Childhood cancer patients' access to cooperative group cancer programs

A population-based study

Abstract

BACKGROUND

The Children's Oncology Group (COG), a merger of the Children's Cancer Group (CCG) and the Pediatric Oncology Group (POG), conducts clinical trials for the treatment of childhood cancer. To assess the feasibility of developing a nationwide childhood cancer registry, the authors attempted to determine whether COG could serve as a resource for identifying all children with cancer.

METHODS

A consolidated file of children age < 20 years who were diagnosed with cancer between 1992–1997 and registered with either CCG or POG was linked with records from the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) Program. Age-specific registration rates and age-adjusted registration rates (AARR) were calculated overall and by year of diagnosis, gender, race/ethnicity, stage of disease at diagnosis, and type of cancer.

RESULTS

Of 10,108 children age < 20 years with cancer who were identified by the 11 SEER registries between 1992–1997, 5796 were registered with CCG or POG. The AARR was 71% for children age < 15 years, 24% for adolescents ages 15–19 years, and 57% for children age < 20 years. Registration rates were stable over the years studied, varied by geographic region, and were found to be higher among children with more advanced disease. Registration rates were highest for children with leukemia, hepatic tumors, and renal tumors, and were lowest for carcinoma and retinoblastoma.

CONCLUSIONS

The results of the current study demonstrate that not all children with cancer are registered by the cooperative groups; however, a national registry program can be achieved by supplementing cases identified through COG with data collected by statewide population-based cancer registries. Such a partnership would be mutually beneficial, allowing COG to achieve 100% registration of children with cancer and, for the statewide cancer registries, improving the timeliness of case-finding and follow-up information for cancer outcomes. Cancer 2003;97:1339–45. © 2003 American Cancer Society.

DOI 10.1002/cncr.11192

Cancer is reported to claim the lives of more children than any other disease, despite significant improvements in its treatment and survival. The American Cancer Society estimates that 12,400 children age < 20 years were diagnosed with cancer in the U.S. in 2000 and that approximately 2300 died of the disease.1 For the majority of childhood cancers, children treated at pediatric oncology centers are reported to have a significant survival advantage compared with those treated elsewhere, and children treated on standardized protocols appear to fare better than those who are not.2–5 Given these advantages, we previously conducted a study in Los Angeles County to determine the proportion of childhood cancer patients diagnosed between 1972–1987 who were registered with the childhood cancer cooperative groups, the Children's Cancer Group (CCG) and the Pediatric Oncology Group (POG).6, 7 We linked CCG and POG records with data from the population-based Los Angeles Cancer Surveillance Program and observed substantial variation in cooperative group registration rates by age and type of cancer, but not by racial ethnic group or socioeconomic status.6, 7

Until recently, CCG and POG have separately conducted clinical trials of new therapies and modifications of existing therapeutic approaches to optimize treatment regimens for children with cancer. These groups have provided state-of-the-art centralized, multidisciplinary treatment for a majority of childhood cancer patients in the U.S. and Canada. They recently merged to form a single entity, the Children's Oncology Group (COG).

Concern regarding potential environmental causes of childhood cancer8 and the cooperative group merger have stimulated interest in the development of a nationwide childhood cancer registry. Of interest is whether COG can serve as a major resource for identifying all children with cancer in the U.S. The goals of COG are to register all children diagnosed with cancer, regardless of the availability of trial protocols and regardless of where they are diagnosed or treated, and to enroll those who are eligible and for whom appropriate consent is obtained onto clinical trials. To assess the completeness of registration of children with cancer in the U.S. by the cooperative groups, we undertook an assessment within the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) registries to determine the proportion of children diagnosed with cancer between 1992–1997 who were registered with CCG or POG.

MATERIALS AND METHODS

Eleven SEER registries in 10 regions participated in the study, including the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah, and the metropolitan areas of Atlanta (which also includes 10 rural Georgia counties), Detroit, Los Angeles County, San Francisco-Oakland Bay Area, San Jose-Monterey, and Seattle-Puget Sound. The two registries located in northern California (the San Francisco-Oakland registry and San Jose-Monterey registry) are maintained together and are referred to in the current study as the Greater San Francisco Region. The Los Angeles Cancer Surveillance Program, located at the University of Southern California (USC), served as the data coordinating and data analysis center. Table 1 lists the principal investigators and coordinators/programmers for the SEER Childhood Cancer Linkage Group.

Table 1. The SEER Childhood Cancer Linkage Group

SEER region	Principal investigator	Coordinator/programmer
SEER: Surveillance, Epidemiology, and End Results. a Dr. Pollock's current institutional affiliation: Center for Epidemiology & Biostatistics, University of Texas Health Science Center at San Antonio, San Antonio, Texas.
Atlanta	John L. Young, Jr., Dr.P.H., CTR	Kevin Ward, M.P.H.
Connecticut	Ruth McCorkle, Ph.D.	George Knafl, Ph.D.
Detroit	Richard Severson, Ph.D.	Fawn Vigneau, J.D.
	Linda Weiss, Ph.D.	Heather Crossley-May, B.A.
Greater San Francisco	Dee W. West, Ph.D.	Rita Leung, B.A.
	Christina Clarke, Ph.D.
Hawaii	Marc T. Goodman, Ph.D.	Brenda Y. Hernandez, Ph.D.
		Michael Green, C.T.R.
Iowa	Charles F. Lynch M.D., Ph.D.	Kathleen McKeen
		Diana Wagner, B.A.
Los Angeles	Leslie Bernstein, Ph.D.	Lihua Liu, Ph.D.
		Frances Wang, M.S.
New Mexico	Charles R. Key, M.D.	Ronald R. Darling, B.A.
		Linda Taylor, C.T.R.
Seattle	Beth A. Mueller, Dr.P.H.	Diane Guay, C.T.R.
		Bill O'Brien, B.A.
Utah	Charles L. Wiggins, Ph.D.
Children's Oncology Group:	Mark Krailo, Ph.D., Brad Pollock, Ph.D.a, and Gregory H. Reaman, M.D.
National Cancer Institute:	Lynn A. Ries, M.S.

A consolidated electronic file of demographic and tumor-related information was created by merging all CCG and POG registration records of cancer patients age ≤ 19 years who were diagnosed with cancer and registered with the cooperative groups in the U.S. and Canada between 1992–1997. The study was limited to invasive cancer and U.S. residents. This file included information regarding the patient's social security number (if available), birth date, date of cancer diagnosis, cancer type (a four-digit code that was translated to an International Classification of Diseases for Oncology—Second Edition [ICDO-2] code), gender, race and ethnicity, zip code of residence at the time of registration, and a numeric code representing the CCG or POG institution at which the patient was registered. To preserve patient confidentiality, the patients' names were not provided in the consolidated file. The USC coordinating center provided each SEER registry with an individual file selected by zip code of residence to assure coverage of the registry's geographic region. All CCG/POG patients with a zip code in the state of Georgia were provided to the Atlanta registry and all patients with a zip code in the state of Washington were provided to the Seattle registry. For a small portion of patients with missing or incomplete residential information (2%), the zip code of the CCG/POG institution at which the patient was registered was used.

Each registry followed a standard protocol that delineated study procedures for handling the data and linkage guidelines. Each registry was responsible for linking the consolidated file for its geographic region to the registry database. The linkage guidelines required computer-assisted matching and manual review. Efforts were made to resolve any unmatched cases in consultation with CCG or POG staff who consulted with the reporting institution at which the child was registered.

Each SEER registry submitted a file to the coordinating center after linkage activities were completed that contained the matched cases, SEER registry cases with no match in the CCG-POG consolidated file, and cases in the CCG-POG file that were not matched to the SEER registry. The coordinating center reviewed the matches and nonmatches. Demographic and tumor-related data included in the SEER registry records were provided for all patients in the SEER registry (matched and unmatched).

The coordinating center conducted all analyses individually for each SEER registry and for all SEER registries combined. Patients were classified into mutually exclusive racial-ethnic groups: Hispanics, non-Hispanic whites, non-Hispanic blacks, non-Hispanic Asians, and non-Hispanics of other racial-ethnic groups. Stage of disease at the time of diagnosis was classified according to the SEER program's Summary Staging Guide.9 A tumor was considered as localized if it was confined to the organ of origin (classified as T1 or T2), regional if it extended beyond the organ of origin (classified as T3, T4, or N1-N3), distant if it had metastasized (classified as M1), or unknown if insufficient information was available to assign a tumor stage. Cancer diagnoses (ICDO-2 codes) were categorized according to the International Classification of Childhood Cancer.10

Analyses were limited to patients diagnosed with invasive cancer. Age-specific registration rates were calculated as the proportion of matched cases out of the total number of children with cancer recorded in SEER registry records for a given age group. Age-adjusted registration rates (AARRs) were provided for both children ages birth–14 years and those ages birth–19 years. These AARRs were standardized to the overall age distribution of childhood cancers in the SEER registries for the years 1992–1997.

RESULTS

The 11 SEER registries identified 10,108 children age < 20 years diagnosed with invasive cancer between 1992–1997, of whom 5796 (AARR of 57.5%) (Table 2) were registered with CCG or POG. The overall registration rate for older adolescents (15–19 years) was 24.4% (723 of 2968). The AARR for children age < 15 years was 71.1% (5073 of 7140). The AARRs varied across the SEER regions. Registration rates appeared to be lowest in Utah and California (the Los Angeles and Greater San Francisco regions) and highest in Western Washington state (Seattle) and Hawaii.

Table 2. Rates of Registration by Cooperative Groups of Children Diagnosed with Cancer by Age Group and SEER Region, 1992–1997

SEER region	Registration Rate (%) by Age Group
	Birth–14 yrsa	15–19 yrs	Birth–19 yrsa
SEER: Surveillance, Epidemiology, and End Results a Registration rates were standardized to the age distribution of children diagnosed with cancer in all Surveillance, Epidemiology, and End Results regions between 1992–1997. b Greater San Francisco region represents the San Francisco-Oakland Bay Area registry and the San Jose-Monterey registry.
Atlanta	73.3 (317/433)	24.3 (46/189)	59.1 (363/622)
Connecticut	79.4 (501/633)	28.7 (72/251)	64.7 (573/884)
Detroit	73.8 (554/751)	29.9 (101/338)	61.0 (655/1089)
Greater San Franciscob	66.5 (745/1115)	24.8 (117/472)	54.3 (862/1587)
Hawaii	84.6 (189/223)	24.5 (24/98)	67.1 (213/321)
Iowa	79.3 (429/547)	33.3 (93/279)	65.9 (522/826)
Los Angeles	64.1 (1239/1923)	18.4 (125/678)	50.8 (1364/2601)
New Mexico	72.2 (205/289)	20.2 (26/129)	57.0 (231/418)
Seattle	84.4 (642/760)	30.8 (84/273)	68.8 (726/1033)
Utah	54.7 (252/466)	13.4 (35/261)	42.7 (287/727)
Total	71.1 (5073/7140)	24.4 (723/2968)	57.5 (5796/10108)

Table 3 shows the age-specific and age-adjusted registration rates by race/ethnicity, gender, and stage of disease at diagnosis (nonleukemia cases only). Age-specific registration rates were found to be highest among children age < 5 years (74.3%). Rates also were high among children in the 5–9 years age range (72.8%) and somewhat lower among children ages 10–14 years at the time of diagnosis (63.3%). Registration rates for older adolescents were found to be 33% those of the youngest age group. AARRs appeared to be highest among non-Hispanic white children (73.3% among children age < 15 years); at least 70% of the non-Hispanic black and Asian children age < 15 years were registered. Slightly more boys than girls were registered with the cooperative groups (72.4% of the boys age < 15 years vs. 69.6% of the girls age < 15 years). Children with localized disease were the least likely to have been registered by the cooperative groups, compared with those with more advanced disease. Among children diagnosed with cancer (excluding leukemia) before age 15 years, 58.1% of those with localized disease were registered with the cooperative groups, whereas 77.3% of those with distant disease were registered.

Table 3. Rates of Registration by Cooperative Groups of Children Diagnosed with Cancer by Age, Gender, Race/Ethnicity, and Disease Stage, 1992–1997, All SEER Regions

Characteristics
Age at diagnosis (yrs)
Birth–4	74.3 (2632/3543)
5–9	72.8 (1254/1722)
10–14	63.3 (1187/1875)
15–19	24.4 (723/2968)

	Ages Birth–14 yrsa	Ages 15–19 yrs	Ages Birth–19 yrsa
SEER: Surveillance, Epidemiology, and End Results. a Registration rates were standardized to the age distribution of children diagnosed with cancer in all Surveillance, Epidemiology, and End Results regions between 1992–1997. b Excludes leukemia cases.
Race/ethnicity
Non-Hispanic white	73.3 (2850/3896)	24.8 (453/1826)	59.2 (3303/5722)
Non-Hispanic black	70.0 (439/628)	27.7 (64/231)	57.7 (503/859)
Hispanic	67.4 (1175/1735)	22.3 (122/547)	54.2 (1297/2282)
Non-Hispanic Asian	71.8 (442/616)	26.8 (66/246)	58.6 (508/862)
Non-Hispanic other	63.9 (165/261)	15.7 (18/115)	49.8 (183/376)
Gender
Male	72.4 (2816/3886)	26.0 (408/1572)	58.9 (3224/5458)
Female	69.6 (2257/3253)	22.6 (315/1394)	55.9 (2572/4647)
Disease stage at diagnosisb
Localized	58.1 (1181/2057)	16.7 (198/1183)	46.0 (1379/3240)
Regional	68.1 (887/1313)	24.0 (168/700)	55.2 (1055/2013)
Distant	77.3 (861/1115)	29.9 (164/549)	63.5 (1025/1664)
Unknown	50.7 (159/314)	15.6 (24/154)	40.4 (183/468)

Age-specific registration rates were relatively stable between 1992–1996 (Fig. 1). The small decline observed in 1997, particularly among children age < 15 years, may be the result of a time lag between diagnosis and referral to a pediatric cancer treatment center affiliated with the cooperative groups because we required patients to be diagnosed and registered by the cooperative groups during the 1992–1997 time interval.

Figure 1. Age-specific rates of registration of childhood cancer patients by cooperative groups between 1992–1997, all Surveillance, Epidemiology, and End Results (SEER) regions.

Children with leukemia were most likely to be registered with the cooperative groups (AARRs of 84.7% for children age < 15 years and 72.9% for children age < 20 years) (Table 4). AARRs also were found to be high for hepatic tumors (81.9% for children ages birth–14 years and 69.7% for children ages birth–19 years) and renal tumors (80.1% for children ages birth–14 years and 63.6% for children ages birth–19 years). Registration rates were lowest for children with carcinomas and other malignant epithelial neoplasms (26.3% for children ages birth–14 years and 20.4% for children ages birth–19 years) and retinoblastoma (30.2% for children ages birth–14 years and 21.2% for children ages birth–19 years). As would be expected, for older adolescents ages 15–19 years, few patients were diagnosed with retinoblastoma, sympathetic nervous system tumors, hepatic tumors, and renal tumors. The registration rates were relatively high for older adolescents with malignant bone tumors (50.2%) and leukemia (44.2%), but not carcinomas (6.3%) or germ cell trophoblastic and other gonadal neoplasms (10.9%).

Table 4. Rates of Registration by Cooperative Groups of Children Diagnosed with Cancer by Tumor Type, 1992–1997, All SEER Regions

ICCC diagnostic group	Ages Birth–14 yrsa	Ages 15–19 yrs	Ages Birth–19 yrsa
SEER: Surveillance, Epidemiology, and End Results; ICCC: International Classification of Childhood Cancer10; CNS: Central nervous system. a Registration rates were standardized to the age distribution of children diagnosed with cancer in all SEER regions from 1992 through 1997.
I Leukemia	84.7 (1985/2341)	44.2 (168/380)	72.9 (2153/2721)
II Lymphomas and reticuloendothelial neoplasms	69.9 (498/683)	26.7 (188/705)	57.4 (686/1388)
III CNS and miscellaneous intracranial and intraspinal neoplasms	61.0 (900/1487)	25.2 (75/298)	50.5 (975/1785)
IV Sympathetic nervous system tumors	69.1 (429/549)	33.3 (5/15)	58.5 (434/564)
V Retinoblastoma	30.2 (95/232)	0.0 (0/1)	21.2 (95/233)
VI Renal tumors	80.1 (352/424)	23.5 (4/17)	63.6 (356/441)
VII Hepatic tumors	81.9 (79/95)	40.0 (6/15)	69.7 (85/110)
VIII Malignant bone tumors	64.6 (214/287)	50.2 (121/241)	60.5 (335/528)
IX Soft tissue sarcomas	67.4 (315/477)	30.0 (64/213)	56.5 (379/690)
X Germ cell trophoblastic and other gonadal neoplasms	47.3 (134/289)	10.9 (52/476)	36.8 (186/765)
XI Carcinomas and other malignant epithelial neoplasms	26.3 (60/252)	6.3 (37/585)	20.4 (97/837)
XII Other and unspecified malignant neoplasms	46.4 (12/24)	13.6 (3/22)	37.0 (15/46)
Total	71.1 (5073/7140)	24.4 (723/2968)	57.5 (5796/10108)

DISCUSSION

CCG and POG coordinated all large-scale clinical trials for the treatment of childhood cancers throughout the U.S. and Canada between 1972–2000. Their approach to cancer treatment is multidisciplinary, bringing together experts in many fields to provide optimal treatment strategies for each type of childhood cancer and to conduct research on the biology, etiology, and late effects of treatment of childhood cancers. Because clinical trials and biologic studies are not necessarily open for all types of childhood cancer at any given time, the cooperative groups attempt to register all new pediatric cancer patients seen at member institutions and affiliated institutions regardless of whether they are eligible for or entered onto a treatment protocol. Given the substantial geographic coverage of member and affiliated institutions, it has been proposed that the newly formed COG, a merger of CCG and POG, might serve as a resource for a nationwide childhood cancer registry. Important questions to resolve before establishing such a registry are what proportion of children with cancer would COG capture and what are the characteristics of children who are and are not registered with the cooperative groups?

Overall, between 1992–1997, 71% of the children age < 15 years who were diagnosed with invasive cancer and who resided in one of the 11 SEER registry regions were registered with CCG or POG. Among those ages 15–19 years, only 24% were registered. Rates varied by registry region and were stable over the time period studied. Registration rates did not appear to differ substantially by gender or racial-ethnic group. Rates were highest for children with leukemia, hepatic tumors, and renal tumors and were lowest for children with carcinoma and retinoblastoma. Patients with more advanced stages of cancer were more likely to be registered than those with localized disease or those for whom no staging information was available. This latter group would include children with clinical diagnoses who had short survival.

The variations in registration rates by age, stage of disease, and geographic region likely reflect patterns of registration by physicians and the availability of protocols specific for each of the types of cancer, despite the efforts of the cooperative groups to register all childhood cancer patients. The registration process requires that a principal investigator at a member or affiliated institution provide information to a clinical research associate at the institution. This information then is transferred to the cooperative group data center. Cooperative group protocols for leukemia treatment have been widely available and successful. Hence, leukemia patients are registered at a high rate. To our knowledge, few cooperative group studies to date have included carcinomas, which are diagnosed most frequently among older adolescent patients. Patients with carcinomas and older adolescent cancer patients are more likely to be treated outside the pediatric oncology center setting and therefore may be missed by the cooperative group registration system. In a few regions of the country, physicians care for pediatric cancer patients at facilities that are not associated with CCG/POG. Older adolescents (ages 15–19 years) are not admitted by some children's hospitals due to age limitations (with these institutions accepting only younger patients). The lower registration rates in older adolescents may arise from referral patterns and competition between medical and pediatric oncologists for patients in this age group.

It is possible that some of the geographic variation in rates is because of variation in the application of the standardized matching protocol at each registry. However, the identification of matches was reviewed centrally and, for matches that appeared questionable, the records were sent back to the individual registries for resolution. Although the registration rates were found to vary geographically, the patterns across all registries were consistent with regard to age, race, gender, diagnostic group, and stage of disease.

Another possible explanation for the incomplete registration by CCG/POG institutions is that a parent may have provided an address that was local when utilizing a CCG/POG facility not near their usual residence within a SEER registry region. The SEER registry might identify the patient through other means as an incident cancer case, but the CCG/POG file we provided the SEER registry would not include that patient. However, we believe that this is unlikely to have had a marked effect on our registration rates. We provided the Atlanta and Seattle registries all data for the states of Georgia and Washington, respectively. One might expect a problem would occur for regions such as Iowa (with patients going to the University of Minnesota or the Mayo clinic) or Connecticut (with patients going to New York or Boston). Yet these two registries have relatively high registration rates. In Los Angeles or the San Francisco Bay Area, in which registration rates are somewhat lower, patients have a much wider choice of health care within the regions and some facilities have pediatric hematology/oncology clinics that are not affiliated with COG. In addition, we had a small portion (2%) of patients whose zip code of residence was missing or incomplete but we were able to match > 50% of these cases (110 matched of 200 with inadequate information).

The pattern of findings for this study is similar to that previously observed for Los Angeles County between 1984–19876; registration rates decreased with increasing age, were substantially lower for older adolescents compared with other age groups, and were higher among patients with more advanced disease. Furthermore, they were highest for children with leukemia, renal tumors, and hepatic tumors and were lowest for those with carcinoma and retinoblastoma. It is interesting to note that recent Los Angeles County registration rates were found to be higher than those previously reported for the 1980s.6 For example, the registration rate for children age < 15 years increased from 60% (in 1984–1987) to 66% (in 1992–1997) among males and from 61% to 63% among females; among children age < 20 years, the registration rates increased from 47% for both males and females to 54% for males and 50% for females.

The results of the current study differ from those of a prior study by Ross et al., who predicted that 94% of children age < 15 years who were diagnosed with cancer would be registered with CCG or POG.11 In this study, the investigators combined CCG and POG data to estimate coverage by the pediatric cancer cooperative groups for some, but not all, tumor types. In combining these databases, it is unclear whether the file was purged of duplicate records and whether it was restricted to invasive cancers. We found duplicate records, both within a cooperative group and across cooperative groups, when creating the consolidated CCG/POG file we used for linkage. Unlike the current study, in which we matched the demographic and diagnostic characteristics of individual records registered by the population-based cancer registry for a geographic region with CCG/POG records, Ross et al. limited the types of cancer of interest to leukemia, lymphoma, central nervous system/brain tumors, neuroblastoma, soft tissue sarcomas, kidney tumors, bone tumors, and retinoblastoma and compared the number of observed CCG/POG cases with the expected number of cases calculated from incidence rates obtained from SEER. In their study they applied SEER rates for 1983–1987 for whites and non-whites to race, gender, and age-specific population estimates for 1990 throughout the U.S.

The COG network provides an excellent framework for timely collection of data concerning children with cancer, including data regarding demographics and diagnosis. This network also can provide timely information on treatment and outcome. However, as this study reveals, the registration of children with cancer by the cooperative groups may not be as complete as once believed.11 Older adolescent cancer patients (those ages 15–19 years) remain “underserved,” as do children with certain tumor types and those with localized cancers.

Within the COG system of member institutions, one means to improve registration would be to forge a relation between the clinical research associate who registers patients with COG and the hospital tumor registry; this would facilitate the identification of all children seen within the facility for registration. We know that certain diagnostic subgroups such as retinoblastoma and brain tumors frequently are missed. A second approach would be to approach pediatric ophthalmologists (or other specialists depending on the tumor type) through their medical subspecialty society and request cooperation to report all cases currently not being registered with COG. Urging changes in the age limitations under which certain children's hospital facilities operate does not appear to be a feasible approach. Each of these approaches requires changes in policy and practice.

Given the current environment of health care management, the most feasible approach to achieve a nationwide registry of children with cancer is to forge a partnership between COG, adult specialists, and existing statewide population-based cancer registries. Statewide population-based cancer registries all participate as members of the North American Association of Central Cancer Registries, which provides data standards and a review of completeness of case ascertainment. Statewide registries presumably are more complete than COG, but less timely, with incidence reporting lagging 18 months to 2 years behind the close of a calendar year. However, these registries can be used to supplement patient ascertainment by COG. Thus, a partnership between COG and statewide population-based cancer registries would be mutually beneficial, enhancing the coverage of COG so that COG achieves its goal of 100% registration of children with cancer, while enhancing the ability of statewide cancer registries to improve timeliness of case-finding and follow-up of cancer outcomes. This approach appears optimal in the creation of a nationwide registration system of children with cancer because this can be organized through an agreement between COG and each of the statewide cancer registries.