|Home | About | Journals | Submit | Contact Us | Français|
Kevin C. Oeffinger, MD, Departments of Pediatrics and Medicine, Memorial Sloan-Kettering Cancer Center, 1275 York Ave., Box 396, New York, NY 10065, United States gro.ccksm@kgniffeo
John Whitton, MS, Fred Hutchinson Cancer Research Center 1100 Fairview Avenue, M4-B402, PO Box 19024, Seattle WA 98109-1024, United States gro.crchf@nottihwj
Wendy Leisenring, ScD, Fred Hutchinson Cancer Research Center 1100 Fairview Avenue, D5- 360, PO Box 19024, Seattle WA 98109-1024, United States gro.crchf@rnesielw
Joseph Neglia, MD, MPH University of Minnesota Cancer Center, Division of Pediatric Hematology/Oncology, Rm D 557 Mayo Bldg/MMC #484, 420 Delaware Street, SE, Minneapolis, MN 55455, United States ude.nmu@ailgenj
Anna Meadows, MD Children’s Hospital of Philadelphia, Division of Oncology, 34th and Civic Center Blvd., Philadelphia, PA 19104, United States ude.pohc.liame@swodaem
Catherine Crotty, MPH University of Chicago Department of Pediatrics, Section of Hematology, Oncology and Stem Cell Transplantation, 5841 S. Maryland Avenue, MC 4060, Chicago, IL 60637, United States ude.ogacihcu.dsb.sdep@yttorcc
David T. Rubin, MD University of Chicago Department of Medicine, Section of Gastroenterology, 5841 S. Maryland Ave. Chicago, IL 60637, United States ude.ogacihcu.dsb.enicidem@niburd
Lisa Diller, MD, Dana-Farber Cancer Institute, Pediatric Oncology, 44 Binney Street, Boston, MA, 02115, United States ude.dravrah.icfd@rellid_asil
Peter Inskip, ScD Radiation Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, 6120 Executive Blvd., Rm. 7052, Bethesda, MD 20892- 7238, United States vog.hin.liam@eppiksni
Susan A. Smith, MPH The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd. Houston, TX 77030, United States gro.nosrednadm@htimsas
Marilyn Stovall, PhD The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd. Houston, TX 77030, United States gro.nosrednadm@llavotsm
Louis S. Constine, MD University of Rochester School of Medicine and Dentistry, 601 Elmwood Ave, Box 647 Rochester, New York 14642, United States ude.retsehcor.cmru@enitsnoc_siuol
Sue Hammond, MD Nationwide Children’s Hospital, 700 Children’s Drive, Columbus, OH 43205, United States firstname.lastname@example.org
Greg T. Armstrong, MD, MSCE St. Jude Children’s Research Hospital, 262 Danny Thomas Place, Memphis, TN, 28105, United States email@example.com
Leslie L. Robison, PhD, St. Jude Children’s Research Hospital, 262 Danny Thomas Place, Memphis, TN, 28105, United States firstname.lastname@example.org
Paul C. Nathan, MD, MSc, The Hospital for Sick Children, Division of Haematology/Oncology, 555 University Avenue, Toronto, ON M5G 1X8, Canada email@example.com
Childhood cancer survivors develop gastrointestinal malignancies more frequently and at a younger age than the general population, but risk factors for their development have not been well characterized.
To determine the risk and associated risk factors for gastrointestinal subsequent malignant neoplasms (SMN) in childhood cancer survivors.
Retrospective cohort study.
The Childhood Cancer Survivor Study, a multi-center study of childhood cancer survivors diagnosed between 1970 and 1986.
14,358 survivors of a malignancy diagnosed at < 21 years who had survived for 5 or more years from initial diagnosis.
Standardized incidence ratios (SIR) for gastrointestinal SMN were calculated using age-specific population data. Multivariate Cox regression models identified associations between risk factors and gastrointestinal SMN development.
At median follow-up of 22.8 years (range: 5.5-30.2), 45 gastrointestinal malignancies were identified. Gastrointestinal SMN risk was 4.6-fold higher in childhood cancer survivors than the general population (95% confidence interval [CI]: 3.5-6.1). Colorectal cancer SIR was 4.2 (95% CI: 2.8-6.3). The highest gastrointestinal SMN risk was associated with abdominal radiation (SIR=11.2, 95% CI: 7.6-16.4). However, survivors not exposed to radiation had a significantly increased risk (SIR=2.4, 95% CI-1.4-3.9). In addition to abdominal radiation, high dose procarbazine (RR=3.2, 95% CI 1.1-9.4) and platinum drugs (RR 7.6, 95% CI: 2.3-25.5) independently increased the gastrointestinal SMN risk.
This cohort has not yet attained an age at which gastrointestinal malignancy risk is greatest.
Childhood cancer survivors, particularly those exposed to abdominal radiation, are at increased risk for gastrointestinal SMN. These findings suggest that surveillance of at-risk childhood cancer survivors should commence at a younger age than recommended for the general population.
Between 5% and 15% of childhood cancer survivors will develop a subsequent malignant neoplasm (SMN) by 20 to 30 years after the diagnosis of their first cancer.(1-5) SMNs are the second leading cause of premature mortality in childhood cancer survivors after recurrence of the primary cancer.(6) There is emerging evidence that childhood cancer survivors develop gastrointestinal malignancies more frequently and at a younger age than the general population.(5, 7, 8) However, most published studies have been conducted in restricted groups of patients (e.g. Hodgkin lymphoma or testicular cancer survivors)(7-9) or have included survivors of both pediatric and adult cancers.(10-12) Although studies have frequently identified radiation exposure as a risk factor for gastrointestinal SMN,(5) (7, 8, 10-12) none have assessed the relationship between the radiation field and the site of the SMN, or the specific impact of the chemotherapy used to treat the primary malignancy.
Since gastrointestinal malignancies, particularly colorectal cancer, can be detected with surveillance, potentially resulting in reduced morbidity and mortality, the Children’s Oncology Group has published guidelines for colorectal cancer surveillance in childhood cancer survivors. (13, 14) These recommend that survivors exposed to more than 30 Gray of abdominal radiation undergo colonoscopy at a minimum of every 5 years beginning at the latter of 10 years after radiation or at age 35 years.
In the current study, we evaluated the risk of gastrointestinal SMN and the clinical and pathologic factors associated with their development in a large, North American cohort of childhood cancer survivors for whom we have detailed information about the specific therapies received for the primary malignancy. The findings of our analysis will facilitate the refinement of the published surveillance guidelines by allowing more precise identification of which groups of childhood cancer survivors are at greatest risk for developing a gastrointestinal SMN.
The Childhood Cancer Survivor Study (CCSS) is a retrospectively assembled and ongoing hospital-based cohort of 14,358 childhood cancer survivors treated at 26 centers in the US and Canada, established in 1994.(15, 16) Eligibility criteria included: 1) diagnosis of leukemia, central nervous system malignancy, Hodgkin lymphoma, non-Hodgkin lymphoma, neuroblastoma, soft-tissue sarcoma, Wilms tumor, or bone cancer; 2) diagnosis between January 1, 1970 and December 31, 1986; 3) less than 21 years at diagnosis; and 4) alive 5 years from diagnosis date. Each institution identified eligible participants. All eligible survivors were recruited to the cohort. The study was approved by Institutional Review Boards at each participating institution. Informed consent was obtained from each participant or a proxy (if the participant was < 18 years of age or deceased) at the time of enrolment by the CCSS Coordinating Center.
Data were collected from eligible subjects using a baseline questionnaire (in 1994-1998) and three follow-up questionnaires (available at http://ccss.stjude.org) in 2000, 2003, and 2005. Next of kin, typically a parent or spouse, were contacted for those subjects who had died after achieving 5-year survivorship. The questionnaires addressed social and demographic factors, medical conditions, health behaviors, cancer recurrence, SMN development, and family history of cancer.
A detailed summary of cancer treatment exposures was obtained from the hospital charts of participants who consented to medical record release. Cumulative dose data was collected for 22 specific chemotherapy agents and qualitative (yes/no) data was collected for 20 additional agents. Exposure to anthracyclines was expressed as the cumulative dose received. Alkylating agent scores were calculated using methods previously described.(15) Radiation records were reviewed by the CCSS Radiation Physics Center for abstraction and exposure assessment.
We defined a gastrointestinal SMN as any malignancy listed in the Surveillance, Epidemiology and End Results (SEER) classification of tumors of the oral cavity and pharynx (excluding the salivary glands) or digestive system (Surveillance Research Program, National Cancer Institute SEER*Stat software, available at www.seer.cancer.gov/seerstat) version 6.3.6).(17) Gastrointestinal SMN cases were ascertained through self- or proxy-report via the baseline and follow-up questionnaires, and by searches of National Death Index data for US participants (the most recent linkage to the National Death Index occurred in 2010 and included all deaths through 2008). All positive responses were screened, and those representing a likely SMN were verified by the CCSS Pathology Review Center. The CCSS pathologist reviewed a copy of the pathology report to confirm the SMN. If the pathology report could not be obtained, the patient or proxy questionnaire response, death certificate (for deceased patients) and/or medical records were reviewed. Only gastrointestinal SMN occurring 5 or more years following the primary cancer diagnosis were included in the analysis. Data collected regarding gastrointestinal SMN cases included histology and specific location(s) of the tumor. All SMN identified and verified before November 2008 are included in this analysis.
Two pediatric oncologists (TH, PN) reviewed operative notes, pathology reports, radiology reports, and any significant correspondence to determine SMN location. The locations were sketched onto an anatomic diagram and reviewed to determine proximity of the SMN to the primary tumor radiation field. The gastrointestinal malignancy was assigned to one of three proximity categories: “in the beam” if the tumor was within the radiation field; “near the beam” (0 to 10 cm from the radiation field edge) or “out of the beam” (>10 cm from the radiation field edge). Specific radiation doses to the site of each SMN were not estimated.
The incidence of gastrointestinal neoplasms in the survivor cohort was compared to the general US population using the SEER database, which is generated from population-based cancer registries across the US. SEER data were used to ascertain the number of GI cancers expected in a general population cohort with the same age and sex distribution as the CCSS cohort. Standardized incidence ratios (SIR) were calculated as the ratio of the observed numbers of cases in the various subgroups of the cohort to the expected numbers in the general population. Poisson regression models were used to calculate SIR, their 95% confidence intervals (CI), and p values. Excess absolute risk (EAR) was calculated as the difference between the number of observed and expected events divided by the number of person years of follow-up, and is expressed per 100,000 person years. Analyses were performed using SAS version 9.1.
Cumulative incidence estimates for gastrointestinal SMN were calculated using the algorithm described by Kalbfleisch and Prentice (pages 168-9). These estimates represent the cumulative number of gastrointestinal SMNs that had occurred by each time point beyond 5 years after the childhood cancer, as a proportion of subjects in the denominator, with appropriate accounting for censoring and treating death as a competing risk such that follow-up for a subject who died is truncated at death.(18)
Univariate Cox regression analysis evaluated the association between gastrointestinal SMN development and primary cancer diagnosis, age at diagnosis of primary cancer, sex, race, treatment era, radiation therapy, chemotherapy (platinum drugs, alkylating agents, epipodophyllotoxins, antimetabolites, plant alkaloids and anthracyclines), smoking history, and family history of cancer. A multivariable Cox regression model was constructed that included significant univariate factors (p<0.10), coupled with variables not statistically significant but considered important to evaluate given our a priori hypotheses or their potential as confounders. The initial multivariable model assessed the following factors that were significant in univariate analysis: primary diagnosis; radiation to brain, neck, abdomen, pelvis; alkylating agents, platinum drugs, antimetabolites and anthracyclines (Appendix 1). Sex, race, treatment era, plant alkaloids and epipodophyllotoxins were also included in the initial model since we had hypothesized a priori that they would impact SMN risk. The final model included those covariates that modified the affect of other factors (as confounders) and excluded those covariates that did not significantly affect time to gastrointestinal SMN development when included in the joint model. Only participants for whom we had complete primary cancer treatment data were included in the final multivariate model. Subsequently, we performed sensitivity analyses in which we included those participants with missing treatment data (procarbazine, platinum, anthracyclines, plant alkaloids and abdominal radiation) in the final model under 3 different assumptions: all missing participants received 1) all treatments; 2) no treatment; and 3) treatment pattern most common for their primary cancer. Age was used as the time scale in all Cox models to account for the strong impact that increasing age has on SMN development.(19) The proportional hazards assumption was examined for the effects of several radiation and chemotherapy risk factors and no significant changes in hazard ratios over time were found. Standard asymptotic inference methods for Cox regression based on the partial likelihood were used to construct 95% CI and to carry out two-sided tests of statistical significance.
The CCSS is a multi-institutional study funded by the National Cancer Institute (NCI, grant # U24 CA55727). Additional support was provided by the Intramural Research Program of the NIH and NCI, and the American Lebanese-Syrian Associated Charities. Dr. Henderson received support from the NCI (grant # K07CA134935).
Figure 1 depicts the individuals identified and included in each stage of the CCSS cohort development. The analytic dataset with complete follow-up information included 14,337 survivors. Of note, medical records were available for 12,592 participants and thus, constituted the dataset utilized for all analyses involving treatment data.
Among 14,337 childhood cancer survivors, 802 SMN (not including non-melanoma skin cancer) were identified in 732 individuals. Of those, 45 (5.6%) gastrointestinal SMN were identified in 45 individuals at a median follow-up of 22.8 years (range: 5.5-30.2 years) from primary diagnosis. The median age at gastrointestinal SMN diagnosis was 33.5 years (range: 9.7-44.8 years). As shown in Table 1, survivors who developed a gastrointestinal SMN were similar to survivors who did not in terms of race, sex, family history of cancer gastrointestinal cancer, and smoking history. The two populations differed statistically in their distribution among treatment eras, exposure to abdominal and pelvic radiation, primary cancer diagnosis and vital status.
The characteristics of those who developed a gastrointestinal SMN are shown in Table 2. Most SMN (80%) occurred more than 24 years after the primary childhood cancer. The most frequent site was the colon followed by the rectum/anus. Twenty-five (56%) of the 45 gastrointestinal SMN were adenocarcinomas. Twenty-three of the 45 patients (51%) were deceased, of whom 15 (65%) had died as a consequence of their gastrointestinal SMN.
Thirty-nine (87%) of the survivors who developed a gastrointestinal SMN had confirmed radiation for their primary cancer. Eighty-two percent of these survivors developed their SMN “in” or “near” their previous radiation field, while 15% developed their SMN “out” of the previous radiation field. The distribution of radiation fields relative to gastrointestinal SMN location was similar among the 25 survivors who developed an adenocarcinoma of the gastrointestinal tract to that observed among those cases with other histologies (data not shown).
We compared the incidence of gastrointestinal SMN in the CCSS cohort to the general population using the SEER database. As shown in Table 3, gastrointestinal malignancy risk was almost 5-fold higher in childhood cancer survivors compared to the general population. Survivors with abdominal radiation history had a greater than 11-fold risk for developing a gastrointestinal malignancy. Wilms tumor survivors had the highest risk, followed by Hodgkin lymphoma survivors. Colorectal malignancy risk was over 4-fold higher in childhood cancer survivors compared to the general population. Similar to all gastrointestinal SMN cases, Wilms tumor survivors had the greatest risk of developing colorectal cancer.
The cumulative incidence of gastrointestinal SMN by 30 years after the primary cancer diagnosis was 0.64% (95% CI: 0.43-0.86%) (Figure 2A). The cumulative incidence in those who received abdominal radiation was 1.97% (95% CI: 1.15-2.80%) by 30-years (Figure 2B). The 30-year cumulative incidence rates were highest among survivors of Hodgkin lymphoma and Wilms tumor at 2.08% (95% CI: 1.08-3.07%) and 1.2% (95% CI: 0.12-2.29%), respectively (Figure 2C). The cumulative incidence of colorectal SMN by 30 years after the primary cancer diagnosis was 0.41% (95% CI: 0.22-0.59%). Abdominal radiation was associated with colorectal SMN cumulative incidence of 1.16% (95% CI: 0.47-1.85) by 30-years (Figure 2D).
Multivariate Cox regression analysis (Table 4) revealed that exposure to abdominal radiation was associated with an over 5-fold increased risk of gastrointestinal SMN. After controlling for abdominal radiation, high-dose procarbazine and platinum exposure were independently associated with an increased gastrointestinal SMN risk. As described in the methods, sensitivity analyses to evaluate the impact of missing treatment variables (including procarbazine, platinum, anthracyclines, plant alkaloids and abdominal radiation exposure data) were carried out and revealed risk estimates consistent with our final model (data not shown).
Since procarbazine was administered orally to all survivors who developed a gastrointestinal SMN, we explored whether the oral administration of other alkylating agents was associated with an increased risk for gastrointestinal SMN when compared with intravenous administration. No effect was observed (data not shown).
To our knowledge, this is the largest study that has focused on the risk for gastrointestinal malignancies in childhood cancer survivors, and considers detailed treatment information, including chemotherapy exposures and radiation fields. We observed that young childhood cancer survivors, particularly Wilms tumor, Hodgkin lymphoma, bone and brain tumors, are at an increased risk for developing gastrointestinal malignancies when compared to the age-matched general population. Gastrointestinal SMN presented in survivors as young as 9 years, and while all observed SMN occurred prior to age 45 years, we expect the incidence to continue to rise as they age.(17) We observed that survivors exposed to abdominal radiation were at particularly increased risk for developing a gastrointestinal SMN. Furthermore, even those survivors not exposed to radiation demonstrated an increase in risk for gastrointestinal SMN.
Given that radiation exposure is a well-established risk factor for SMN development, we anticipated that survivors treated with abdominal radiation would be at greatest risk for the development of a gastrointestinal malignancy. However, 13/45 of the gastrointestinal SMNs occurred outside of the radiation field, or in survivors who had not received any radiation therapy as part of their primary cancer treatment. It is plausible that genetic predisposition may have contributed to these tumors even though self-report of a family history of cancer was not associated with an increased risk of gastrointestinal SMN.
Procarbazine and platinum drugs were independently associated with the risk of GI SMN in the radiation field, suggesting that these two agents may potentiate the carcinogenic effect of radiation. This observation of an independent, dose-related risk of procarbazine with abdominal radiation supports van den Belt-Dusebout’s findings of an association between procarbazine and the development of gastric cancer in testicular and Hodgkin lymphoma survivors who were treated with abdominal radiation.(9) However, similar to their study, we did not find that procarbazine without abdominal radiation was associated with an increased risk of GI SMN. This may have been because the number of individuals treated with procarbazine without radiation was too small for analysis. It has also been suggested in another study by van Belt-Dusebout (20) that cisplatin is associated with an increased risk of SMN (all types) in testicular cancer survivors treated without radiation. Whether cisplatin, without abdominal radiation, is associated with an increased risk of GI SMN cannot be ascertained in our study. It is important to note that cisplatin may simply be a proxy for an underlying germline mutation such as TP53. Cisplatin is primarily used in the treatment of osteosarcoma or CNS tumors, both of which are included in the classical definition of Li-Fraumeni syndrome.(21, 22) Both of the childhood cancer survivors treated with cisplatin without abdominal radiation who subsequently developed a GI SMN had an osteosarcoma as their primary diagnosis. Indeed, of the 11 survivors with a GI SMN either outside of the radiation field or without a history of radiation, 9 had primary cancers that have been associated with a TP53 germline mutation (osteosarcoma, 4; CNS tumor, 3; leukemia 2). Recently, Ruijs et al (23) reported an excess risk of colorectal and pancreatic cancer in TP53- positive families.
Our results were interpreted in the context of relevant findings from other published work. Briefly, we searched English-language MEDLINE/PubMED, OVID and Embase from 1950 to 2011 including the following search terms: second primary neoplasms, abdominal, anal, and digestive neoplasms. Most studies identified were in cohorts of adult Hodgkin lymphoma or testicular cancer survivors, and were not limited to childhood cancer survivors. None of these studies focused exclusively on gastrointestinal SMN in childhood cancer survivors. Recently, the British Childhood Cancer Survivor Study, which consists of an older population of childhood cancer survivors compared to the CCSS, demonstrated that digestive SMN contribute the largest absolute excess risk of any malignancy in childhood cancer survivors older than 40 years. Digestive SMN accounted for 18% of the excess cancer risk in older survivors in that cohort,(5) suggesting that we will continue to observe an increase in the incidence of gastrointestinal SMN as our cohort ages. Most published studies that demonstrated an increased risk of solid tumors in the gastrointestinal tract attributed the increased risk to the radiation therapy used to treat the primary cancer.(7, 8, 10-12, 24) In the British Childhood Cancer Survivor Study report, abdomino-pelvic radiation increased the risk of digestive SMN by 3.3-fold, although detailed and validated radiation fields and chemotherapy exposures were not examined.(5) Bhatia described SMN among 1,380 childhood Hodgkin lymphoma survivors and reported a significantly elevated excess risk at a young age of both gastric and colorectal cancers.(7) In addition to radiation exposure, younger age at primary cancer diagnosis significantly increased risk. In a British cohort of Hodgkin lymphoma survivors, Swerdlow and colleagues only observed a borderline association between radiation therapy and gastrointestinal SMN (SIR = 1.7; 95% CI 1.0-2.5). However, patients who received mixed modality therapy (i.e. radiation and chemotherapy) had a greater than 3-fold increase in risk as compared to the general population (SIR = 3.3, 95% CI 2.1-4.8).(10)
There are limitations to our study that must be considered when interpreting the results. The total number of observed gastrointestinal SMN was small and therefore it was not feasible to examine the effect of demographic factors such as race and geography on the risk for their development. Further, analyses aimed at distinguishing between the effects of treatment and primary diagnoses (as representing inherent genetic propensity) were not possible given the small numbers. Since SMN are self-reported in the CCSS, under-reporting is possible. Family history is also self-reported and so may be inaccurate. The analysis was limited to only those gastrointestinal SMNs that occurred 5 or more years after the primary cancer diagnosis; thus, our study could not identify risks for early gastrointestinal SMNs.
Since cure of the primary childhood malignancy is a priority, we do not advocate for modification of the current treatment protocols used for childhood cancers in order to decrease the long-term risk for gastrointestinal SMN. However, pediatric oncologists strive to reduce or eliminate late toxicity without impacting the probability of cure, and so the necessity of therapies such as radiation is under constant scrutiny. Our observations should enable researchers and clinicians to better identify those survivors at highest risk for gastrointestinal SMN, potentially facilitating the implementation of better surveillance in clinical practice. Colorectal cancer is a malignancy known to have improved outcomes with early detection in the general population and in other high-risk populations.(25-28) As such, the Children’s Oncology Group currently recommends survivors exposed to more than 30 Gray of abdominal radiation undergo colonoscopy at a minimum of every 5 years beginning at the latter of 10 years after radiation or at age 35 years. If the findings of this study are confirmed, physicians should also consider chemotherapy exposures when determining the indications for early colorectal cancer surveillance in childhood cancer survivors.
We would like to acknowledge Eneida Mendonca, MD., PhD, for her assistance with literature review.
Grant Support: The Childhood Cancer Survivor Study is a collaborative, multi-institutional study funded by the National Cancer Institute (NCI, grant # U24 CA55727; L.L. Robison, Principal Investigator). Additional support was provided by the Intramural Research Program of the NIH and NCI, and the American Lebanese-Syrian Associated Charities. The primary investigator, Dr. Henderson, received support from the NCI (grant # K07CA134935, Health Beliefs and Behaviors: Cohort Studies in Childhood Cancer Survivors).
Funding Source: This work was supported by the National Cancer Institute (grant number U24- CA55727, to Leslie L. Robison, Principal Investigator).