used for all statistical analyses including the mediation analysis, and R version 3.2.4 was used for statistical graphics.

Among the 23 603 eligible, consented survivors (Figure 1), 46% were female, the mean age at primary diagnosis was 7.7 years, and the most common initial diagnoses were acute lymphoblastic leukemia (ALL), Hodgkin lymphoma, and astrocytoma. Mean follow-up ranged from 15.7 years for survivors diagnosed in the 1990s to 27.6 years for those diagnosed in the 1970s. Over the course of 374 638 person-years at risk, 1639 survivors experienced 3115 subsequent neoplasms, including 1026 subsequent malignancies, 233 benign meningiomas, and 1856 nonmelanoma skin cancers (Table 1). The distribution of subsequent neoplasms by primary cancer diagnosis is reported in eTable 1 in the Supplement and the distribution of observed and expected subsequent malignancies, by decade of initial cancer diagnosis, is reported in eTable 2 in the Supplement. The most frequently observed subsequent malignancies were breast and thyroid cancer.

Complete treatment data were available for 83% of the cohort. Between 1970-1999, there were substantial changes in therapies. Radiation therapy decreased from 77% of survivors treated in the 1970s to 54% in the 1980s and 33% in the 1990s. Median radiation treatment dose decreased from 30 Gy (interquartile range, 24-44) in the 1970s to 26 Gy (interquartile range, 18-52) in the 1990s. Although the proportion of children treated with alkylating agents and anthracyclines increased over time, median doses decreased. The proportion of children treated with epipodophyllotoxins and platinum agents also increased over the 3 decades; however, whereas the median cumulative dose of platinum increased with each treatment decade, the median cumulative dose of epipodophyllotoxins increased substantially in the 1980s and then decreased in the 1990s (Table 1).

At 15 years from initial diagnosis, the cumulative incidence of subsequent neoplasms was 2.9% (95% CI, 2.5%-3.3%) among individuals diagnosed in the 1970s, 2.4% (95% CI, 2.1%-2.7%) among those diagnosed in the 1980s, and 1.5% (95% CI, 1.3%-1.8%) among those diagnosed in the 1990s (1970s vs 1980s, P = .02; 1970s vs 1990s, P < .001; 1980s vs 1990s, P < .001) (Figure 2). At 15 years, the cumulative burden of subsequent neoplasms per 100 survivors was 3.6 among those diagnosed in the 1970s, 2.8 among those diagnosed in the 1980s, and 1.7 among those diagnosed in the 1990s (1970s vs 1980s, P = .02; 1970s vs 1990s, P < .001; 1980s vs 1990s, P = .001) (Figure 2). After 20 years from diagnosis, among survivors from the 1970s and 1980s, the steep increase in cumulative burden was secondary to recurrent nonmelanoma skin cancer events.

A significantly lower 15-year cumulative incidence of subsequent malignancies was observed in those diagnosed in the 1990s (1.3% [95% CI, 1.1%-1.5%]), compared with the 1980s (1.7% [95% CI, 1.5%-2.0%]; P < .001) and the 1970s (2.1% [95% CI, 1.7%-2.4%]; P < .001) (Figure 2; eTable 3 in the Supplement). A similar decline was seen for nonmelanoma skin cancers but not for meningiomas. When assessing incidence by primary cancer diagnosis, declines between decades were seen for Hodgkin lymphoma and Wilms tumor, but only survivors of Hodgkin lymphoma demonstrated a statistically significant decrease in 15-year cumulative incidence of subsequent neoplasms across decades (eFigure 2 in the Supplement). Among the most common subsequent malignancies, only soft-tissue sarcomas (0.26% [95% CI, 0.13%-0.38%] for the 1970s vs 0.13% [95% CI, 0.06%-0.21%] for the 1990s, P = .03) and breast cancers (0.27% [95% CI, 0.14%-0.40%] for the 1970s vs 0.08% [95% CI, 0.02%-0.14%] for the 1990s, P = .003) had significant decreases in 15-year cumulative incidence from the 1970s to the 1990s (eFigure 3 in the Supplement).

Survivors treated with radiation experienced a higher cumulative incidence of all types of subsequent neoplasms, for all treatment decades (eFigure 4 in the Supplement). Cumulative burden for nonmelanoma skin cancers, compared with cumulative incidence, showed a more pronounced difference based on receipt of radiation therapy, exemplifying the number of radiation-exposed survivors with multiple events. Among irradiated survivors, a significant decrease in the 15-year cumulative incidence of nonmelanoma skin cancers was observed for the most recent treatment decade (1970s, 1.0% [95% CI, 0.7%-1.3%]; 1980s, 0.9% [95% CI, 0.6%-1.1%]; 1990s, 0.2% [95% CI, 0.1%-0.4%]; 1970s vs 1980s, P = .27; 1970s vs 1990s, P < .001; 1980s vs 1990s, P < .001).

Reference absolute rates per 1000 person-years were 4.21 (95% CI, 3.05-5.81) for subsequent neoplasms, 1.12 (95% CI, 0.84-1.57) for subsequent malignancies, 0.16 (95% CI, 0.06-0.41) for meningiomas, and 1.71 (95% CI, 0.88-3.33) for nonmelanoma skin cancers. Lower SIRs were observed by decade of diagnosis for survivors whose attained age was 20 to 29 years (1970s, 5.7 [95% CI, 4.7-6.7]; 1980s, 4.8 [95% CI, 4.0-5.6]; 1990s, 3.6 [95% CI, 2.7-4.6]; P = .004) and 30 to 39 years (1970s, 5.6 [95% CI, 4.8-6.4]; 1980s, 4.9 [95% CI, 4.1-6.0]; 1990s, 3.1 [95% CI, 1.8-5.0]; P = .03) (Figure 3; eTable 4 in the Supplement). Decreases in SIRs across treatment decades within specific subsequent malignancy types were not observed (eFigure 5 in the Supplement). Among survivors of Hodgkin lymphoma with attained age 20 years or older, SIRs for subsequent malignancies decreased over time (20-29 years: 1970s, 10.7 [95% CI, 7.7-14.4]; 1980s, 6.8 [95% CI, 4.4-10.0]; 1990s, 5.5 [95% CI, 3.2-9.0]; P = .02 and 30-39 years: 1970s, 10.2 [95% CI, 8.2-12.5]; 1980s, 8.4 [95% CI, 6.3-11.0]; 1990s, 5.2 [95% CI, 2.4-9.8]; P = .04) (eTable 5 in the Supplement).

Multivariable analysis demonstrated that female survivors experienced increased rates of subsequent malignant neoplasms (RR, 1.7 [95% CI, 1.5-2.0]; P < .001) and meningiomas (RR, 1.4 [95% CI, 1.0-2.0]; P = .05) compared with male survivors. Treatment with high doses of alkylating agents and platinum agents were also associated with increased rates of subsequent malignancies, and therapeutic radiation at all dose increments was associated with increased rates of subsequent malignant neoplasms, meningiomas, and nonmelanoma skin cancers (Table 2; crude data reported in eTable 6 in the Supplement).

After adjusting for sex, age at diagnosis, and attained age, RRs declined for every 5-year increment of treatment era for subsequent neoplasms (RR, 0.81 [95% CI, 0.76-0.86]; P < .001), subsequent malignant neoplasms (RR, 0.87 [95% CI, 0.82-0.93]; P < .001), meningiomas (RR, 0.85 [95% CI, 0.75-0.97]; P = .03), and nonmelanoma skin cancers (RR, 0.75 [95% CI, 0.67-0.84]; P < .001) (Table 3). Inclusion of all treatment variables in the model attenuated statistically significantly the treatment era–associated declines in rates of subsequent neoplasms (P < .001), subsequent malignant neoplasms (P < .001), meningiomas (P = .03), and nonmelanoma skin cancers (P < .001). Further mediation analyses were conducted by modifying the adjustment from all treatment variables to specific components of treatment variables (ie, maximum radiation dose and all other treatments, including chemotherapy drug doses and splenectomy). These analyses revealed that radiation therapy dose changes were the chief contributor to the era-associated decline of subsequent neoplasm rates and that radiation therapy dose changes were the only component of the treatment variables significantly associated with the decline of subsequent neoplasm rates over time (Table 3).

Survival after childhood cancer has improved substantially over the last 5 decades. As the number of survivors has increased, so has the focus on late outcomes of cancer therapy. Cohort studies, including the CCSS, devoted to understanding the late health consequences of childhood cancer therapies⁶ have previously documented the effect of subsequent neoplasms and quantified risk associated with specific therapies, particularly therapeutic radiation.^7,26-28 Accordingly, efforts have been directed toward eliminating the use of radiation therapy when possible or decreasing the volume, dose, or both.^14,15 An example of this is the near elimination of cranial radiation among children newly diagnosed with ALL.²⁹ As treatment with radiation has decreased, some chemotherapy regimens have intensified.¹⁴ This evolution in delivered therapies has reduced late mortality among survivors⁴; however, the association with specific outcomes, including subsequent neoplasms, has not been investigated. Quiz Ref IDThe current analysis, including more than 23 000 survivors of childhood cancer treated over 3 decades, demonstrated that the cumulative incidence rates of subsequent neoplasms, subsequent malignant neoplasms, meningiomas, and nonmelanoma skin cancers were lower among survivors treated in more recent treatment eras and that modifications of primary cancer therapy were associated with these declines.

Survivors of Hodgkin lymphoma are at particularly high risk for subsequent malignancies.^30-33 The study findings demonstrated a decreased 15-year cumulative incidence of subsequent neoplasms for Hodgkin lymphoma survivors treated in the 1990s compared with earlier decades, with cumulative incidence of subsequent malignant neoplasms at 15 years significantly lower in the 1990s compared with the 1970s. In contrast, a recent report of Dutch long-term survivors of Hodgkin lymphoma demonstrated that the cumulative incidence of second cancers did not decrease across treatment decades.³⁴ Within the current study, the SIR for subsequent malignant neoplasms decreased over time among survivors of Hodgkin lymphoma with attained age of 20 years or older. Intensified alkylator dosing has been used in Hodgkin lymphoma since the 1980s to compensate for decreasing therapeutic radiation.¹⁴ At high cumulative doses, alkylating agents are associated with increased rates of subsequent malignant neoplasms, which may have attenuated the expected decline in subsequent malignancies among survivors of Hodgkin lymphoma, particularly among the Dutch cohort, for which a decrease in incidence was not observed.

Quiz Ref IDTemporal treatment modifications have also been made using improved risk stratification of children with ALL and Wilms tumor,^14,15 among others, which have led to decreased late mortality from subsequent malignant neoplasms among survivors.⁴ In nearly all patients with ALL, cranial irradiation has been replaced with intensive intrathecal therapy. Meningiomas are among the most frequently observed subsequent neoplasms in survivors of ALL, and, given the latency of more than 20 years to development of meningiomas,⁷ it is likely that the full effect of omitting cranial irradiation among more recently treated survivors has not yet been observed. Temporal changes in Wilms tumor therapy include reduction in the dose and even elimination of therapeutic radiation in low-risk populations, which has likely contributed to the decreased cumulative incidence of subsequent malignant neoplasms in the 1990s compared with the 1970s.

These data further document the increased cumulative incidence, cumulative burden, and elevated risk for subsequent malignant neoplasms, meningiomas, and nonmelanoma skin cancers in survivors treated with radiation therapy.⁷ Despite reduced use of therapeutic radiation, radiation continues to be an important component of treatment for many children. As the use of radiation therapy decreases, it is possible that other associations may emerge. Specifically, to maintain and improve cure rates, chemotherapy dosing, the proportion of patients receiving various agents, or both have been increasing. Although there has been a decrease in the median cumulative doses of alkylating agents and anthracyclines, the proportion of survivors receiving these agents has increased. For epipodophyllotoxins and platinum agents, increased median cumulative doses were given in the 1990s compared with the 1970s, and the proportion of survivors treated with these agents increased over time. There was an increased rate of subsequent malignant neoplasms with higher cumulative doses of alkylating agents and platinum agents. These associations may become more prominent as use of therapeutic radiation continues to decline. Additionally, the role of genetic susceptibility in the development of subsequent neoplasms may become more evident, and interactions between host genetics and chemotherapy doses may emerge as well.

This study represents, to our knowledge, the first comprehensive report of subsequent neoplasms from the CCSS since the expansion of the cohort to include survivors initially diagnosed through 1970-1999. Both the British Childhood Cancer Survivor Study³⁵ and a cohort of childhood cancer survivors from Nordic cancer registries³⁶ have reported on subsequent neoplasms from patients diagnosed over extended periods (British, 1940-1991; Nordic, 1943-2005); however, this is the first study, to our knowledge, to report changes in subsequent neoplasm incidence and SIRs as a function of treatment era and treatment variables.

The mediation analysis^23-25 demonstrated that the decline of subsequent neoplasm rates was mediated by treatment variable changes over time. This occurred in conjunction with reduction in treatment doses and decreased splenectomy frequency across treatment eras (Table 1) and with statistically significant associations of treatment variables with subsequent neoplasm rates when controlling for era of treatment and treatment variables (Table 2). In this analysis, the age-specific SIRs and overall cumulative incidence of subsequent malignant neoplasms at 15 years consistently decreased for more recent treatment eras. As observed in previous CCSS reports on subsequent neoplasms,³⁷ SIRs are greatest at younger attained ages because they measure observed counts relative to expected counts in the age- and sex-matched general population. Although significant decreases were observed from the 1970s to the 1990s in 15-year cumulative incidence of breast cancer, soft-tissue sarcoma, and nonmelanoma skin cancer, significant decreases were not observed for subsequent leukemia, central nervous system malignancies, or thyroid malignancies (eFigure 3 and eFigure 4 in the Supplement). Furthermore, significant decreases in SIRs for specific subsequent malignancies, including breast cancer, soft-tissue sarcoma, leukemia, central nervous system malignancies, and thyroid malignancies were not observed when stratified by decade of diagnosis and attained age (eFigure 5 in the Supplement).

Quiz Ref IDThe lack of significant changes in individual malignancies is in contrast to the overall decrease in subsequent malignancies with more recent treatment eras and the decrease in use of therapeutic radiation; however, the observed numbers of subsequent malignancies are small for each subgroup and the confidence intervals are wide, indicating possible insufficient power to detect significant differences and also suggesting additional mechanisms contributing to subsequent risk of malignancy. Ongoing follow-up of survivors from the latest treatment decade is needed to determine changes in risk over time, particularly given the long latency from primary diagnosis to many subsequent malignant neoplasms.

This study has important limitations. Although the CCSS is a large, well-characterized cohort, it is not completely representative of the childhood cancer survivor population, and there is potential for selection bias, given that 33% of eligible survivors were not included in this analysis. Selected primary diagnoses, including retinoblastoma, germ cell tumor, and hepatoblastoma, were not included. Children with heritable retinoblastoma are at significant risk of subsequent neoplasms, and their exclusion may have resulted in underestimation of subsequent malignancy risk. Subsequent neoplasms were initially self-reported, which may have led to underreporting of neoplasms, particularly those occurring in the more distant past. Additionally, the design of the CCSS cohort to include only 5-year survivors a priori excludes consideration of cancers occurring prior to 5 years. Also, therapies for subsequent neoplasms are not completely ascertained, which limits further exploration of the role of treatments among survivors who develop multiple subsequent neoplasms. In addition, interpretation of results should be made within the context of the multiple statistical tests performed to address the hypotheses of interest.

Among survivors of childhood cancer, the risk of subsequent malignancies at 15 years after initial cancer diagnosis remained increased for those diagnosed in the 1990s, although the risk was lower compared with those diagnosed in the 1970s. This lower risk was associated with reduction in therapeutic radiation dose.

Corresponding Author: Lucie M. Turcotte, MD, MPH, MS, Division of Pediatric Hematology/Oncology, University of Minnesota, 420 Delaware St SE, MMC 484, Minneapolis, MN 55455 (turc0023@umn.edu).

Author Contributions: Drs Turcotte and Neglia had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Turcotte, Henderson, Leisenring, Armstrong, Robison, Neglia.

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

Drafting of the manuscript: Turcotte, Liu, Neglia.

Critical revision of the manuscript for important intellectual content: Yasui, Arnold, Hammond, Howell, Smith, Weathers, Henderson, Gibson, Leisenring, Armstrong, Robison, Neglia.

Statistical analysis: Liu, Yasui, Neglia.

Obtained funding: Armstrong, Robison, Neglia.

Administrative, technical, or material support: Yasui, Hammond, Smith, Weathers, Gibson, Armstrong, Robison, Neglia.

Supervision: Hammond, Leisenring, Armstrong, Neglia.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Arnold reported receiving grants from the National Cancer Institute (NCI). Ms Smith reported receiving a grant from St Jude’s Children’s Research Hospital. Ms Weathers reported receiving a grant from NCI. Dr Henderson reported receiving a grant from Seattle Genetics. Dr Leisenring reported receiving a grant from NCI. No other authors reported disclosures.

Funding/Support: This work was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health (UL1TR000114, B.R. Blazar, principal investigator) and the National Cancer Institute (CA55727, G.T. Armstrong, principal investigator). Support to St Jude Children’s Research Hospital also provided by the Cancer Center Support (CORE) grant (CA21765, C. Roberts, principal investigator) and the American Lebanese-Syrian Associated Charities (ALSAC).

Role of the Funder/Sponsor: The funders/sponsors of this study 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; or decision to submit the manuscript for publication.