Incidence of and Risk Factors for Histologically Confirmed Solid Benign Tumors Among Long-term Survivors of Childhood Cancer

Key Points

Question

What are the incidence of and treatment-related risk factors for histologically confirmed solid benign tumors in long-term survivors of childhood cancer?

Findings

In this record linkage study of 5843 Dutch 5-year survivors of childhood cancer, 9.3% developed a benign tumor after a median follow-up of 22.7 years and at a median age of 30.0 years; the highest cumulative incidence of benign tumors was found in survivors treated with radiotherapy only (16.7%). Unlike breast fibroadenoma, nervous system tumors, uterine leiomyoma, and osteochondroma appeared to be associated with prior local radiotherapy.

Meaning

These findings in survivors of childhood cancer are important to enable early diagnosis, to develop organ-specific guidelines that include some benign tumors, and to guide in-depth etiologic studies.

This Dutch national record linkage study quantifies the incidence of and treatment-related risk factors for histologically confirmed solid nonskin benign tumors among survivors of childhood cancer.

Abstract

Importance

Survivors of childhood cancer (CCSs) face risk of developing subsequent tumors. Solid benign tumors may be cancer precursors; benign tumors and cancers may share etiologic factors. However, comprehensive data on the risk for solid benign tumors are lacking.

Objective

To quantify the incidence of and treatment-related risk factors for histologically confirmed solid nonskin benign tumors among CCSs.

Design, Setting, and Participants

This record linkage study involves the Dutch Childhood Oncology Group–Long-Term Effects After Childhood Cancer (DCOG-LATER) cohort of 6165 individuals diagnosed with childhood cancer at younger than 18 years from January 1, 1963, through December 31, 2001, in 7 Dutch pediatric centers and who survived at least 5 years after the diagnosis. Study groups eligible for record linkage from 1990 onward included 5843 CCSs (94.8%) and 883 siblings. Benign tumors were identified from the population-based Dutch histopathology and cytopathology registry (PALGA). Follow-up was completed on May 1, 2015. Data were analyzed from January 1, 1990, through May 1, 2015.

Main Outcomes and Measures

Cumulative incidence of any subsequent benign tumor for cohort strata and multivariable Cox proportional hazards regression models (hazard ratios [HRs]) were used to evaluate potential risk factors for 8 major benign tumor subtypes.

Results

Of the 5843 eligible CCSs (55.9% male), 542 (9.3%) developed a histologically confirmed subsequent benign tumor after a median follow-up of 22.7 years (range, 5.0-52.2 years). Among women, abdominopelvic radiotherapy inferred dose-dependent increased risks for uterine leiomyoma (n = 43) for doses of less than 20 Gy (HR, 1.9; 95% CI, 0.5-7.0), 20 to less than 30 Gy (HR, 3.4; 95% CI, 1.1-10.4), and at least 30 Gy (HR, 5.4; 95% CI, 2.4-12.4) compared with no abdominopelvic radiotherapy (P = .002 for trend). High-dose radiotherapy to the trunk was not associated with breast fibroadenoma (n = 45). Of 23 osseous and/or chondromatous neoplasms, 16 occurred among leukemia survivors, including 11 after total body irradiation (HR, 37.4; 95% CI, 14.8-94.7). Nerve sheath tumors (n = 55) were associated with radiotherapy (HR at 31 years of age, 2.9; 95% CI, 1.5-5.5) and a crude indicator of neurofibromatosis type 1 or 2 status (HR, 5.6; 95% CI, 2.3-13.7). Subsequent risk for benign tumors was higher than the risks for subsequent nonskin solid malignant neoplasms and for benign tumors among siblings.

Conclusions and Relevance

This record linkage study uses a unique resource for valid and complete outcome assessment and shows that CCSs have an approximately 2-fold risk of developing subsequent benign tumors compared with siblings. Site-specific new findings, including for uterine leiomyoma, osteochondroma, and nervous system tumors, are important to enable early diagnosis; this information will be the first step for future surveillance guidelines that include some benign tumors in CCSs and will provide leads for in-depth etiologic studies.

Introduction

Survivors of childhood cancer (CCSs) are at high risk for late morbidities, including new tumors. Current research predominantly focuses on the burden of subsequent malignant neoplasms.^1,2,3,4 Benign tumors are of importance too; benign tumors and subsequent malignant neoplasms may share etiologic factors and clinical manifestations that may affect quality of life or life expectancy.⁵ In addition, benign tumors may be cancer precursors, including for example colorectal adenoma⁶ and thyroid nodules,⁷ offering potential opportunities for early detection of precancerous growths.^8,9,10

For some organs, elevated risks in CCSs were shown after radiotherapy, including benign tumors of the central nervous system,^11,12 salivary and/or thyroid gland,^13,14,15 and colorectal tract,^16,17 mostly based on self-reports or retrospective medical record review. Benign tumors can occur as part of specific familial cancer predisposition syndromes, such as neurofibromatosis type 1 or 2 (NF1/NF2).^18,19,20

We provide a comprehensive examination of the incidence of histologically confirmed subsequent benign tumors in long-term CCSs vs siblings. Furthermore, we evaluated treatment-related risk factors for benign tumors among CCSs. Our study used a record linkage approach combining the strengths of the Dutch Childhood Oncology Group–Long-Term Effects After Childhood Cancer (DCOG-LATER) cohort study and the population-based nationwide network and registry of histopathology and cytopathology in the Netherlands (PALGA).²¹

Methods

The DCOG-LATER cohort included 6165 individuals diagnosed with any type of childhood cancer at younger than 18 years from January 1, 1963, through December 31, 2001, in 7 Dutch pediatric centers who survived at least 5 years after the diagnosis. Data on prior diagnosis and treatment of primary tumors and recurrences are available in the DCOG-LATER registry. All cohort members were traced for vital status and address using national and municipal registries. The CCSs who participated in the 2013-2014 questionnaire survey (n = 3172) invited their respective siblings. In all, 883 of 1663 approached siblings (53.1%) were eligible for record linkage studies. The study protocol was declared exempt from review of medical intervention research by the institutional review boards of participating centers. More details were reported elsewhere.¹

Definition and Ascertainment of Subsequent Benign Tumors

Histologically confirmed solid benign tumors (hereinafter referred to as benign tumors) were identified by linkage with PALGA, a database containing results of all pathology examinations performed in the Netherlands, which reached nationwide coverage during 1990.²¹ PALGA pathology reports contain short digital summary excerpts using thesaurus codes (based on Systematized Nomenclature of Medicine codes) to classify the pathologists’ review. The linkage was based on pseudonymized family name, sex, and date of birth. Potentially eligible excerpts were defined as those with benign tumor morphology codes (M8.XXX/0-M9.XXX/0), excluding skin, until May 2015. Resulting excerpts were manually reviewed by 2 of us (J.L.K. and C.M.R.) to identify eligible cases. Where necessary, an experienced late-effects specialist was consulted (H.J.vdP.). Benign tumors were classified according to morphology codes from the International Classification of Diseases for Oncology, Third Edition.²² We used data for subsequent malignant neoplasms to December 31, 2012 (ascertainment previously reported¹) to compare the magnitude of the cumulative incidences of benign tumors and subsequent malignant neoplasms in our cohort and to identify groups of CCSs who developed both.

Statistical Analyses

Data were analyzed from January 1, 1990, through May 1, 2015. Survivors who declined approval for use of health care data (152 [2.5%]) and those who died, emigrated, or were lost to follow-up before 1990 (170 [2.8%]) were excluded. Follow-up started 5 years after the diagnosis of childhood cancer or January 1, 1990, whichever came last, and ended on the date of diagnosis of the first benign tumor, death, last known vital status (emigration, loss to follow-up), or end of the study (May 1, 2015), whichever came first. In analyses of any benign tumor, follow-up ended on the date of first occurring benign tumor; in analyses on specific benign tumor types, follow-up ended on the first occurring benign tumor of that particular type.

Multivariable Cox proportional hazards regression models were used to estimate risks of benign tumors associated with treatment and demographic factors (hazard ratios [HRs]). Cumulative incidences of benign tumors for survivors and siblings were calculated, accounting for death as a competing risk. To estimate risks in CCSs for 4 primary and 4 secondary outcome types, we used multivariable Cox proportional hazards regression models, including more detailed treatment variables (prescribed radiotherapy dose and 6 main chemotherapeutic groups) (eMethods in the Supplement). Risk models for meningioma²³ and colorectal adenoma¹⁷ were reported previously. Attained age was used as the time scale to account for age-related increasing rates of benign tumors.²⁴ The proportional hazards assumption was tested. If violated, risk estimates were presented at the median attained age of all CCSs. P < .05 was considered statistically significant, and all statistical tests were 2 sided. Analyses were performed using Stata software (version 13; StataCorp).

Results

Cohort Characteristics

This analysis included 5843 five-year CCSs (3269 male [55.9%] and 2574 female [44.1%]) who contributed 99 123 person-years at risk from January 1, 1990, through May 1, 2015, and 883 siblings contributing 19 177 person-years. Leukemia (1962 [33.6%]), lymphoma (950 [16.3%]), and central nervous system tumors (781 [13.4%]) were the most frequent childhood cancers (Table 1). For CCSs, median time since childhood cancer diagnosis was 22.7 years (range, 5.0-52.2 years) and median attained age at end of follow-up was 30.0 years (range, 5.8-67.5 years). For siblings, median attained age was 31.7 years (range, 9.5-73.3 years). In total, 542 CCSs and 56 siblings developed 1 or more benign tumors. Among CCSs with a benign tumor, 63 (11.6%) also developed a solid subsequent malignant neoplasm, including 35 with a benign tumor after a solid subsequent malignant neoplasm, 24 with a benign tumor before a solid subsequent malignant neoplasm, and 4 with synchronous benign tumor and subsequent malignant neoplasm diagnoses. In all, 484 cohort members died during follow-up, including 46 with benign tumors; 2 meningioma were listed as confirmed cause of death. After adjustment for attained age and sex, the incidence of benign tumors in CCSs vs siblings was significantly elevated (HR, 1.9; 95% CI, 1.5-2.6).

Table 1. Multivariable Cox Regression Models for Risk of Any Benign Tumor Among 5-Year Survivors of Childhood Cancer in the DCOG-LATER Cohort^a.

Characteristic	No. of Participants		HR (95% CI)
	All	Cases	Model 1b	Model 2c
Sex
Male	3269	247	1 [Reference]	1 [Reference]
Female	2574	295	1.5 (1.3-1.8)	1.5 (1.3-1.8)
Age at childhood cancer diagnosis, y
0-4	2641	223	1.5 (1.2-1.8)	1.3 (1.0-1.7)
5-9	1583	143	1.1 (0.9-1.4)	1.1 (0.9-1.4)
10-17	1619	176	1 [Reference]	1 [Reference]
Period of childhood cancer diagnosis
1963-1984	1791	290	1 [Reference]	1 [Reference]
1985-1994	2113	179	1.4 (1.1-1.8)	1.0 (0.8-1.3)
1995-2001	1939	73	1.4 (1.0-1.9)	1.0 (0.7-1.4)
Treatmentd
Surgery only or no recorded treatment	603	36	1 [Reference]	NA
CT without RT	2898	138	0.9 (0.6-1.3)	NA
RT without CT	445	74	NAe	NA
RT and CT	1854	290	2.5 (1.7-3.5)	NA
Childhood cancer typef
Hodgkin lymphoma	395	39	NA	1 [Reference]
Leukemiag	1962	157	NA	0.9 (0.7-1.4)
Non-Hodgkin lymphomah	555	46	NA	0.9 (0.6-1.4)
Medulloblastoma	153	26	NA	2.0 (1.2-3.3)
Nonmedulloblastoma CNS tumor	628	63	NA	1.4 (0.9-2.0)
Neuroblastoma	313	32	NA	1.1 (0.7-1.8)
Renal tumor	578	50	NA	0.9 (0.6-1.4)
Bone tumor	343	32	NA	0.8 (0.5-1.3)
Soft tissue sarcoma	426	51	NA	1.2 (0.8-1.9)
Germ cell tumor	224	26	NA	1.2 (0.7-2.0)
Other and unspecifiedi	266	20	NA	0.7 (0.4-1.2)

Abbreviations: CNS, central nervous system; CT, chemotherapy; DCOG-LATER, Dutch Childhood Oncology Group–Long-Term Effects After Childhood Cancer; HR, hazard ratio; NA, not applicable; RT, radiotherapy.

^aOwing to missing values, includes 538 cases of first benign tumor.

^bAccounts for demographic risk factors and treatment-related risk factors.

^cAccounts for childhood cancer type.

^dIncludes primary treatment and all recurrences; radiotherapy (yes/no) were missing for 30 survivors; chemotherapy (yes/no), for 34; and hematopoietic cell transplant (yes/no), for 68. Also includes 371 survivors who had hematopoietic cell transplant and 58 cases with benign tumors, of whom 210 survivors had total body irradiation, including 41 cases with benign tumors.

^eProportional hazards assumption was violated for RT without CT, meaning that the HR was not constant over time. At the age of 30.0 years (median age of this subcohort that received RT without CT), the HR was 3.3 (95% CI, 1.5-3.5). The HR was higher at ages below 30.0 years and lower at ages above 30.0 years (HR, 3.3 × 0.95 [years from age 30.0]). The magnitude of change is denoted by the factor 0.95.

^fDiagnostic groups included all malignant neoplasms covered by International Classification of Childhood Cancer, Third Edition (ICCC-3), as well as multifocal Langerhans cell histiocytosis and selected nonmalignant ependymomas and astrocytomas.

^gIncludes 1667 survivors with acute lymphoblastic leukemia (126 cases with benign tumors), 187 with acute myelogenous leukemia (23 cases with benign tumors), and 108 with other leukemia (8 cases with benign tumors).

^hIncludes all morphology codes specified in the ICCC-3 under lymphomas and reticuloendothelial neoplasms, except for Hodgkin lymphomas.

ⁱIncludes all morphology codes specified in the ICCC-3 under other malignant epithelial neoplasms and malignant melanomas and other and unspecified malignant neoplasms as well as nonmalignant multifocal Langerhans cell histiocytosis.

Demographic and Treatment-Related Risk Factors for Any Benign Tumor

Risk factors for the first benign tumor of any histologic type or location per individual (further referred to as any benign tumor) were evaluated in multivariable Cox proportional hazards regression models (Table 1) (model 1). Female CCSs had a significantly higher risk than male CCSs to develop any benign tumor (HR, 1.5; 95% CI, 1.3-1.8). No sex difference was observed when benign tumors of sex-specific organs were excluded. Children diagnosed with cancer before 5 years of age had a significantly increased risk of any benign tumor compared with children diagnosed at 10 to 17 years of age (HR, 1.5; 95% CI, 1.2-1.8). Of note, younger age at diagnosis equals longer follow-up time in models with attained age as time scale. Furthermore, a diagnosis of childhood cancer after 1985 inferred a 1.4-fold risk for benign tumors compared with those diagnosed before 1985 (HR for 1985-1994, 1.4 [95% CI, 1.1-1.8]; HR for 1995-2001, 1.4 [95% CI, 1.0-1.9]). A history of radiotherapy only (HR for 30 years of age, 2.3; 95% CI, 1.5-3.5 [proportional hazards assumption violated]) and combined radiotherapy and chemotherapy (HR, 2.5; 95% CI, 1.7-3.5) significantly increased the risk for benign tumor compared with surgery only or no recorded treatment. Relative to survivors of pediatric Hodgkin lymphoma (reference group), only survivors of medulloblastoma (HR, 2.0; 95% CI, 1.2-3.3) were at statistically significantly increased risk for benign tumors (Table 1) (model 2).

Cumulative Incidence

The cumulative incidence of BTs by age 30 years was highest after radiotherapy only (16.7%; 95% CI, 9.0%-26.3%) and after radiotherapy plus chemotherapy (11.2%; 95% CI, 8.1%-14.7%) with values of 4% to 6% for those treated otherwise and for siblings (Figure, A). By childhood cancer type, the highest cumulative incidences at 30 years of age were seen after medulloblastoma (11.5%; 95% CI, 6.5%-17.9%), neuroblastoma (10.5%; 95% CI, 7.0%-14.7%), and renal tumors (7.6%; 95% CI, 5.2%-10.5%) (eFigure 1 in the Supplement). Cumulative incidences at 30 years of age varied slightly by period of childhood cancer diagnosis, but did not seem to decrease for most recently treated patients, with cumulative incidences of 7.3% (95% CI, 5.8%-9.1%) for a childhood cancer diagnosis from 1963 to 1984; 8.4% (95% CI, 6.8%-10.2%), 1985 to 1994; and 7.5% (95% CI, 5.7%-9.5%), 1995 to 2001 (Figure, B). eFigure 2 in the Supplement also provides results using time since diagnosis as time scale. The cumulative incidences at 45 years of age for benign tumors (female CCSs, 24.7% [95% CI, 21.6%-27.9%]; male CCSs, 15.1% [95% CI, 13.1%-17.3%]) (eFigure 3 in the Supplement) were much higher than those for nonskin solid subsequent malignant neoplasms in the comparable subsample of a previous DCOG-LATER study of subsequent malignant neoplasms¹ (females, 10.7% [95% CI, 8.4%-13.2%]; males, 5.7% [95% CI, 4.3%-5.7%]).

Figure. Cumulative Incidence of Benign Tumors for 5-Year Survivors of Childhood Cancer by Attained Age.

Participants are stratified by treatment (A) (including siblings) and period of diagnosis (B). Both analyses account for death as a competing risk. CT indicates chemotherapy; RT, radiotherapy.

The most common benign tumors in CCSs were meningiomas (89 [16.4%]), and among siblings, fibroepithelial neoplasms (12 [21.4%]) (eTable 1 in the Supplement). In addition, nerve sheath tumors (50 [9.2%]) and osseous and/or chondromatous neoplasms (23 [4.2%]) were more frequently diagnosed in CCSs compared with siblings (1 [1.8%] for each tumor type).

Multivariable Risk Factor Analyses for Primary Outcome Groups

Among female CCSs, 50 myomatous neoplasms were diagnosed, including 43 uterine leiomyomas (eTable 2 in the Supplement). The most important risk factor for leiomyomas was abdominopelvic radiotherapy (including total body irradiation [TBI]), with a significant dose-dependent increase compared with female CCSs without such radiotherapy (HR for dose <20 Gy, 1.9 [95% CI, 0.5-7.0]; HR for dose of 20 to <30 Gy, 3.4 [95% CI, 1.1-10.4]; and HR for dose ≥30 Gy, 5.4 [95% CI, 2.4-12.4]) (Table 2); the test for trend reached significance (P = .002), however, not among exposed-only participants (P = .78). None of the chemotherapy groups was associated with uterine leiomyoma risk.

Table 2. Multivariable Cox Regression Analyses for Risk of Selected Benign Tumor Types Among 5-Year Survivors of Childhood Cancer in the DCOG-LATER Cohort^a.

Characteristic	No. of Participants		Uterine Leiomyoma (n = 43)		Breast Fibroadenoma (n = 45)		Nerve Sheath Tumor (n = 55)b,c		Osseous and/or Chondromatous Neoplasm (n = 23)d
	All	Women	No. of Cases	HR (95% CI)	No. of Cases	HR (95% CI)	No. of Cases	HR (95% CI)	No. of Cases	HR (95% CI)
Sex
Male	3269	NA	NA	NA	NA	NA	30	1 [Reference]	15	1 [Reference]
Female	2574	2574	43	NA	45	NA	25	1.0 (0.6-1.7)	8	0.7 (0.3-1.8)
Age childhood cancer diagnosis, y
0-4	2641	1216	11	1.0 (0.4-2.3)	16	0.7 (0.3-1.5)	28	1.7 (0.8-3.8)	14	2.3 (0.6-8.5)
5-9	1583	647	11	1.2 (0.6-2.6)	10	0.7 (0.3-1.5)	12	0.8 (0.3-1.7)	5	1.0 (0.3-4.2)
10-17	1619	711	11	1 [Reference]	19	1 [Reference]	15	1 [Reference]	4	1 [Reference]
Period of childhood cancer diagnosis
1963-1984	1791	837	34	1 [Reference]	25	1 [Reference]	22	1 [Reference]	7	1 [Reference]
1985-1994	2113	911	8	0.7 (0.3-1.7)	11	0.5 (0.2-1.2)	22	1.4 (0.7-3.0)	9	0.9 (0.3-2.8)
1995-2001	1939	826	1	0.4 (0.1-3.6)	9	0.7 (0.3-1.9)	11	1.5 (0.6-4.0)	7	1.1 (0.3-4.2)
Chemotherapy
No	1050	NA	NA	NA	NA	NA	NA	NA	2	1 [Reference]
Yes	4759	NA	NA	NA	NA	NA	NA	NA	20	0.9 (0.2-4.2)
Alkylating agents
No	2820	1336	20	1 [Reference]	15	1 [Reference]	35	1 [Reference]	NA	NA
Yes	2987	1224	23	1.6 (0.7-3.4)	29	2.1 (1.0-4.6)	19	0.5 (0.2-1.1)	NA	NA
Anthracyclines
No	3165	1438	22	1 [Reference]	21	1 [Reference]	38	1 [Reference]	NA	NA
Yes	2642	1123	21	1.9 (0.9-4.1)	23	1.3 (0.6-2.6)	16	1.1 (0.5-2.4)	NA	NA
Epipodophyllotoxin
No	4573	2071	37	1 [Reference]	35	1 [Reference]	38	1 [Reference]	NA	NA
Yes	1232	488	6	1.4 (0.5-4.1)	9	0.9 (0.4-2.4)	16	3.8 (1.5-9.2)	NA	NA
Vinca alkaloids
No	1599	748	14	1 [Reference]	10	1 [Reference]	25	1 [Reference]	NA	NA
Yes	4210	1813	29	0.9 (0.9-2.0	34	1.1 (0.5-2.6)	29	0.7 (0.4-1.5)	NA	NA
Platinum agents
No	5035	2190	39	1 [Reference]	37	1 [Reference]	45	1 [Reference]	NA	NA
Yes	770	369	4	1.7 (0.5-5.6)	7	1.7 (0.7-4.6)	9	0.7 (0.3-2.0)	NA	NA
Antimetabolites
No	3087	1434	26	1 [Reference]	28	1 [Reference]	38	1 [Reference]	NA	NA
Yes	2721	1127	17	1.1 (0.5-2.4)	16	0.7 (0.3-1.4)	16	1.2 (0.5-2.9)	NA	NA
Abdominopelvic RT dose, Gye
None	NA	2275	27	1 [Reference]	NA	NA	NA	NA	NA	NA
<20	NA	120	3	1.9 (0.5-7.0)	NA	NA	NA	NA	NA	NA
≥20 to <30	NA	65	4	3.4 (1.1-10.4)	NA	NA	NA	NA	NA	NA
≥30	NA	92	9	5.4 (2.4-12.4)	NA	NA	NA	NA	NA	NA
Trunk RT dose, Gyf
None	NA	2159	NA	NA	26	1 [Reference]	NA	NA	NA	NA
<20	NA	142	NA	NA	12	3.8 (1.8-8.1)	NA	NA	NA	NA
≥20 to <30	NA	88	NA	NA	4	2.7 (0.9-7.9)	NA	NA	NA	NA
≥30	NA	163	NA	NA	2	1.0 (0.2-4.5)	NA	NA	NA	NA
RT
No	3505	NA	NA	NA	NA	NA	16	1 [Reference]	NA	NA
Yes	2308	NA	NA	NA	NA	NA	38	NAg	NA	NA
RT
No	3690	NA	NA	NA	NA	NA	NA	NA	16	1 [Reference]
Yes without TBI	2123	NA	NA	NA	NA	NA	NA	NA	6	1.0 (0.4-3.0)
TBI
No	5590	NA	NA	NA	NA	NA	NA	NA	11	1 [Reference]
Yes	210	NA	NA	NA	NA	NA	NA	NA	11	37.4 (14.8-94.7)
NF1/NF2 probability
No	3894	NA	NA	NA	NA	NA	20	1 [Reference]	NA	NA
Yes, CNS tumors	759	NA	NA	NA	NA	NA	21	5.6 (2.3-13.7)	NA	NA
Yes, other childhood cancer types	1190	NA	NA	NA	NA	NA	14	2.4 (0.9-6.2)	NA	NA

Abbreviations: CNS, central nervous system; DCOG-LATER, Dutch Childhood Oncology Group–Long-Term Effects After Childhood Cancer; HR, hazard ratio; NA, not applicable; NF1/NF2, neurofibromatosis type 1 or 2; RT, radiotherapy; TBI, total body irradiation.

^aModels include only 44 female survivors with a breast fibroadenoma, 54 survivors with a nerve sheath tumor, and 22 survivors with osseous and/or chondromatous neoplasms owing to missing values. Numbers in each subgroup are larger than those cited in eTable 1 in the Supplement for the first benign tumor per person, because for specific outcome analyses any first benign tumor of a particular type is counted rather than any first benign tumor number cited in eTable 1.

^bIncludes 17 tumors in the head and neck (30.9%), 19 in the trunk (34.5%), 13 in the extremities (23.6%), and 6 unknown location (10.9%).

^cFor HR interaction of RT and NF1/NF2 indicator, unlikely NF1/NF2 with no RT (HR, 0.2; 95% CI, 0.1-0.5), unlikely NF1/NF2 with RT (1 [reference]), likely NF1/NF2 with no RT (HR, 1.3; 95% CI, 0.5-3.6), and likely NF1/NF2 with RT (HR, 2.8; 95% CI, 1.2-6.8).

^dIncludes 1 tumor in the head and neck (4.3%), 3 in the trunk (13.0%), 18 in the extremities (78.3%; 15 in the lower extremities [65.2%]), and 1 unknown location (4.3%).

^eIncludes TBI dose (n = 3). Spinal RT (n = 4) was not included.

^fIncludes TBI dose (n = 1) and 50% of abdominopelvic RT dose. Spinal RT was not included (n = 3).

^gProportional hazards assumption was violated for RT, meaning that the HR was not constant over time. At the age of 30.7 years (median age of this subcohort that received RT), the HR was 2.9 (95% CI, 1.5-5.5). The HR was higher at ages below 30.7 years and lower at ages above 30.7 years (HR, 2.9 × 0.94 [years from age 30.7]). The magnitude of change is denoted by the factor 0.94.

In the subgroup of fibroepithelial neoplasms (eTable 2 in the Supplement), 46 fibroadenomas of the breast were diagnosed (1 male and 45 female CCSs; age <50 years). Women treated with a radiotherapy dose of less than 20 Gy to the trunk (12 cases) had a significantly increased risk of breast fibroadenoma compared with women without such exposure (26 cases; HR, 3.8; 95% CI, 1.8-8.1), whereas those treated with a cumulative dose of 20 to less than 30 Gy (4 cases; HR, 2.7; 95% CI, 0.9-7.9) and at least 30 Gy (2 cases; HR, 1.0; 95% CI, 0.2-4.5) did not (P = .07 for trend among all survivors and P = .28 for trend among those with exposure only) (Table 2). In addition, female survivors who had received alkylating agents (29 cases of 1224 female cohort members) had a statistically significantly increased risk (HR, 2.1; 95% CI, 1.0-4.6; P = .048).

Nerve sheath tumors (n = 55) occurred across the body, including 29 schwannomas and 24 neurofibromas (eTable 2 in the Supplement). Six schwannomas concerned acoustic neuromas, 3 of which occurred after cranial radiotherapy for astrocytoma. In multivariable analyses, risk of nerve sheath tumors was elevated for the following 3 factors: (1) radiotherapy, the effect of which slightly decreased with increasing attained age (HR at 31 years of age, 2.9; 95% CI, 1.5-5.5 [proportional hazards assumption violated]); (2) an indicator variable for NF1/NF2–associated childhood cancer types (a definition is found in eTable 3 in the Supplement), with HRs of 5.6 (95% CI, 2.3-13.7) for central nervous system tumors and 2.4 (95% CI, 0.9-6.2) for other childhood cancer types among subgroups at increased NF1/NF2 probability vs all other CCSs; and (3) epipodophyllotoxins (HR, 3.8; 95% CI, 1.5-9.2) (Table 2). A significant effect of epipodophyllotoxins remained after adjustment for cranial radiotherapy and TBI. Because NF1/NF2 status may affect the association of prior radiotherapy on nerve sheath tumor, we also conducted analyses of interaction. No evidence in our data suggested that the effect of radiotherapy was stronger according to our crude surrogate indicator for NF1/NF2 status (Table 2).

Twenty-three subsequent osseous and/or chondromatous neoplasms occurred, including 21 osteochondromas (eTable 2 in the Supplement). These tumors occurred most often (16 of 23) among leukemia survivors, many of whom (11 of 16) had received hematopoietic cell transplant, all including TBI (HR, 37.4; 95% CI, 14.8-94.7) (Table 2). We did not find an effect of other radiotherapy or for 6 chemotherapy groups.

Multivariable Risk Factor Analyses for Secondary Outcome Groups

Multivariable Cox proportional hazards regression models for secondary outcomes are provided in eTable 4 in the Supplement. Patients treated with chemotherapy and radiotherapy had elevated risk for adenomas (HR, 2.8; 95% CI, 1.1-7.1) and for fibromatous neoplasms (HR, 4.4; 95% CI, 1.7-11.0), compared with those who had surgery only or no recorded treatment.

Discussion

In our record linkage study with high-quality treatment and outcome assessments, we observed the highest cumulative incidence of benign tumors by 30 years of age for CCSs treated with radiotherapy only (16.7%; 95% CI, 9.0%-26.3%). Notable findings of our exploratory risk factor analyses include the role of abdominopelvic radiotherapy in uterine leiomyoma, elevated risk for osteochondromas among survivors of leukemia, and strong effects of radiotherapy and a crude indicator of NF1/NF2 status for the risk of nerve sheath tumors. Moreover, we provide a comparison of risk for benign tumors in CCSs and a sibling group to provide some insight into background risk, because age-specific reference rates are not available for any of the reported outcomes. We observed that CCSs have an approximately 2-fold risk to develop subsequent benign tumors compared with siblings.

Our study shows that female survivors have a higher risk than male survivors, a difference attributable to sex-specific benign breast and uterine tumors. Moreover, the cumulative incidence at 45 years of age was more than twice as high for solid benign tumors than for nonskin solid malignant neoplasms, possibly reflecting differences in background risks of benign and malignant tumors. Although biopsies for suspected subsequent malignant neoplasms can contribute to excess risk for benign tumors, and 11.6% of patients with benign tumors also experienced a subsequent malignant neoplasm, only 0.7% of all benign tumors concerned synchronous benign tumor and subsequent malignant neoplasm diagnoses in a single organ. In all, survivors experienced higher rates of benign tumors than the sibling group.

We found evidence of a radiotherapy-related increased risk of uterine leiomyomas among women who had abdominopelvic radiotherapy (minimal dose, 2 Gy). Although the strongest dose-response test to support causality, the P value for trend among exposed patients was not significant, and this finding is consistent with a significant dose-response effect (odds ratio at 1 Gy, 1.61; 95% CI, 1.12-2.31) among 1190 female atomic bomb survivors who underwent ultrasonographic examination (238 cases).²⁵ This exploratory finding warrants confirmation in other study populations, including studies using more specific measures of uterine radiation exposure (volume-based reconstructed doses).

Of all fibroadenomas, more than 50% were diagnosed among women without chest-directed radiotherapy, that is, in the large group of women not undergoing survivorship mammography or magnetic resonance imaging screening; all were diagnosed before the starting age of population-based screening eligibility (50 years).^26,27 Two cases were synchronous with a diagnosis of breast cancer; 2 other fibroadenomas occurred after a breast cancer diagnosis, likely detected on routine imaging in breast cancer aftercare. There was no monotonic association of fibroadenoma risk with increasing dose of radiotherapy to the trunk. In summary, we did not find evidence of an increased risk of breast fibroadenoma nor of a strong role of prior childhood cancer therapy.

Adenomas are among the most prevalent benign tumors in our cohort and are of interest as potential cancer precursors. Teepen et al¹⁷ previously reported a positive association between abdominopelvic radiotherapy and colorectal adenoma. Other site-specific analyses^13,14,15 were hampered by small numbers of cases; several known radiosensitive tissues were represented (eg, salivary, thyroid, and parathyroid glands). In line with a recent US report among 1296 CCSs (6 cases [0.5%]),²⁸ we identified 4 (0.1%) liver adenomas.

Nerve sheath tumors are known to be associated with predisposition syndromes and prior radiation exposure.^20,29 Although we confirmed strong effects of radiotherapy and a crude indicator of NF1/NF2 status, exploratory analyses showed no evidence that the effect of radiotherapy was stronger among NF1/NF2–associated childhood cancer types. As reported previously,²³ meningioma represented the most common benign tumor (16.4%) primarily after cranial radiotherapy for leukemia and central nervous system tumors. Medulloblastoma survivors had high-dose, full-volume cranial radiotherapy, which offers a likely explanation for the elevated overall risk for benign tumors among CCSs with medulloblastoma vs Hodgkin lymphoma. In all, nervous system tumors accounted for a quarter of benign tumors.

Osteochondromas have been reported in other studies, in particular after treatment involving radiotherapy for leukemia and neuroblastoma.^30,31,32,33 Herein we report on 21 cases, of whom 11 had received hematopoietic cell transplant, all including TBI. There appeared to be no association with other radiotherapy. Although the relevance of the TBI finding is unclear, it may be an indicator of other hematopoietic cell transplant–related exposures.

Strengths and Limitations

To our knowledge, this study is the first comprehensive evaluation of risk for benign tumors among long-term CCSs. Strengths include the availability of cohort-based detailed individual treatment information and objective and uniform data on histologically confirmed benign tumors from PALGA²¹; we collected data on benign tumors for more than 95% of the study population. Most benign tumors are not covered in cancer registries, so existing studies rely on patient self-report with medical verification or on medical record review. Another strength of this record linkage study is the comparison with a sibling group of CCSs. We hypothesize that siblings represent appropriate controls because general health awareness among siblings and CCSs may be more alike than for population-based comparison groups vs CCSs. There is some potential for survival bias and selection bias of the sibling sample; unlike the survivor cohort, siblings were recruited in 2013 to 2014, and inclusion depended on questionnaire participation (ie, siblings had to be alive). We expect small levels of bias because of low age-specific mortality rates and in view of prior work from the DCOG-LATER-VEVO (Fertility in Female Childhood Cancer Survivors) study group.³⁴

A possible limitation of research on indolent benign tumors in CCSs is that survivors receive medical attention in survivorship care programs, and physicians may exert a lower threshold to order imaging for suspect lesions.^26,33,35 However, most benign tumors identified in this study were not detected in screening programs, but rather, by symptoms. Cohort members receive guideline-based care without imaging-based screening, except for women at elevated risk of breast cancer who received mammography (and sometimes magnetic resonance imaging) from 2010 onwards (<10% of cohort).²⁶ Also, we had no information on individual genetic predisposition. Finally, benign tumors diagnosed before 1990 were not captured by PALGA; by including only follow-up time from 1990, we have limited the underestimation of benign tumors owing to left truncation. Also, a slight chance of false-positive findings exists with the PALGA linkage on family name, sex, and birth date due to administrative twins. In addition, PALGA excerpts contain information on consecutive biopsies; sometimes the excerpt texts were not sufficiently specific to distinguish specimens from one lesion from specimens taken from different lesions or new lesions in the same organ. We therefore focused in the analyses on the first diagnosed benign tumor per outcome group. We had no information on benign tumors detected on imaging without histologic confirmation.

For follow-up of survivors of childhood cancer, it will be relevant to be aware of clinical symptoms related to uterine leiomyoma among women with a history of abdominopelvic radiotherapy, of osteochondroma among leukemia survivors, and of meningioma and nerve sheath tumors after (cranial) radiotherapy. Timely attention for these symptoms is relevant for early diagnosis and can help to diminish burden for survivors. In addition, early detection of benign tumors as precursors of subsequent malignant neoplasms (eg, colorectal adenoma, thyroid nodules) might decrease morbidity and mortality.^8,36,37 Of interest are efforts of the International Guideline Harmonization Group to define or update surveillance recommendations for late effects (http://www.ighg.org).^8,27 In this group, the benefits and harms of active surveillance of benign tumors should be discussed. For future research, investigation of the roles of prior therapy and genetic susceptibility among CCSs should be evaluated combining organ-specific volume-based reconstructed doses and results of genotyping.

Conclusions

In conclusion, our record linkage study shows a considerable cumulative incidence of solid benign tumors in CCSs, which is higher than for subsequent solid malignant neoplasms, and higher compared with siblings. We also identified risk factors for benign tumors. These findings are important to enable early diagnosis and to serve as the first step for future surveillance guidelines, including some benign tumors in CCSs. Also, they provide leads for further in-depth etiologic studies.

Supplement.

eMethods. Statistical Analysis and Radiotherapy Assessment

eTable 1. Patient and Benign Tumor Characteristics of the DCOG-LATER and Sibling Cohorts

eTable 2. Overview of Benign Tumor Types According to ICD-O–Defined Morphology Subgroups

eTable 3. Childhood Cancer Types Associated With NF1 and NF2, Based on Literature

eTable 4. Multivariable Cox Regression Analyses for Risk of Adenoma, Lipomatous and Fibromatous Neoplasm, and Blood Vessel Tumor in 5-Year Survivors of Childhood Cancer in the DCOG-LATER Cohort

eFigure 1. Cumulative Incidence of Benign Tumors for 5-Year Survivors of Childhood Cancer (CCSs) by Attained Age and Time Since Childhood Cancer Diagnosis According to Childhood Cancer Type, Accounting for Death as Competing Risk

eFigure 2. Cumulative Incidence of Benign Tumors for 5-Year Survivors of Childhood Cancer (CCSs) by Time Since Childhood Cancer Diagnosis, Accounting for Death as Competing Risk

eFigure 3. Cumulative Incidence of Benign Tumors (BTs) and Subsequent Malignant Neoplasms (SMNs) for Female and Male 5-Year Survivors of Childhood Cancer (CCSs) by Attained Age, Accounting for Death as a Competing Risk

eReference.