Lung Cancer as a Subsequent Malignant Neoplasm in Survivors of Childhood Cancer

This article is featured in Highlights of This Issue, p. 2141

With five-year survival following a childhood cancer diagnosis exceeding 85% and nearly 500,000 pediatric cancer survivors now alive in the United States (1, 2), the identification of late effects after cancer therapy, including subsequent malignant neoplasms (SMN; ref. 3) and chronic medical conditions (4), is crucial to providing continued quality care to long-term survivors (5). SMNs account for the greatest proportion of deaths in the survivor population after primary cancer recurrence (6, 7).

Lung cancer is the leading cause of cancer mortality in both adult men and women in the general population (8). Previous reports from large cohorts of survivors of childhood cancer have described limited numbers of subsequent lung cancers (9–12). Risk for subsequent lung cancer has been reported to be greatest among patients with a primary diagnosis of Hodgkin lymphoma (13–16). Much of what is understood about subsequent lung cancer comes from adult survivors of Hodgkin lymphoma, and risk factors include therapeutic chest radiation and a history of smoking (16–19).

Lung cancer screening has been debated in high-risk groups and the use of low-dose helical CT (LDCT) has been shown to reduce mortality in selected, high-risk populations (20, 21). The current recommendations from the American Cancer Society have specific guidelines for lung cancer screening that have been adopted by some to guide screening in pediatric cancer survivors, though the benefit of LDCT is still debated (8, 22). This study seeks to provide additional information about the risk factors, characteristics, and outcomes of subsequent lung cancers among survivors of childhood cancer to further develop the rationale for screening for subsequent lung cancer.

The Childhood Cancer Survivor Study (CCSS) is a retrospective cohort study with longitudinal follow-up of 25,654 five-year survivors of childhood cancer diagnosed at one of 31 participating institutions in the United States and Canada between 1970 and 1999. Participants were less than 21 years of age when initially diagnosed with cancer (leukemia, Hodgkin lymphoma, non-Hodgkin lymphoma, central nervous system cancer, renal tumor, neuroblastoma, rhabdomyosarcoma, or bone cancer). Human subjects committees at participating institutions approved all protocol and contact documents. Informed consent was obtained from participants or their parents if they were minors. Minor (<18 years of age) participants were re-consented at 18 years of age. The study was conducted in accordance with the Declaration of Helsinki. The CCSS methodology has been previously described (23, 24).

Cases of lung cancer occurring five or more years after childhood cancer diagnosis were ascertained by self-report and confirmed by pathology report review, or if unavailable, medical record and/or death certificate review. Other pulmonary outcomes were similarly assessed by self-report. Neoplasms considered eligible for inclusion had International Classification of Diseases of Oncology (ICD-O-3) site codes of C33.9, C34, C38.4, C38.8, C39.8, and C39.9. All lung neoplasms with an ICD-O-3 histology 5^th digit code of 3 (malignant) were included. Demographic characteristics and cancer therapies received in the five years following diagnosis, including surgery, chemotherapy, and radiation, were ascertained through medical record abstraction, as described (24, 25). Specifically, chemotherapy exposures, including cumulative doses of anthracyclines (doxorubicin equivalents; refs. 26, 27), platinating agents (28, 29), and alkylating agents (expressed as cyclophosphamide equivalent dose; ref. 30) were abstracted. Radiation exposure was evaluated by body site, field size, and by maximum target dose to the chest (the sum of the target doses from all overlapping chest directed fields). For field size, exposures were specified as none, small, or large. Large chest field sizes were identified as those which covered all or at least half of the chest and generally did not have lung blocking. The following were considered to be large chest fields for the purpose of this study: whole chest, other chest large, other chest large right, other chest large left, and total body irradiation (TBI). All other chest radiation fields were considered small. Smoking status was also ascertained through self-report from baseline and follow-up questionnaires of participants ≥ 18 years of age at time of questionnaires, minor participants were not asked about smoking status (31). Participants were considered smokers if they had ever smoked > 100 cigarettes in their lifetime and if documented prior to diagnosis of lung cancer. All others were presumed to be “never.”

Descriptive statistical analysis was performed regarding the distributions of age at diagnosis, cancer stage at diagnosis, and histology. The cumulative incidence of lung SMNs was estimated from five years after diagnosis to the first occurrence of lung cancer, treating death as a competing risk and censoring at the date of last contact. Standardized incidence ratios (SIRs) were calculated using age-, sex-, race-, and calendar-specific incidence rates of lung cancer from the Surveillance, Epidemiology and End Results (SEER) database (2). Lung cancer incidence data of the SEER population (not limited to individuals aged greater than 21 years) was downloaded using SEER*Stat, including the number of malignant cases and the population size in each of the strata defined jointly by age, sex, race, and calendar year. Indirect standardization method was used to compute the expected count of malignant lung cancer in the CCSS cohort in each stratum using the age-sex-race-calendar year-specific rates of the SEER population as reference. The SEER data covered from 1975–2016: for the 2017 CCSS data, SEER 2016 data was used. Nine regions of US population were selected, San Francisco-Oakland SMSA, Connecticut, Detroit (Metropolitan), Hawaii, Iowa, New Mexico, Seattle (Puget Sound), Utah, Atlanta (Metropolitan), from the SEER data to match with CCSS data sets. The data set used was SEER Research Data, 9 Registries, Nov 2018 Sub (1975–2016). We compared CCSS survivors' first lung cancer (SMN) versus the SEER population's first primary lung cancer. The stratum specific expected counts were summed across the strata to compute the total expected count and subsequently SIR by dividing the total observed count in the CCSS by the total expected count.

Cox proportional hazards models with age as the time scale were used to assess associations between survivor and treatment characteristics and the risk of subsequent lung cancer. In a treatment model, chest radiation dose was forced in the model and other characteristics were selected by backward selection. The criteria used for the backward selection removed covariates with the largest P value successively until all the covariates in the model had a P value less than or equal to 0.05. Cancer diagnosis was not included in this model due to concern for collinearity between diagnosis and subsequent treatment. Trend test was conducted by assigning sequential integers to the five chest radiotherapy dose groups of no chest radiotherapy, >0–10 Gy; >10–30 Gy, >30–40 Gy, and >40 Gy, and tested the statistical significance of this integer-valued continuous variable for linear trend. Individual chemotherapeutic agents were also evaluated adjusting for sex and race, independent of radiation given the overwhelming effect of radiation when included in the model. All statistical tests were two-sided and P values less than 0.05 were considered as statistically significant.

Among the 25,654 survivors included in the study, 42 (0.16%) were diagnosed with a subsequent malignant lung neoplasm. Demographic and treatment characteristics of survivors who developed lung SMN versus those who did not were assessed (Table 1). The 42 identified lung SMNs included 26 carcinomas, 6 mesotheliomas, and 10 others (Table 2). Two benign neoplasms were also identified (neurilemoma and solitary fibrous tumor of the lung) but not counted as a lung SMN in this analysis. Mean age of survivors at time of lung cancer diagnosis was 45 years (SD, 10.3) and mean duration from primary cancer diagnosis to lung cancer was 29 years (SD, 8.7). The most common primary diagnoses of survivors who developed lung SMN were Hodgkin lymphoma (66.7%) and bone cancer (16.7%). Survivors who developed a lung SMN were older at last follow up (45 years vs. 33 years, P < 0.001) and had longer duration of follow-up (29 years vs. 24 years, P < 0.001). No difference was seen in the distribution of sex or race between groups. Survivors with a lung SMN were more likely to have previous bleomycin exposure (14.6% vs. 5.6%, P = 0.01). Survivors with lung cancer were more likely to report a positive family history of cancer (42.9% vs. 21.5%, P < 0.001), history of smoking (50% vs. 30.4%, P = 0.006), and other pulmonary dysfunction (35.7% vs. 17.5%, P = 0.002), specifically reporting pulmonary fibrosis (14.3% vs. 3.4%, P < 0.001).

Chest radiation dose and radiation field size differed significantly among survivors who developed lung cancer compared with those who did not (P < 0.001). Five of the 42 (11.9%) individuals with a lung SMN received only stray radiation (scatter and leakage) to the chest and were included in the 0–10 Gy group; five (11.9%) survivors had no direct chest radiation exposure. All five of these survivors had a primary diagnosis of bone cancer (predominantly osteosarcoma). Two were male and three were female, age at primary diagnosis ranged between 8–15 years, and age at time of lung cancer diagnosis ranged between 35–49 years. One had a positive family history of cancer (brother with an unspecified neoplasm of the tonsil) and only one had a positive smoking history.

The 30-year cumulative incidence of lung SMN was 0.16% [95% confidence interval (CI), 0.09%–0.23%] as seen in the cumulative incidence plot in Fig. 1A. When cumulative incidence was stratified by chest radiation dose, it was observed that greater cumulative incidence of lung SMN was observed in those who were exposed to greater chest radiation doses (Fig. 1B). The cumulative incidence curve was also stratified on the basis of smoking history and demonstrated that cumulative incidence of lung SMN was higher in those who were current or past smokers compared with those who had never smoked (Fig. 1C).

Survivors who developed subsequent lung cancer were more likely to have died (n = 28) at the time of last follow-up than survivors without lung SMN (66.7% vs. 15.6%). Among survivors who developed subsequent lung cancer and were deceased at the time of last follow-up, 86.2% (n = 24) had cause of death reported as lung SMN, 7.1% (n = 2) had cause of death as a result of heart failure, and 3.6% (n = 1) had cause of death as a result of kidney failure. One cause of death was unknown. The median survival time (any cause of death, n = 28) was 3.76 years, (95% CI, 1.57–4.93). For those who had a reported cause of death as subsequent neoplasm (n = 24), the mean interval between diagnosis of lung SMN and death was 2.68 years (SD, 4.40 years). To compare the survival after lung cancer in this cohort to SEER, a Cox regression model adjusting for age at diagnosis of lung cancer was used and found a rate ratio of mortality following lung cancer in CCSS versus SEER of 0.70 (95% CI, 0.48–1.04).

Survivors had a four-fold greater risk for developing lung cancer compared with the general population (SIR, 4.0; 95% CI, 2.9–5.4) and the absolute excess risk (AER) was 0.1 per 1000 person-years (95% CI, 0.04–0.08; Table 3). When evaluating distribution of cancer histology compared with the general population as reflected in SEER, survivors had a similar risk for developing small cell and squamous cell carcinoma, a four-fold greater risk of developing adenocarcinoma (SIR, 4.2; 95% CI, 2.4–6.8), and a 12-fold greater risk of developing other carcinomas (SIR, 12.5; 95% CI, 5.4–24.6; Table 2). In comparison with the general population, survivors had a nearly 60-fold greater risk for developing mesothelioma (SIR, 59.4; 95% CI, 21.7–129.3; Table 2). Table 3 shows additional unadjusted SIRs, AERs, and cumulative incidences based on demographic, diagnostic, and treatment characteristics. Of specific interest was the SIRs associated with the primary diagnoses of Hodgkin lymphoma (SIR, 9.3; 95% CI, 6.2–13.4) and bone cancer (SIR, 4.4; 95% CI, 1.8–9.1). Absolute excess risk was greatest for those with an attained age of 60 years or older (22.6/1,000 person-years, 95% CI, 3.52–68.65). SIRs of lung cancer were also calculated stratified by histologic classification of lung cancer (Table 2) and demonstrated increased risk for development of adenocarcinoma (SIR, 4.2; 95% CI, 2.4–6.8), other carcinomas (SIR, 12.5; 95% CI, 5.4–24.6), and mesothelioma (SIR, 59.5; 95% CI, 21.7–129.3).

To assess risk for lung cancer specifically after radiation exposure, a treatment model was assessed, forcing radiation dose in the model with backward selection for sex, age at diagnosis, chemotherapy exposures and smoking history (Table 4). Chest radiation field size was not included in the model due to its collinearity with the radiation dose. Only chest radiation and older age at diagnosis of initial childhood cancer were associated with subsequent lung SMN risk, specifically those with chest doses of 10–30 Gy (HR, 3.4; 95% CI, 1.05–11.0), 30–40 Gy (HR, 4.6; 95% CI, 1.5–14.3), and >40 Gy (HR, 9.1; 95% CI, 3.1–27.0): P_trend < 0.001. Chemotherapy exposures and smoking history were not found to be independently associated with increased risk for lung SMN when adjusted for radiation dose and age at diagnosis (Supplementary Table S1A and S1B).

When exposure to chemotherapy agents were evaluated individually adjusting for sex and race and without adjusting for chest radiotherapy (Table 5), a significant association was observed between exposure to bleomycin and subsequent lung SMN (HR, 4.7; 95% CI, 1.9–11.6). An association (HR, 2.3; 95% CI, 0.9–5.9) between bleomycin exposure and subsequent lung SMN was also observed when adjusted for chest radiotherapy and age at diagnosis though it was not statistically significant (Table 5). A similar association (HR, 7.0; 95% CI, 0.8–62.4) was observed between high cumulative doses of platinating agents (>750 mg/m²) and subsequent lung SMN when adjusted for chest radiotherapy and age at diagnosis, but again was not statistically significant (Table 5). An association between exposure to alkylators and subsequent lung SMN was not observed (Table 5).

To assess the risk of lung SMN based on smoking history, the HR for positive smoking history compared to never smoking, adjusting for sex and race, was calculated and found to be not significant (HR, 1.8; 95% CI, 0.96–3.3), similar to what was observed in the treatment model (Table 4). However, when the five patients with bone tumors who developed lung cancer without radiation exposure were removed from the analysis, a significant association was observed between smoking history and lung SMN when adjusting for radiation dose alone (HR, 2.4; 95% CI, 1.2–4.6) and both radiation dose and age at diagnosis (HR, 2.1; 95% CI, 1.1–4.1).

We investigated the risk of subsequent lung cancer in childhood cancer survivors and report, to our knowledge, the largest number of observed lung cancers to date among this population. Other studies of childhood cancer survivors have described subsequent lung cancers (10–14); however, the numbers of lung cancer cases observed have been small (N = 4–20) and two of the four previous publications focused solely on survivors of Hodgkin lymphoma. Within the CCSS cohort, we found childhood cancer survivors had a four-fold greater risk of developing lung cancer compared with the general population. Specifically, survivors had increased risk for developing adenocarcinoma, other carcinomas (not small cell or squamous cell), and mesothelioma. When looking at specific diagnosis and treatment factors, an increased risk of subsequent lung cancers was associated with greater doses of chest radiotherapy, older age at childhood cancer diagnosis, and a primary diagnosis of Hodgkin lymphoma, all consistent with previous reports among survivors who were originally diagnosed with their primary cancer as adults (15–19). To our knowledge, this study is the first to demonstrate an increased risk of subsequent lung cancer among survivors of bone cancer. Though prior studies have demonstrated increased risk of subsequent lung cancer following alkylating agent chemotherapy exposure and smoking history among survivors of adult Hodgkin lymphoma (17, 19), these exposures were not identified to be independent risk factors in our study of survivors of childhood cancer.

Radiotherapy has been reported multiple times as a risk factor for subsequent lung cancer in survivors of adult-onset cancers, most notably of Hodgkin lymphoma (17–19) and breast cancer (32–34). This risk factor was observed in our study among survivors of childhood cancer as well. Prior studies of survivors of adult-onset cancers have demonstrated that this risk occurs in a dose-dependent manner (17, 19), which we now demonstrate occurs among survivors of childhood cancer as well. Our study was the first to assess risk for subsequent lung cancer related to radiation field size along with maximum target dose to the chest. We demonstrated that risk for subsequent lung cancer was nearly eight-fold greater for those who received radiation to smaller lung fields versus nearly five-fold greater for those who received radiation to larger lung fields, when both groups were compared to the general population. This could perhaps be due to the fact that field size most typically is inversely related to dose, as such those who had smaller lung fields likely received a greater dose to that area compared with survivors who received treatment to larger lung fields.

It is noteworthy that the majority of our cases in this study were from the 1970–1979 treatment era (n = 27) compared to the 1980–1989 treatment era (n = 13) and 1990–1999 treatment era (n = 2). This is, in part, due to shorter follow up times, but may also reflect a more limited use of radiation in later decades. As such, it will be important to continue to trend the incidence of subsequent lung cancers in future decades to better understand risk with more contemporary treatment regimens.

We did not demonstrate independent risk between subsequent lung cancer and history of smoking when adjusting for other diagnostic and treatment factors, though our survivors who developed lung cancer were more likely to have had a history of smoking compared with survivors who did not develop lung cancer and cumulative incidence of lung SMN was greater in those who were past or current smokers compared with those who had never smoked. This is unique from prior studies of survivors of adult-onset cancers, which have demonstrated an increased risk of subsequent lung cancer in survivors who smoked (17, 19). This lack of association in our study may be due to how smoke exposure is measured in the CCSS cohort (yes/no as described in Methods), as well as under self-reporting or misclassification of smoking status. The measurement does not account for dose response or pack-years, which may yield different findings. In addition, the prevalence of smoking in the CCSS cohort was much lower than that seen in prior studies, which may be a reflection of changing norms in society, limiting the power of our study to detect an association between smoking history and subsequent lung cancer.

Similarly, our study did not demonstrate independent risk for lung cancer secondary to alkylating agents, which had been previously reported (17). This may be secondary to the effect of radiotherapy overwhelming other independent risk factors, though even when not adjusted for radiation exposure, an association between alkylators and lung SMN was not seen. In contrast, we did observe a significant association between bleomycin and lung SMN, when not adjusted for chest radiotherapy, and a strong association when adjusted for chest radiotherapy. This may be due to bleomycin's common use in pediatrics for the treatment of Hodgkin lymphoma. Similarly, the strong association observed between exposure to high doses of platinating agents and lung SMN, may be related to the fact that the individuals exposed to platinating agents in our study were most likely those to be treated for bone tumors. Future larger studies could help better elucidate the contribution of chemotherapeutic agents to this risk.

A primary diagnosis of Hodgkin lymphoma is a well-established risk factor for subsequent lung cancer (13–16), which was replicated in our study. This was most recently evaluated in the Hodgkin lymphoma cohort of the Late Effects Study Group where Hodgkin lymphoma survivors who were male and received chest radiotherapy prior to age 10 years were found to be at greatest risk of subsequent lung cancer (11). Most of the risk for subsequent lung cancer related to prior diagnosis of Hodgkin lymphoma is thought to be attributable to radiotherapy exposure, particularly given the greater doses and lack of lung blocking that occurred in earlier treatment eras (16–19).

We observed an increased risk of subsequent lung cancer after a primary diagnosis of bone cancer. Given that the bone cancers in our study were predominantly osteosarcoma, this may be related to germline mutations of p53, which have been associated with osteosarcomas in adolescence and lung cancer in middle adulthood (35–37). This association with germline mutations could also explain the five cases of lung cancer that occurred without radiation exposure in our study.

Though there are currently no standardized guidelines for lung cancer screening among survivors of childhood cancers, findings of this study suggest survivors of Hodgkin lymphoma and bone cancer may benefit from lung cancer screening with LDCT. Survivors of other childhood cancers who received greater doses of chest radiotherapy may benefit from lung cancer screening as well. Previous trials to establish lung cancer screening guidelines have raised concern that prior radiation exposure may result in greater rates of false positivity by LDCT (38, 39), thereby limiting its benefit among cancer survivors. This is thought to be due to radiation-induced lung disease, which can occasionally appear as a consolidation or scarring on imaging, and on occasion be mistaken for malignancy, resulting in radiologic false positive findings (40). However, there are limited reports that false positivity rates among survivors of adult-onset cancers are comparable to the general population (41). The mean age at time of lung cancer diagnosis in our survivor population was 44 years, suggesting that childhood cancer survivors may develop lung cancer earlier than their general population counterparts within SEER where mean age at time of lung cancer diagnosis was 61 years (2). Consequently, survivors of childhood cancer may benefit from earlier screening than what is suggested by the current general recommendations, which include heavy smokers >55 years of age (21).

Important limitations of this study must be mentioned. Though the CCSS is a large and well-characterized cohort, it does not represent all childhood cancer survivors and there is a potential for participation bias. Selected primary diagnoses are also not included within the cohort, most significantly retinoblastoma. Children with heritable retinoblastoma are at increased risk of subsequent neoplasms and may be at increased risk specifically for subsequent lung cancer, which has been reported previously among British childhood cancer survivors of retinoblastoma (12, 42, 43). Subsequent neoplasms as a whole are initially self-reported, which may lead to missed identification of affected survivors. Additionally, the outcome of interest (subsequent lung cancer) occurs many years after initial diagnosis, so loss to follow-up may result in bias and underreporting.

In summary, survivors of childhood cancer are at increased risk of subsequent lung cancer compared to the general population, with greatest risks observed among survivors who received greater doses of chest radiotherapy, and those with a primary diagnoses of Hodgkin lymphoma or bone cancer. Independent risks related to chemotherapy exposures and smoking history were not observed. However, our study was limited in its ability to assess smoking history. As such we would advocate that survivors of childhood cancer do not smoke, so as to reduce their overall risk for secondary cancers including lung cancer given the strong data in the literature to support this public health recommendation. Associations between subsequent lung cancers and radiotherapy, provide further evidence that radiotherapy exposures should be limited as much as possible to avoid long-term health consequences among survivors of childhood cancer. Additionally, associations between subsequent lung cancers and specific primary diagnoses may be secondary to underlying germline mutations. In aggregate, this suggests that there may be a subset of survivors who would benefit from lung cancer screening. Specifically, our study reflected that survivors who developed lung cancer had a greater rate of mortality compared to those that did not, and the majority died of this subsequent neoplasm. Although the survival of survivors after lung cancer was similar to that of the general population, earlier detection may reduce the significant mortality burden seen in survivors with lung cancer. Further follow up in other large childhood cancer survivor cohorts is necessary to understand the risk for this aging population, which may better inform screening recommendations, particularly as relates to which specific subsets of patients would benefit and at what age screening should begin, as our study demonstrated that survivors of childhood cancer developed subsequent lung cancer at an earlier age compared with the general population.

A.C. Dietz reports other support from bluebird bio, Inc and Shape Therapeutics outside the submitted work. G.T. Armstrong reports grants from NIH during the conduct of the study. R.M. Howell reports grants from St. Jude Children's Research Hospital during the conduct of the study. S.A. Smith reports other support from St. Jude Research Hospital during the conduct of the study. D.A. Mulrooney reports grants from NCI CA55727 (G.T. Armstrong, PI) and NCI Cancer Center Support (CORE grant CA21765; C. Roberts, principal investigator) during the conduct of the study. Y. Yuan reports grants from St Jude Children's Research Hospital during the conduct of the study. No disclosures were reported by the other authors.

T. Ghosh: Conceptualization, data curation, investigation, visualization, writing–original draft, writing–review and editing. Y. Chen: Data curation, software, formal analysis, investigation, methodology, writing–review and editing. A.C. Dietz: Conceptualization, visualization, writing–review and editing. G.T. Armstrong: Resources, data curation, funding acquisition, methodology, project administration, writing–review and editing. R.M. Howell: Resources, data curation, methodology, writing–review and editing. S.A. Smith: Resources, data curation, methodology, writing–review and editing. D.A. Mulrooney: Writing–review and editing. L.M. Turcotte: Conceptualization, supervision, methodology, writing–review and editing. Y. Yuan: Data curation, methodology, writing–review and editing. Y. Yasui: Data curation, formal analysis, supervision, investigation, methodology, writing–review and editing. J.P. Neglia: Conceptualization, data curation, supervision, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We would like to acknowledge the NCI and the American Lebanese-Syrian Associated Charities (ALSAC) for providing funding support for this study. We would like to acknowledge the Childhood Cancer Survivor Study (CCSS) for providing access to previously collected data used in this study. This work was supported by the NCI (CA55727, to 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).

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.