Endocrine and Cardiovascular Late Effects among Adult Survivors of Childhood Brain Tumors

Transcription

1 663 Endocrine and Cardiovascular Late Effects among Adult Survivors of Childhood Brain Tumors Childhood Cancer Survivor Study James G. Gurney, Ph.D. 1 Nina S. Kadan-Lottick, M.D., M.S.P.H. 1 Roger J. Packer, M.D. 2 Joseph P. Neglia, M.D., M.P.H. 1 Charles A. Sklar, M.D. 3 Judy A. Punyko, M.S. 1 Marilyn Stovall, Ph.D. 4 Yutaka Yasui, Ph.D. 5 H. Stacy Nicholson, M.D., M.P.H. 6 Suzanne Wolden, M.D. 7 Dawn E. McNeil, M.D. 8 Ann C. Mertens, Ph.D. 1 Leslie L. Robison, Ph.D. 1 1 Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota. 2 Department of Neurology, Children s National Medical Center, Washington, DC 3 Department of Pediatrics, Memorial Sloan- Kettering Cancer Center, New York, New York. 4 Department of Radiation Physics, University of Texas, M. D. Anderson Cancer Center, Houston, Texas. 5 Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, Seattle, Washington. 6 Division of Pediatric Hematology/Oncology, Oregon Health and Science University, Portland, Oregon. 7 Department of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, New York, New York. 8 Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, Maryland. Supported by National Cancer Institute grant U24 CA and the Children s Cancer Research Fund at the University of Minnesota. The following are Childhood Cancer Survivor Study (CCSS) institutions and investigators: University of California-San Francisco, CA (Arthur Ablin, M.D., Institutional Principal Investigator); University of Alabama, Birmingham, AL (Roger Berkow, M.D., BACKGROUND. Survivors of childhood brain tumors (CBTs) are at high risk for a variety of late adverse effects. Most research on long-term effects of CBTs has been comprised of single-institution case series without comparison groups. Research on CBT late effects often is focused on neurologic and sensory outcomes, with less emphasis on other potential targets such as the endocrine and circulatory systems. The current study was conducted to contrast the incidence of endocrine and cardiovascular conditions among CBT survivors as a function of treatment and to determine the risk of occurrence of these conditions relative to a sibling comparison group. METHODS. As part of the Childhood Cancer Survivor Study (CCSS), treatment data were collected from medical records and self-reported late effects were ascertained from a survey questionnaire of 1607 CBT patients who survived their disease for 5 or more years. For comparison purposes, questionnaire data were also collected from 3418 randomly selected siblings of participants in CCSS. RESULTS. One or more endocrine conditions were reported by 43% of CBT survivors. Compared with siblings, CBT survivors had a significantly increased risk of late-onset ( 5 years postdiagnosis) hypothyroidism (relative risk [RR] 14.3; 95% confidence interval [95% CI] ), growth hormone deficiency (RR 277.8; 95% CI ), the need for medications to induce puberty (RR 86.1; 95% CI ), and osteoporosis (RR 24.7; 95% CI ). One or more cardiovascular conditions were reported by 18% of CBT survivors, with an elevated late-onset risk for stroke (RR 42.8; 95% CI ), blood clots (RR 5.7; 95% CI ), and angina-like symptoms (RR 2.0; 95% CI ). Very few late effects were evident among those treated with surgery only, but risks were consistently elevated for those treated with radiation and surgery, and higher still for those who also received adjuvant chemotherapy. CONCLUSIONS. Childhood brain tumor survivors are at a significantly increased risk for several adverse endocrine and cardiovascular late effects, particularly if they were treated with radiation and chemotherapy. Lifetime medical surveillance and Institutional Principal Investigator); International Epidemiology Institute, Rockville, MD (John Boice, Sc.D., member CCSS Steering Committee); University of Washington, Seattle, WA (Norman Breslow, Ph.D., member CCSS Steering Committee); University of Texas Southwestern Medical Center at Dallas, TX (George R. Buchanan, M.D., Institutional Principal Investigator); Dana-Farber Cancer Institute, Boston, MA (Lisa Diller, M.D., Institutional Principal Investigator; Holcombe Grier, M.D., former Institutional Principal Investigator; Frederick Li, M.D., member CCSS Steering Committee); Texas Children s Center, Houston, TX (Zoann Dreyer, M.D., Institutional Principal Investigator); Seattle Children s Hospital, Seattle, WA (Debra Friedman, M.D., M.P.H., Institutional Principal Investigator; Thomas Pendergrass, M.D., former Institutional Principal Investigator); Roswell Park Cancer Institute, Buffalo, NY (Daniel M. Green, M.D., Institutional Principal Investigator and member CCSS Steering Committee); Hospital for Sick Children, Toronto, ON (Mark Greenberg, M.B., Ch.B., Institutional Principal Investigator); St. Louis Children s Hospital, MO (Robert Hayashi, M.D., 2003 American Cancer Society

2 664 CANCER February 1, 2003 / Volume 97 / Number 3 follow-up for potential toxicities are necessary because treatment-related complications may occur many years after therapy. Cancer 2003;97: American Cancer Society. DOI /cncr KEYWORDS: brain neoplasms, central nervous system neoplasms, epidemiology, treatment effects, surgery, radiation, chemotherapy, late effects, survivorship, children, adolescents, young adults, hypothyroidism, growth deficiency, stroke. The survival rate of patients with childhood brain (CBT) and central nervous system (CNS) tumors has improved modestly in recent years. In the aggregate, the 5-year relative survival rate for individuals diagnosed with malignant CBTs in the United States between 1992 and 1998 was 69.9%, compared with 54.8% for individuals diagnosed between 1974 and The case fatality rate of CBTs varies as a function of tumor biology, histology, site, extent of tumor resection, and age at diagnosis. 2,3 The survival probability is generally good for low-grade neoplasms such as pilocytic astrocytomas and is generally poor for neoplasms such as disseminated medulloblastoma, high-grade gliomas, and subsets of ependymomas. Because of the vulnerability and limited repair properties of brain tissue and its neural functioning, survivors of CBTs are at relatively high risk for a variety of adverse effects. 3,4 The National Cancer Institute (NCI) estimated that 88,614 individuals (all ages) were alive in 1999 with a current or past primary malignancy of the brain. 5 Unlike the NCI estimate, Davis et al. 6 combined both malignant and nonmalignant primary brain tumors. They estimated brain tumor prevalence at 350,000 persons of all ages in the year 2000 in the United States. The prevalence of primary brain tumors represents the surviving population of persons once treated or still under treatment for a brain tumor, and thus the population at risk for seeking care for related adverse effects. Given the potential long-term morbidity associated with brain tumors, these numbers illustrate the need to better understand the potential late adverse effects that survivors and their primary care providers may encounter. Childhood brain tumors and their associated treatments can result in a multitude of physical, medical, cognitive, and psychosocial adverse outcomes. 3,4,7 Most research reports on the long-term effects of CBTs have been comprised of single-institution case series without comparison groups. The CBT late effects research usually focuses on neurologic and sensory outcomes, with less emphasis on other potential targets such as the endocrine and circulatory systems. Institutional Principal Investigator; Teresa Vietti, M.D., former Institutional Principal Investigator); St. Jude Children s Research Hospital, Memphis, TN (Melissa Hudson, M.D., Institutional Principal Investigator and member CCSS Steering Committee); University of Michigan, Ann Arbor, MI (Raymond Hutchinson, M.D., Institutional Principal Investigator); Stanford University School of Medicine, Stanford, CA (Michael P. Link, M.D., Institutional Principal Investigator; Sarah S. Donaldson, M.D., member CCSS Steering Committee); Children s Hospital of Philadelphia, PA (Anna Meadows, M.D., Institutional Principal Investigator and member CCSS Steering Committee; Bobbie Bayton, member CCSS Steering Committee); Children s Hospital, Oklahoma City, OK (John Mulvihill, M.D., member CCSS Steering Committee); Children s Hospital, Denver, CO (Brian Greffe, M.D., Institutional Principal Investigator; Lorrie Odom, M.D., former Institutional Principal Investigator); Children s Health Care-Minneapolis, MN (Maura O Leary, M.D., Institutional Principal Investigator); Columbus Children s Hospital, OH (Amanda Termuhlen, M.D., Institutional Principal Investigator; Frederick Ruymann, M.D., former Institutional Principal Investigator; Stephen Qualman, M.D., member CCSS Steering Committee); Children s National Medical Center, Washington, DC (Gregory Reaman, M.D., Institutional Principal Investigator; Roger Packer, M.D., member CCSS Steering Committee); Children s Hospital of Pittsburgh, PA (A. Kim Ritchey, M.D., Institutional Principal Investigator; Julie Blatt, M.D., former Institutional Principal Investigator); University of Minnesota, Minneapolis, MN (Leslie L. Robison, Ph.D., Institutional Principal Investigator and member CCSS Steering Committee; Ann Mertens, Ph.D., member CCSS Steering Committee; Joseph Neglia, M.D., M.P.H., member CCSS Steering Committee; Mark Nesbit, M.D., member CCSS Steering Committee; Stella Davies, M.D., Ph.D., member CCSS Steering Committee); Children s Hospital Los Angeles, CA (Kathy Ruccione, R.N., M.P.H., Institutional Principal Investigator); Memorial Sloan-Kettering Cancer Center, New York, NY (Charles Sklar, M.D., Institutional Principal Investigator and member CCSS Steering Committee); National Cancer Institute, Bethesda, MD (Malcolm Smith, M.D., member CCSS Steering Committee; Martha Linet, M.D., member CCSS Steering Committee); Mayo Clinic, Rochester, MN (W. Anthony Smithson, M.D., Institutional Principal Investigator; Gerald Gilchrist, M.D., former Institutional Principal Investigator); University of Texas M. D. Anderson Cancer Center, Houston, TX (Louise Strong, M.D., Institutional Principal Investigator and member CCSS Steering Committee; Marilyn Stovall, Ph.D., member CCSS Steering Committee); Riley Hospital for Children, Indianapolis, IN (Terry A. Vik, M.D., Institutional Principal Investigator; Robert Weetman, M.D., former Institutional Principal Investigator); Fred Hutchinson Cancer Center, Seattle, WA (Yutaka Yasui, Ph.D., Institutional Principal Investigator; John Potter, M.D., Ph.D., former Institutional Principal Investigator and member CCSS Steering Committee); and University of California-Los Angeles, CA (Lonnie Zeltzer, M.D., Institutional Principal Investigator and member CCSS Steering Committee). Address for reprints: James G. Gurney, Ph.D., Division of Pediatric Epidemiology, University of Minnesota, Mayo Mail Code 715, 420 Delaware St. SE, Minneapolis, MN 55455; Fax: (612) ; gurney@epi.umn.edu Received June 18, 2002; revision received August 20, 2002; accepted September 16, 2002.

3 Late Effects among Brain Tumor Survivors/Gurney et al. 665 As part of the Childhood Cancer Survivor Study (CCSS), a 25-institution follow-up study of persons who survived a childhood neoplasm for 5 years or longer, we collected detailed treatment data from medical records. In addition, we ascertained information on late effects for 1607 CBT survivors through self-report or parental report. For comparison purposes, we also collected questionnaire data on more than 3418 randomly selected siblings of participants in CCSS. This article provides results from our evaluation of nonneurologic outcomes. Neurologic and psychosocial outcomes among CBT survivors in CCSS will be reported separately. Because so few adverse outcomes were reported in our study for respiratory, digestive, and urinary systems, this analysis focuses on endocrine and cardiovascular-related effects for both case sibling comparisons and for treatment-related effects among CBT cases alone. MATERIALS AND METHODS A detailed accounting of the methodology for CCSS was published previously. 8 An abbreviated description is provided below. Inclusion Criteria The CCSS includes individuals who received their primary treatment at 1 of 25 collaborating institutions and who survived at least 5 years after diagnosis. Eligibility was restricted to those with a primary CBT, leukemia, Hodgkin disease, non-hodgkin lymphoma, kidney tumor, neuroblastoma, soft tissue sarcoma, or bone tumor. Individuals were diagnosed between 1970 and 1986 and were 20 years of age or younger at the time of diagnosis. The CCSS protocols and documents were approved by the Human Subjects Committees at the University of Minnesota (the study coordinating center) and at each collaborating institution. Each subject provided informed consent to participate in the study and provided consent for release of medical records. To recruit the sibling comparison group, a random sample of participating cases was selected and one sibling was randomly chosen for recruitment. The collaborating institutions identified 20,276 five-year survivors (cases). At the time of this analysis, 14,054 were enrolled and completed an interview, 3132 declined to participate, 2996 were lost to followup and were never offered enrollment, and 94 subjects are pending inclusion into the study cohort. Of the 1818 participating cases with a CBT who completed the baseline questionnaire, we obtained treatment information on 1607 (88%). These cases were included in our analysis. Data Collection The CCSS staff obtained medical records from the referring institution to abstract diagnosis and treatment information, including surgical procedures and chemotherapy and radiotherapy regimens. Our coinvestigators at M. D. Anderson Cancer Center determined radiation field dosimetry, including maximum doses to the brain, spine, and thyroid. Participants completed a 24-page questionnaire with a wide range of information on demographic characteristics, health habits, and medical conditions. A copy of the survey instrument is available for review and downloading at The participant, or a parent/guardian for minors and cognitively impaired persons, either completed the questionnaire by mail or by telephone with a trained interviewer. The questions on medical conditions were introduced with Have you ever been told by a doctor or other health care professional that you have or have had...? Respondents who provided an affirmative answer were asked to provide the age when the condition first occurred. Data Analysis Our analytic approach was developed and supported by the study s statistical coordinating center at Fred Hutchinson Cancer Research Center. The aims of our analysis were to 1) calculate rates of occurrence for each medical condition (outcome), stratified by time period of reported first occurrence; 2) compare rates of occurrence between the brain tumor survivors and the sibling comparison group (case sibling comparisons); and 3) compare the hazard ratio, reported as the relative risk (RR), of late outcomes (first occurrence 5 years from diagnosis) among the survivors as a function of treatment group (case case comparisons). For calculating rates based on person-time accrual, the time periods of interest were diagnosis to end of treatment, end of treatment to 5 years postdiagnosis, and 5 years postdiagnosis to date of interview. We consider a first occurrence during the latter period to be an incident late effect. We limited our case case comparisons for assessing cancer treatment effects to this time period. We combined missing outcome responses with those of no and not sure and categorized them as no occurrence. When a yes response was recorded without an age at first occurrence, we imputed the time period of occurrence using the multiple imputation methodology 9 employed for event-time imputations by Taylor et al., 10 with slight modifications. Our imputation methodology employed piece-wise exponential models to predict the missing time of occurrence, dependent on age at

4 666 CANCER February 1, 2003 / Volume 97 / Number 3 diagnosis, age at interview, gender, histology, and treatment characteristics. We calculated incidence rate estimates and their 95% confidence intervals (95% CI) by standard formulas 9 to make inferences from multiple-imputed datasets. We used Cox proportional hazard regression models, controlling for gender, to estimate the hazard rate (risk) of developing each medical condition among survivors, relative to siblings, for each time period. Because 399 of the 3418 participating siblings were related specifically to a CBT case, we adjusted the standard error estimates using the sandwich method for correlated data analysis. 11 We used similar Cox regression techniques for the case case comparisons by treatment group, comparing the hazard ratio (RR) of outcome occurrence 5 years or more after diagnosis. We controlled for gender, age at diagnosis (4 or younger, 5 9, 10 20), and histology (astrocytoma/glioma, primitive neuroectodermal tumor [PNET]/medulloblastoma, ependymoma, or other histology). We categorized cases into one of four treatment groups: surgery only; surgery and radiation; surgery, radiation, and chemotherapy; and other therapy. Only 65 cases fell into the heterogenous other therapy category. Although these cases are included in our analysis, we do not report all results for this group. Very few outcome events occurred in the surgery only treatment group. As a result, we used the surgery and radiation treatment group as our reference category for the case case RR estimates. To ensure meaningful evaluations of the associations of outcome occurrence with treatment and other factors of interest, we restricted our statistical analysis to outcomes that were reported by 21 cases with known age at first outcome occurrence. RESULTS Cases were younger and were more likely to be male than those in the sibling comparison group (Table 1). Among cases, 34% were 4 years old or younger at diagnosis and 66% had an astrocytoma/glioma tumor. About 11% of cases were deceased when their family was contacted. Including cases younger than 18 years at interview, about 37% of case questionnaires were completed by a parent or other family member. The treatment distribution of cases is also shown in Table 1. Surgery with radiation was the most frequent therapeutic modality (42%), followed by surgery, radiation, and chemotherapy (28%), and then surgery only (26%). The distribution of histologic subtypes as a function of treatment group is shown in Table 2. The surgery only treatment group was composed primarily of children diagnosed with astrocytoma/glioma tumors (92%). The surgery, radiation, and chemotherapy TABLE 1 Characteristics of Participating Childhood Brain Tumor Survivors (n 1607) and the Sibling Comparison Group (n 3418) of the Childhood Cancer Survivor Study Characteristics No. cases (%) No. siblings (%) Age at Interview (yrs) (38.6) 1001 (29.3) (43.1) 1221 (35.7) (17.1) 941 (27.5) (1.2) 255 (7.5) Gender Males 873 (54.3) 1645 (48.1) Females 734 (45.7) 1773 (51.9) Household income $20, (7.8) 135 (3.9) $20,000 39, (23.8) 789 (23.1) $40, (44.9) 1990 (58.2) Unknown 377 (23.5) 504 (14.8) Vital status Alive 1433 (89.2) 3418 (100) Dead 174 (10.8) 0 (0) Age at diagnosis (yrs) (34.0) (29.9) (36.1) Histology Astrocytoma/glioma 1066 (66.3) Medulloblastomas/PNET 343 (21.3) Ependymoma 118 (7.3) Other CNS tumors 80 (5.0) Treatment Surgery only 414 (25.8) Surgery and radiation 682 (42.4) Surgery, radiation, and chemotherapy 446 (27.8) Other treatments 65 (4.0) PNET: primitive neuroectodermal tumor; CNS: central nervous system. treatment group contained equal proportions of PNET/medulloblastoma (42%) and astrocytoma/glioma patients (42%). Table 3 shows the distribution of the maximum radiation dose among cases who received surgery and radiation, with or without chemotherapy. Cases in the chemotherapy group received higher doses of radiation to the brain, spine, and thyroid than those who did not receive chemotherapy. Endocrine Outcomes Our statistical evaluation of endocrine-related conditions includes hypothyroidism, deficiency of growth hormone, history of receiving growth hormone injection, osteoporosis (including brittle, weak, or fragile bones), need for medication to induce puberty, primary amenorrhea (among female respondents 16 years of age or older), and history of taking female hormones to have a menstrual period (among female respondents 16 years of age or older). Conditions that

5 Late Effects among Brain Tumor Survivors/Gurney et al. 667 TABLE 2 Distribution of Selected Adverse Outcomes by Histology within Treatment Group Treatment groups S S R S R C Others Outcome a No. (%) No. (%) No. (%) No. (%) Total All CNS cases Astrocytoma/glioma 382 (92.3) 447 (65.5) 188 (42.2) 49 (75.4) 1066 PNET/medulloblastoma 4 (1.0) 140 (20.5) 189 (42.4) 10 (15.4) 343 Ependymoma 9 (2.2) 53 (7.8) 55 (12.3) 1 (1.5) 118 Other CNS 19 (4.6) 42 (6.2) 14 (3.1) 5 (7.7) 80 Total 414 (100) 682 (100) 446 (100) 65 (100) 1607 Hypothyroidism Astrocytoma/glioma 10 (83.3) 57 (51.4) 50 (40.0) 6 (85.7) 123 PNET/medulloblastoma 1 (8.3) 40 (36.0) 59 (47.2) 0 (0.0) 100 Ependymoma 0 (0.0) 3 (2.7) 10 (8.0) 0 (0.0) 13 Other CNS 1 (8.3) 11 (9.9) 6 (4.8) 1 (14.3) 19 Total 12 (100) 111 (100) 125 (100) 7 (100) 255 Growth hormone deficiency Astrocytoma/glioma 7 (70.0) 80 (55.6) 57 (33.9) 11 (73.3) 155 PNET/medulloblastoma 1 (10.0) 40 (27.8) 89 (53.0) 2 (13.3) 132 Ependymoma 0 (0.0) 8 (5.6) 18 (10.7) 0 (0.0) 26 Other CNS 2 (20.0) 16 (11.1) 4 (2.4) 2 (13.3) 24 Total 10 (100) 144 (100) 168 (100) 15 (100) 337 Received growth hormone injections Astrocytoma/glioma 4 (80.0) 57 (51.8) 40 (30.5) 7 (70.0) 108 PNET/medulloblastoma 1 (20.0) 33 (30.0) 74 (56.5) 1 (10.0) 109 Ependymoma 0 (0.0) 8 (7.3) 11 (8.4) 0 (0.0) 19 Other CNS 0 (0.0) 12 (10.9) 6 (4.6) 2 (20.0) 20 Total 5 (100) 110 (100) 131 (100) 10 (100) 256 Osteoporosis Astrocytoma/glioma 7 (100.0) 20 (87.0) 12 (44.4) 0 (0.0) 39 PNET/medulloblastoma 0 (0.0) 2 (8.7) 8 (29.6) 2 (100.0) 12 Ependymoma 0 (0.0) 1 (4.4) 5 (18.5) 0 (0.0) 6 Other CNS 0 (0.0) 0 (0.0) 2 (7.4) 0 (0.0) 2 Total 7 (100) 23 (100) 27 (100) 2 (100) 59 Medications needed to induce puberty Astrocytoma/glioma 4 (80.0) 29 (59.2) 16 (35.6) 2 (40.0) 51 PNET/medulloblastoma 0 (0.0) 10 (20.4) 18 (40.0) 1 (20.0) 29 Ependymoma 0 (0.0) 2 (4.1) 4 (8.9) 0 (0.0) 6 Other CNS 1 (20.0) 8 (16.3) 7 (15.6) 2 (40.0) 18 Total 5 (100) 49 (100) 45 (100) 5 (100) 104 Arrhythmia Astrocytoma/glioma 6 (85.7) 8 (72.7) 4 (50.0) 0 (0.0) 18 PNET/medulloblastoma 0 (0.0) 0 (0.0) 2 (25.0) 0 (0.0) 2 Ependymoma 1 (14.3) 3 (27.3) 1 (12.5) 0 (0.0) 5 Other CNS 0 (0.0) 0 (0.0) 1 (12.5) 0 (0.0) 1 Total 7 (100) 11 (100) 8 (100) 0 (100) 26 Stroke Astrocytoma/glioma 12 (100.0) 25 (86.2) 16 (45.7) 2 (100.0) 55 PNET/medulloblastoma 0 (0.0) 2 (6.9) 13 (37.1) 0 (0.0) 15 Ependymoma 0 (0.0) 0 (0.0) 4 (11.4) 0 (0.0) 4 Other CNS 0 (0.0) 2 (6.9) 2 (5.7) 0 (0.0) 4 Total 12 (100) 29 (100) 35 (100) 2 (100) 78 Blood clots Astrocytoma/glioma 17 (89.5) 18 (62.1) 11 (44.0) 0 (0.0) 46 PNET/medulloblastoma 1 (5.3) 7 (24.1) 7 (28.0) 1 (100.0) 16 Ependymoma 0 (0.0) 2 (6.9) 3 (12.0) 0 (0.0) 5 Other CNS 1 (5.3) 2 (6.9) 4 (16.0) 0 (0.0) 7 Total 19 (100) 29 (100) 25 (100) 1 (100) 74 Angina-like symptoms b Astrocytoma/glioma 22 (88.0) 38 (77.6) 16 (31.4) 3 (75.0) 79 PNET/medulloblastoma 1 (4.0) 9 (18.4) 25 (49.0) 1 (25.0) 36 Ependymoma 0 (0.0) 0 (0.0) 7 (13.7) 0 (0.0) 7 Other CNS 2 (8.0) 2 (4.1) 3 (5.9) 0 (0.0) 7 Total 25 (100) 49 (100) 51 (100) 4 (100) 129 S: surgery only; S R: surgery and radiotherapy; S R C: surgery, radiotherapy, and chemotherapy, PNET: primitive neuroectodermal tumor; CNS: central nervous system. a Self-reported condition that occurred after diagnosis of the brain tumor. b Exercise-induced chest pain, shortness of breath, or irregular heart beats.

6 668 CANCER February 1, 2003 / Volume 97 / Number 3 TABLE 3 Distribution of Maximum Radiation Doses among Those Who Received Surgery and Radiation (n 682) Versus Surgery, Radiation, and Chemotherapy (n 446) Radiation dose (cgy) No. S R (%) No. S R C (%) Brain 4000 a 50 (7.3) 19 (4.3) (16.9) 64 (14.4) (44.7) 201 (45.1) (18.5) 119 (26.7) Dose unknown 86 (12.6) 43 (9.6) P Spinal 1000 b 427 (62.6) 193 (43.3) (11.7) 108 (24.2) (13.1) 102 (22.9) Dose unknown 86 (12.6) 43 (9.6) P Thyroid 150 b 354 (51.9) 155 (34.8) (16.1) 74 (16.6) (19.4) 174 (39.0) Dose unknown 86 (12.6) 43 (9.6) P S R: surgery and radiation; S R C: surgery, radiation, and chemotherapy; cgy: centigrays. a Includes indirect radiation to the brain. b Includes scatter doses. occurred too infrequently to include in our statistical analysis were diabetes (n 17), hyperthyroidism (n 20), thyroid nodules (n 17), thyroid enlargement (n 12), and low sperm count (n 20 males). Overall, 43% of cases (37% of males and 50% of females) reported one or more endocrine-related medical conditions. One endocrine-related condition only was reported by 18% of cases, two conditions by 12%, three conditions by 7%, and the remaining 5% of cases reported four or more conditions. Case sibling comparisons Cases were considerably more likely than siblings to report endocrine outcomes (Table 4). Growth hormone deficiency was reported by 21% of cases. The incidence rate among cases ranged from 9.6 to 26.9 per 1000 person-years, depending on the time period of first occurrence. Relative to the sibling comparison group, cases had more than a 270-fold elevation in risk for a growth hormone deficiency to occur as a late effect, i.e., 5 or more years after diagnosis. As expected, risk elevations for receiving growth hormone injections were similar to the risk elevations for growth hormone deficiency. The risk for hypothyroidism, osteoporosis, and medications needed to reach puberty were also substantially elevated among cases relative to siblings. Of the 533 female cases 16 years or older, 6.4% reported never having had a menstrual period compared with 0.3% of the 1513 comparably aged female siblings (data not shown). In addition, 27.7% of female cases versus 15.3% of female siblings (16 years or older) reported taking hormones to have a menstrual period. We had no information on how many patients reached puberty before their treatment commenced. Case case treatment comparisons Endocrine late effects (Table 5) were quite rare among cases who received surgery only. In contrast, the risk for endocrine late effects was consistently higher for cases who received surgery and radiation and was higher still for cases who received chemotherapy, radiation, and surgery. The surgery only treatment group was one-fifth as likely to report a growth hormone deficiency as a late effect compared with cases in the surgery with radiation treatment group (RR 0.2, P 0.001). Conversely, the chemotherapy with radiation and surgery treatment group had an elevated risk for growth hormone deficiency relative to the surgery with radiation treatment group (RR 2.7, P 0.001). Similar patterns and magnitudes of effect are apparent for each outcome (Table 5). The higher risk from radiotherapy for any endocrine outcome to occur was not dose dependent. Compared with cases who received surgery only, the RRs for any reported endocrine outcome among cases who received brain doses of less than 5000, , and greater than or equal to 5500 cgy were 3.4 (P ), 3.8 (P ), and 2.9 (P ), respectively. We attempted to explain the higher risk within the surgery, radiation, and chemotherapy group, relative to the surgery and radiation group who did not receive chemotherapy, by exploring differences in maximum radiation dose to the brain, thyroid, or spine between the two treatment groups. When we adjusted for the differences in the radiation doses in our statistical models, however, our RR estimates remained virtually unchanged. We also attempted to evaluate chemotherapy-specific effects. However, the relatively small number of outcomes in this treatment group coupled with the wide diversity in multimodal chemotherapy regimens prevented us from isolating any single agent or combination of agents associated with a disproportionately high outcome risk. Radiation site-specific effects were also difficult to isolate because virtually all patients who received radiotherapy had exposure at some level to the brain, spine, and thyroid. We did find that the risk of hypothyroidism in cases who received a radiation dose of 2500 cgy or higher to the thyroid was more than twice that of cases who received less than 2500 cgy of radiation to the thyroid (RR 2.7,

7 Late Effects among Brain Tumor Survivors/Gurney et al. 669 TABLE 4 Endocrine-Related Outcomes by Reported Time Period of First Occurrence Outcome No. among cases Rate among cases a RR b 95% CI P value Hypothyroidism Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Growth hormone deficiency Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Received growth hormone injections Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Osteoporosis or brittle bones Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Medications needed to induce puberty Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview RR: relative risk; CI: confidence interval. a Rate per 1000 person-years. b Relative risk of occurrence adjusted for gender. The sibling rate is the reference category. P ). Similarly, although we did not have radiation dosimetry to the pituitary region specifically, the risk for growth hormone deficiency among cases who received greater than or equal to 2500 cgy to the spine was also elevated (RR 3.0, P ), relative to cases who received lower doses to the spine. Cardiovascular Outcomes The cardiovascular outcomes included in our analysis were arrhythmia, stroke, blood clots, and angina-like symptoms (exercise-induced chest pain, shortness of breath, or irregular heart beat). Conditions with too few reported outcomes to analyze were rheumatic heart disease (n 1), arteriosclerosis (n 2), congestive heart failure (n 3), myocardial infarction (n 6), coronary heart disease (n 1), angina pectoris (n 2), pericarditis (n 5), pericardial constriction (n 0), stiff or leaky valves (n 4), heart catheterization (n 5), and heart biopsy (n 1). Although not strictly in this outcome category, we were also interested in the occurrence of lung fibrosis. Only six cases of lung fibrosis were reported. Therefore, we could not evaluate lung fibrosis in any detail. One or more adverse heart or circulatory problems was reported by 18% of cases (16% of males and 19% of females). One cardiovascular outcome only was reported by 14% of cases and the remaining 3% reported two or more such outcomes. Case sibling comparisons Several cardiovascular outcomes, although rare, were higher in cases relative to siblings (Table 6). A stroke was reported by 78 cases (4.9%), including 38 that occurred 5 or more years after diagnosis. The risk for stroke as a late effect was more than 40-fold higher in cases compared with siblings. The hazard rate for angina-like symptoms as a late effect (i.e., 4.8 per 1000 person-years for cases) was twice that of the sibling comparison group. Cases were not at increased risk for arrhythmia relative to siblings. Case case treatment comparisons Stroke and blood clots had at least threefold higher risks of occurrence 5 or more years after diagnosis among cases in the chemotherapy, radiation, and surgery treatment group compared with cases in the radiation and surgery treatment group who did not receive chemotherapy (Table 7). However, the absolute frequencies of these complications were low. Because of statistical adjustments, the cardiovascular outcome differences by treatment (Table 7) presumably are independent of histology, age at diagnosis, and gender.

8 670 CANCER February 1, 2003 / Volume 97 / Number 3 TABLE 5 Endocrine-Related Outcomes by Treatment Group among Cases with Reported First Occurrence 5 or More Years after Diagnosis Outcome No. RR a 95% CI P value Hypothyroidism Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Growth hormone deficiency Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Received growth hormone injections Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Osteoporosis Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Medications needed to induce puberty Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy RR: relative risk; CI: confidence interval. a Relative risk of occurrence adjusted for gender, age at diagnosis, and histology. Consistent with the findings of endocrine-related late effects, further adjustment for maximum radiation doses to the brain, spine, or thyroid did not change the magnitude of the RRs. DISCUSSION We draw several conclusions from this unique followup study of survivors of childhood brain and CNS tumors. First, long-term survivors of CBTs are at elevated risk for a number of adverse endocrine and cardiovascular events, even many years after conclusion of treatment. Fifty percent of the long-term CBT survivors in our study reported one or more adverse effects related to either the endocrine or cardiovascular system. Second, for reasons presumably independent of gender, histology, age at diagnosis, and radiation dose, chemotherapy consistently increased the risk of adverse late effects beyond the late effects caused by radiation. Our efforts to understand this difference with available study data were not fruitful. Conversely, long-term CBT survivors who were treated with surgery only, almost all of whom had astrocytomas/glioma tumors, were at very low risk of late effects. We restricted our case group to individuals who survived at least 5 years. Therefore, the hazard rates within the two early time periods, i.e., diagnosis period to end of treatment and end of treatment to 5 years postdiagnosis, likely underestimate the true incidence among all children diagnosed with brain tumors. This conservative bias occurs because we could not include outcomes for cases who died within the first 5 years and therefore were not eligible for the study. We believe it is important to report outcome rates in the two early time periods, despite the potential underestimations, because some outcomes occur acutely from the tumor or treatment and result in permanent or long-term consequences that should not be disregarded due to our study design, which included only 5-year or more survivors. For those medical conditions unrelated to case fatality, the risk estimates we provide are likely unbiased. Another limitation of the study is our reliance on self-reported incidence and time of first occurrence for medical conditions. Consistent with a smaller previous study, 4 growth hormone deficiencies and hypothyroidism were frequent late effects in our study population. In a series of 144 CBT survivors, Livesey et al. 12 found that 97% had some evidence of growth hormone deficiency at a median follow-up of 9.6 years after treatment. In a series of 32 medulloblastoma patients who had not reached final growth before diagnosis, 24 (71%) exhibited abnormal growth rates and 14 of these patients showed abnormal growth hormone by pharmacologic evaluation. 13 Growth hormone deficiency and hypothyroidism are often subclinical conditions. 14,15 Our study relied on selfreport of a diagnosis, rather than on laboratory assessment as did previous evaluations Consequently, our results may have underestimated the rates of hypothyroidism and growth hormone deficiency. It is also possible that surveillance for conditions such as hypothyroidism, growth hormone deficiency, and osteoporosis was greater for cases than for siblings, resulting in slightly elevated RRs due to differential underdiagnosis in the sibling comparison group. The functions regulated by the growth and thyroid hormones are particularly important in the growing child. Diagnosis and treatment of hypothyroidism, even when subclinical, is required to optimize growth, cognition, and progression to puberty. Similarly, timely treatment of growth hormone deficiency is essential to maximize linear growth. 15 Less well described in previous studies of brain tumors are the higher incidences we observed of lateonset osteoporosis, stroke, blood clots, and angina-

9 Late Effects among Brain Tumor Survivors/Gurney et al. 671 TABLE 6 Cardiovascular-Related Outcomes by Time Period of Reported First Occurrence Outcome No. among cases Rate among cases a RR b 95% CI P value Arrhythmia Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Stroke Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Blood clots Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview Angina-like symptoms c Diagnosis to end of treatment End of treatment to 5 yrs postdiagnosis Five yrs postdiagnosis to interview RR: relative risk; CI: confidence interval. a Rate per 1000 person-years. b Relative risk of occurrence adjusted for gender. The sibling rate is the reference category. c Exercise-induced chest pain, shortness of breath, or irregular heart beats. TABLE 7 Cardiovascular-Related Outcomes by Treatment Group among Cases with Reported First Occurrence 5 or More Years after Diagnosis Outcome No. RR a 95% CI P value Arrhythmia Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Stroke Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Blood clots Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy Angina-like symptoms b Surgery Surgery and radiation Reference Reference Surgery, radiation, and chemotherapy RR: relative risk; CI: confidence interval. a Relative risk of occurrence adjusted for gender, age at diagnosis, and histology. b Exercise-induced chest pain, shortness of breath, or irregular heart beats. like symptoms. Fibrosis of the pericardium, cardiomyopathy, and atherosclerotic heart disease have been reported in patients who received chest irradiation for Hodgkin disease and breast carcinoma. 16 In a study of 26 childhood brain carcinoma patients still surviving a mean of 6 years after diagnosis, Heikens et al. 17 reported elevated systolic blood pressure, elevated lowdensity lipoprotein cholesterol levels, lower high-density lipoprotein levels, and thicker carotid intimamedia thickness, compared with normal controls. These findings are important because intima-media thickness is associated with the risk of both myocardial infarction and stroke. 18 We hypothesized that spinal irradiation was associated with a higher risk of cardiovascular problems because of presumed exposure to the heart and because pediatric patients who receive spinal irradiation for malignancies are at risk for significant cardiac dysfunction. 19 Our data did not clearly reveal this relation. We found that spinal radiation doses were similar among the 273 participants with any reported cardiovascular outcome compared with 1334 participants without any such outcome (P 0.21). Likewise, it is unlikely that the cardiovascular conditions reported are a result of exposure to anthracyclines, a known cardiotoxic agent. Anthracyclines are rarely used in the treatment of CBTs because they do not effectively cross the blood-brain barrier. Only 16 patients in our study received anthracyclines, 5 of whom reported cardiovascular outcomes (3 with stroke). Adjuvant chemotherapy was associated with a higher risk of late adverse effects compared with radiation and surgery alone. Patients who received chemotherapy and radiation were exposed to higher

10 672 CANCER February 1, 2003 / Volume 97 / Number 3 doses of brain and spinal radiotherapy than did patients who received radiotherapy but not chemotherapy. Statistical adjustment for radiation dose and histologic classification did not explain the apparent additional risk associated with chemotherapy. We cannot rule out the possibility of residual confounding from these factors. We also evaluated chemotherapy agents for added toxicity within particular drug classes (e.g., alkylating agents, alkaloids, platinum-containing agents, antimetabolites, topoisomerase inhibitors, antibiotics, steroids). Treatment regimens were too heterogenous and many outcomes too infrequent to allow associations between particular agents and specific outcomes. Some studies 20,21 have shown a higher incidence of primary hypothyroidism in pediatric brain tumor survivors treated with radiation and chemotherapy compared with pediatric patients treated with radiation alone. However, Chin et al. 22 did not find this association. The CCSS cohort years of diagnosis, , spanned an era when organized clinical trials of CNS tumor treatment generally did not include chemotherapy. This is changing as the use of chemotherapy in treatment regimens for patients with CNS tumors is increasing. If chemotherapy causes higher risks of late effects, as our data suggest, children treated on current protocols may experience higher rates of long-term adverse outcomes than reported in our study. Fortunately, many long-term CBT and CNS tumor survivors will not suffer long-term hormonal or cardiovascular sequelae, especially those treated with surgery only. Nonetheless, CBT survivors are at a significantly increased risk for several potentially debilitating outcomes that are preventable or treatable. Timely growth hormone and thyroid hormone replacement are safe and effective therapies. 23 The increased risk for stroke, blood clots, and anginalike symptoms perhaps warrants aggressive conformity to a healthy lifestyle that includes a prudent diet, abstinence from smoking, regular exercise, and monitoring for hypertension and hyperlipidemia. Our data suggest that awareness of late effects is particularly important in individuals who received chemotherapy in addition to radiotherapy and surgery. Finally, lifetime medical surveillance and follow-up for potential toxicities are necessary because treatment-related complications may occur many years after therapy. REFERENCES 1. Ries LAG, Eisner MP, Kosary CL, et al. SEER cancer statistics review, Bethesda: National Cancer Institute. Available from URL: 4 Dec 2002]. 2. Gurney JG, Smith MA, Bunin GR. CNS and miscellaneous intracranial and intraspinal neoplasms. In: Ries LAG, Smith MA, Gurney JG, et al. Cancer incidence and survival among children and adolescents: United States SEER program (NIH Publication No ). Bethesda: National Cancer Institute, 1999: Strother DR, Pollack IF, Fisher PG, et al. Tumors of the central nervous system. In: Pizzo PA, Poplack DG. Principles and practice of pediatric oncology. Philadelphia: Lippincott Williams & Wilkins, 2002: Anderson DM, Rennie KM, Ziegler RS, Neglia JP, Robison LR, Gurney JG. Medical and neurocognitive late effects among survivors of childhood central nervous system tumors. Cancer. 2001;92: Mariotto A, Gigli A, Capocaccia R, et al. Complete and limited duration cancer prevalence estimates. In: Ries LAG, Eisner MP, Kosary CL, et al. SEER cancer statistics review, Bethesda: National Cancer Institute, Available from URL: 4 Dec 2002]. 6. Davis FG, Kupelian V, Freels S, McCarthy B, Surawicz T. Prevalence estimates for primary brain tumors in the United States by behavior and major histology groups. Neurooncology. 2001;3: Packer RJ, Meadows AT, Rorke LB, Goldwein JL, D Angio G. Long-term sequelae of cancer treatment on the central nervous system in childhood. Med Pediatr Oncol. 1987;15: Robison LL, Mertens AC, Boice JD, et al. Study design and cohort characteristics of the Childhood Cancer Survivor Study: a multi-institutional collaborative project. Med Pediatr Oncol. 2002;38: Rubin DB. Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons, Taylor JM, Munoz A, Bass SM, Saah AJ, Chmiel JS, Kingsley LA. Estimating the distribution of times from HIV seroconversion to AIDS using multiple imputation: multicentre AIDS cohort study. Stat Med. 1990;9: Therneau MT, Grambsch PM. Modeling survival data extending the Cox model. New York: Springer, Livesey EA, Hindmarsh PC, Brook CG, et al. Endocrine disorders following treatment of childhood brain tumours. Br J Cancer. 1990;61: Oberfield SE, Allen JC, Pollack J, New MI, Levine LS. Longterm endocrine sequelae after treatment of medulloblastoma: prospective study of growth and thyroid function. J Pediatr. 1986;108: Tomita T. Long-term effects of treatment for childhood brain tumors. Neurosurg Clin North Am. 1992;3: Sklar CA, Constine LS. Chronic neuroendocrinological sequelae of radiation therapy. Int J Radiat Oncol Biol Phys. 1995;31: Stewart JR, Fajardo LF, Gillette SM, Constine LS. Radiation injury to the heart. Int J Radiat Oncol Biol Phys. 1995;31: Heikens J, Ubbink MC, van der Pal HP, et al. Long term survivors of childhood brain cancer have an increased risk for cardiovascular disease. Cancer. 2000;88: O Leary DH, Polak JF, Kronmal RA, Manolio TA, Burke GL, Wolfson SK Jr. Carotid-artery intima and media thickness as a risk factor for myocardial infarction and stroke in older adults: Cardiovascular Health Study Collaborative Research Group. N Engl J Med. 1999;340:14 22.

11 Late Effects among Brain Tumor Survivors/Gurney et al Jakacki RI, Goldwein JW, Larsen RL, Barber G, Silber JH. Cardiac dysfunction following spinal irradiation during childhood. J Clin Oncol. 1993;11: Ogilvy-Stuart AL, Shalet SM, Gattamaneni HR. Thyroid function after treatment of brain tumors in children. J Pediatr. 1991;119: Livesey EA, Brook CG. Thyroid dysfunction after radiotherapy and chemotherapy of brain tumours. Arch Dis Child. 1989;64: Chin D, Sklar C, Donahue B, et al. Thyroid dysfunction as a late effect in survivors of pediatric medulloblastoma/primitive neuroectodermal tumors: a comparison of hyperfractionated versus conventional radiotherapy. Cancer. 1997;80: Sklar C, Whitton J, Mertens A, et al. Abnormalities of the thyroid in survivors of Hodgkin s disease: data from the Childhood Cancer Survivor Study. J Clin Endocrinol Metab. 2000;85: