An Updated Assessment of 43,110 Patients Enrolled in the Childhood Cancer Research Network (CCRN): A Children’s Oncology Group Report

Abstract

Background:

The Childhood Cancer Research Network (CCRN) was established by the Children’s Oncology Group (COG) as a resource for epidemiologic studies of childhood cancer. Our objective was to evaluate the representativeness of CCRN and identify factors associated with enrollment.

Method:

The number of U.S. childhood cancer patients diagnosed <20 years of age enrolled in CCRN (2008-2015) was compared to expected counts, calculated from Surveillance, Epidemiology, and End Results incidence rates and U.S. Census population estimates. Observed-to-expected ratios and corresponding 95% confidence intervals (CI) were estimated across sex, race, diagnosis age, calendar year, and cancer diagnosis groups. Multivariable linear regression models were generated to evaluate the association between open COG Phase III therapeutic trials and CCRN enrollment rates.

Result:

The 43,110 cases enrolled in CCRN represented 36% of the expected childhood cancers diagnosed from 2008-2015 (N=120,118). CCRN enrollment ratios [95% CI] were highest among males (0.38 [0.37-0.38]), non-Hispanics (0.35 [0.35-0.36]), and those diagnosed at 1-4 years of age (0.50 [0.50-51]). Enrollment ratios varied by diagnosis group, with leukemia, myeloproliferative diseases, and myelodysplastic diseases (0.55 [0.54-0.55]) and renal tumors (0.55 [0.53-0.58]) having the highest enrollment. After adjusting for year of diagnosis and cancer diagnosis, there was a 3.1% [0.6-5.6%] increase in CCRN enrollment during windows of open COG therapeutic trials.

Conclusion(s):

Despite enrolling only 36% of newly-diagnosed cases, CCRN remains a valuable resource for investigators conducting childhood cancer etiology and survivorship research. The results of this study may inform efforts to improve enrollment on current and future COG nontherapeutic registry protocols.

PRECIS:

The Childhood Cancer Research Network (CCRN) was established by the Children’s Oncology Group (COG) as a childhood cancer registry protocol. This manuscript reviews more than 43,000 childhood cancer cases enrolled in CCRN with the goal of evaluating the representativeness of CCRN and identify factors associated with study enrollment.

INTRODUCTION

In the United States, more than 15,000 individuals less than 20 years of age are diagnosed with cancer annually (1, 2). Notably, the incidence of pediatric cancer has increased by 60% since the 1970s (3). At the same time, advances in curative therapy have significantly improved outcomes, whereby nearly 85% of children and adolescents diagnosed with cancer are expected to achieve long-term survival (2). While survival has improved, long-term survivors of pediatric malignancies are vulnerable to numerous chronic health conditions as a consequence of the treatment they received (4, 5). In spite of the increasing incidence and clinical significance of childhood cancer, the relatively rarity of specific malignancies poses a major challenge to conducting well-powered investigations.

The Children’s Oncology Group (COG) has demonstrated a commitment to establishing registries that can be leveraged for studies of childhood cancer etiology and outcomes. For example, COG launched the Childhood Cancer Research Network (CCRN) in December 2007 as a childhood cancer case recruitment protocol across the United States (U.S.) and Canada, which included a consent to recontact for future research. Between 2008 and 2017, more than 50,000 children treated at COG member institutions were enrolled on CCRN, providing a rich resource for epidemiologic investigations. In fact, several investigators have successfully leveraged CCRN for etiologic studies of several childhood cancers (6-12). Despite these accomplishments, however, questions remain about the overall representativeness of the resource.

In a 2014 study, Musselman et al. compared the distribution of demographic and clinical features for 18,580 participants enrolled in CCRN between 2009 and 2011 to cases captured by the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program (13). In this report, less than 50% of the expected number in individuals diagnosed with cancer before 20 years of age were enrolled in CCRN, despite early reports suggesting >95% of invited patients agreed the participate (14). The proportion of eligible patients enrolled in CCRN generally declined with increasing age and differences were observed for specific cancer types, racial and ethnic groups. A proportion of these observed differences are likely explained by disease-specific referral patterns, but it is also possible that open COG therapeutic protocols could promote enrollment in CCRN. Now that CCRN is closed and enrollment in the more comprehensive COG case recruitment Project:EveryChild protocol (http://www.projecteverychild.org/) is ongoing, there is a need to evaluate the representativeness of the expanded CCRN cohort. Moreover, evaluating the impact of therapeutic trials on enrollment in CCRN may inform the development of future nontherapeutic registry protocols, especially as this resource is utilized in studies exploring outcomes among survivors of childhood cancer. Therefore, we sought to assess the influence of demographic and clinical factors on participation patterns in CCRN and evaluate the impact of open therapeutic trials on enrollment rates for cases enrolled in CCRN between 2008 and 2015.

MATERIALS AND METHODS

All cases diagnosed at a COG member institution in the United States (US) and enrolled in CCRN between January 1, 2008 and December 31, 2015 were included in the analysis. This timeline was selected to capture cases enrolled between the first full year of operation and prior to the opening of the successor COG registry protocol, Project:EveryChild. After pediatric central institutional review board (IRB) review, IRBs at all participating COG institutions reviewed and approved the CCRN protocol (ACCRN07). The full inclusion criteria for enrollment eligibility in CCRN are included in the supplemental materials.

Observed Cases

To evaluate the representativeness of patients enrolled in CCRN, we compared the observed distribution of cases in CCRN to the expected number estimated from SEER-derived rates and US Census Bureau population counts. Patients with a newly diagnosed International Classification of Diseases for Oncology (ICD-O) histology behavior code of 2 (carcinoma in situ) or 3 (malignant) cancer before 20 years of age at a US COG member institutions were eligible for CCRN. All patients with newly diagnosed central nervous system (CNS) tumors prior to 20 years of age were eligible for CCRN regardless of behavior code. In order to be consistent with cases captured in SEER, our analysis excluded non-CNS tumors enrolled in CCRN with a histology behavior code of 2, resulting in 21 individuals with a histology behavior code of 2 being excluded from our analysis. A total of 43,110 CCRN cases were included in our analysis. Information on diagnosis, age, sex, race (White, Black, American Indian/Alaskan Navtive, Asian/Pacific Islander), ethnicity (Hispanic and non-Hispanic), and tumor histology/location were obtained for all CCRN cases.

Expected Cases

To estimate the expected number of pediatric cancer cases diagnosed between 2008 and 2015, we multiplied SEER-generated cancer rates by U.S. population estimates. Estimates of the total U.S. population were obtained from the Population Estimates Program of the U.S. Census Bureau. In collaboration with the National Center for Health Statistics, the Population Estimates Program (https://www.cdc.gov/nchs/index.htm) provides population estimates for the U.S. for each year of interest by year of age, bridged race (White, Black or African America, American Indian or Alaska Native, Asian or Pacific Islander), Hispanic origin (not Hispanic/Latino, Hispanic/Latino), and sex. Specifically, we obtained total population estimates for 2008-2009 from census counts in ‘July 1, 2000-July 2009 Revised Bridged-Race Intercensal’ data, 2010 from ‘2010 Bridged-Race’ data, and 2011-2015 from ‘Vintage 2018 Bridged-Race Postcensal’ data records. Total population counts were restricted to population <20 years of age and stratified by sex, race, ethnicity, and age for each year.

Age-adjusted incidence rates were obtained from the 21 Registries SEER database (November 2018 submission) for each year between 2008 and 2015. Expected case counts were obtained by multiplying the SEER-based incidence rates by the corresponding total population counts obtained from U.S. Census Bureau data. Expected counts were calculated for each sex, race, ethnicity, diagnosis age, International Classification of Childhood Cancer (ICCC) category (leukemia, lymphoma, CNS, peripheral nervous system, retinoblastoma, hepatic, renal, bone, soft tissue, and germ cell/gonadal), and cancer diagnosis subtypes within each ICCC category.

Statistical Analysis

Observed cases enrolled in CCRN were divided by expected case counts to obtain observed-to-expected ratios, and the corresponding 95% confidence intervals (CIs) were calculated based on the Vandenbroucke method (15). Information on some variables was incomplete or missing for a subset of CCRN participants, including race (n=1,675), ethnicity (n=1,247), and sex (n=12). Individuals missing information on these variables were excluded from models evaluating these variables. Observed-to-expected ratios were calculated both overall and by sex (male, female), race (White, Black, American Indian/Alaskan Navtive, Asian/Pacific Islander),ethnicity (Hispanic, non-Hispanic), diagnosis age (<1, 1-4, 5-9, 10-14, 15-19 years), enrollment year (2008-2015), and ICCC category. Next, we compared ICCC-specific observed-to-expected ratios across diagnosis age, race, ethnicity, and cancer diagnosis subtypes to examine differences in observed-to-expected ratios across these factors. To evaluate the impact of therapeutic trials on CCRN enrollment rates, we identified Phase III frontline therapeutic clinical trials that were actively enrolling between 2008 and 2015 on the COG website (https://childrensoncologygroup.org/). Non-cancer, non-U.S., and trials closed prior to 2008 or opened after 2015 were excluded. A complete list of selected trials is available in the supplementary materials. For each cancer subtype with an eligible trial, dates of trial opening and closing were recorded. Because SEER incidence rates are based on mid-year population estimates, and are not readily available for specific months, we compared annual observed-to-expected enrollment rates on the basis of open therapeutic trials. Trials were classified as open for a particular year during the analysis if they were open to enrollment for at least one month of the calendar year. Linear regression models were fitted with the observed-to-expected ratios treated as the dependent variable and the following independent variables: 1) the presence of open clinical trials (any/none), 2) enrollment year, and 3) cancer subtype. All calculations and analyses were conducted in R v3.6.2, with statistical significance defined at a two-sided p-value <0.05.

RESULTS

Between 2008 and 2015, 43,110 patients with childhood cancer aged <20 years of age were enrolled in the CCRN across the U.S. (Table 1). This number represented 36% [0.36-0.36] of the expected newly diagnosed childhood cancer cases (N=120,118), based on SEER-derived estimates for the same period. All participants included in the original evaluation of CCRN catchment were included in the current study unless they specifically requested removal from the protocol (13), comprising approximately 43% of the total population included in this analysis. Overall, 56% of CCRN participants were male, 79% were non-Hispanic, 79% were White, and 40% were diagnosed at <5 years old. CCRN enrollment ratios [95% CI] were higher among males (0.38 [0.37-0.38]) compared to female (0.34 [0.33-0.34]) and higher among non-Hispanics (0.35 [0.35-0.36]) compared to Hispanics (0.32 [0.32-0.33]) and American Indian/Alaska Natives (0.37 [0.33-0.41]) followed by Whites (0.34 [0.33-0.34]), Blacks (0.31 [0.30-0.32]), and Asian/Pacific Islanders (0.25 [0.24-0.27]).

Table 1.

Observed and expected counts of pediatric cancer patients diagnosed and enrolled in COG CCRN between 2008 and 2015 in the United States

Variable	Observed		Expected		Observed Expected Ratio	95% CI
	Counts	Proportion	Counts	Proportion
Sex
Male	24091	0.56	63859	0.53	0.38	0.37-0.38
Female	19019	0.44	56259	0.47	0.34	0.33-0.34
Race/Ethnicity
Non-Hispanic White	27105	0.63	71979	0.60	0.38	0.37-0.38
Non-Hispanic Black	4584	0.11	14467	0.12	0.32	0.31-0.33
Non-Hispanic American Indian/Alaska Native	229	0.01	975	0.01	0.23	0.21-0.34
Non-Hispanic Asian/Pacific Islander	1379	0.03	5497	0.05	0.25	0.24-0.26
Hispanic	8802	0.20	26777	0.22	0.33	0.32-0.34
Non-Hispanic Othera	1011	0.02	--	--	--	--
Diagnosis age, years
<1	3075	0.07	7718	0.06	0.40	0.38-0.41
1-4	14298	0.33	28347	0.24	0.50	0.50-0.51
5-9	9112	0.21	21388	0.18	0.43	0.42-0.43
10-14	8746	0.20	24877	0.21	0.35	0.34-0.36
15-19	7879	0.18	37589	0.31	0.21	0.21-0.21
Diagnosis year
2008	4664	0.11	14708	0.12	0.32	0.31-0.33
2009	6379	0.15	15030	0.13	0.42	0.41-0.43
2010	6197	0.14	14662	0.12	0.42	0.41-0.43
2011	6238	0.14	14898	0.12	0.42	0.41-0.43
2012	5877	0.14	14927	0.12	0.39	0.38-0.40
2013	5171	0.12	14890	0.12	0.35	0.34-0.36
2014	4994	0.12	15301	0.13	0.33	0.32-0.34
2015	3591	0.08	15730	0.13	0.23	0.22-0.24
ICCC group
I. Leukemias, myeloproliferative diseases, and myelodysplastic diseases	16972	0.39	31124	0.26	0.55	0.54-0.55
II. Lymphomas and reticuloendothelial neoplasms	5791	0.13	17037	0.14	0.34	0.33-0.35
III. CNS and miscellaneous intracranial and intraspinal neoplasm	5807	0.13	21095	0.18	0.28	0.27-0.28
IV. Neuroblastoma and other peripheral nervous cell tumors	2879	0.07	5576	0.05	0.52	0.50-0.54
V. Retinoblastoma	667	0.02	2201	0.02	0.30	0.28-0.33
VI. Renal tumors	2575	0.06	4648	0.04	0.55	0.53-0.58
VII. Hepatic tumors	691	0.02	1725	0.01	0.40	0.37-0.43
VIII. Malignant bone tumors	2682	0.06	6120	0.05	0.44	0.42-0.45
IX. Soft tissue and other extraosseous sarcomas	2868	0.07	8348	0.07	0.34	0.33-0.36
X. Germ cell tumors, trophoblastic tumors, and neoplasms of gonads	1390	0.03	8076	0.07	0.17	0.16-0.18
XI. Other malignant epithelial neoplasms and melanomas	665	0.02	13573	0.11	0.05	0.05-0.05
XII. Other and unspecified malignant neoplasms	123	0.00	426	0.00	0.29	0.24-0.34

Abbreviations: COG, Children’s Oncology Group; CCRN, Childhood Cancer Research Network; ICCC, international classification of childhood cancer; CNS, central nervous system; CI, confidence interval

^aSEER does not define non-Hispanic other race/ethnicity

The overall observed-to-expected CCRN enrollment ratio exceeded 40% for cases diagnosed <10 years of age (Table 1). The enrollment rates were highest for cases diagnosed between 1 and 4 years of age (50%) and declined with increasing age groups, with the lowest rates observed in cases diagnosed between 15 and 19 years of age (21%). The trend between CCRN enrollment rate and age at diagnosis was similar across ICCC cancer diagnosis groups (Supplemental Table S1), although overall enrollment rates differed between cancer diagnosis groups. For example, overall CCRN enrollment rates exceeded 50% for cases of leukemia, myeloproliferative diseases, and myelodysplastic diseases (55%); neuroblastoma and peripheral nervous cell tumors (52%); and renal tumors (55%). In contrast, germ cell tumors, trophoblastic tumors, and neoplasms of gonads and other malignant epithelial neoplasms and melanomas had observed CCRN enrollment rates of 17% and 5%, respectively. Similar to the differences observed across diagnosis age groups, racial and ethnic disparities in the observed-to-expected persisted across cancer diagnosis groups (Supplemental Table S2-3). With the exceptions of a few diagnosis groups, the highest observed-to-expected enrollment ratios were generally observed among non-Hispanic and White patients.

Observed-to-expected CCRN enrollment ratios varied by cancer diagnosis subtypes (Supplemental Table S4). For example, among the leukemia cancer group, observed-to-expected enrollment ratio for lymphoid leukemia was 59% compared to 39% for acute myeloid leukemia. Among CNS tumors, the observed-to-expected enrollment ratio for medulloblastoma was 48% compared to <40% for most other common pediatric CNS malignancies. In addition to lymphoid leukemias, diagnosis-specific enrollment rates exceeded 50% for several major diagnosis subtypes, including neuroblastoma, Wilms tumor, and Ewing sarcoma. By comparison, observed-to-expected enrollment ratios were <30% for several important diagnosis subtypes, including glioma, astrocytoma, and melanoma.

COG Phase III frontline therapeutic clinical trials were open during the entire enrollment period for the five diagnosis groups with the highest overall observed-to-expected enrollment ratios, including acute lymphoid leukemia, Wilms tumor, Ewing sarcoma, neuroblastoma, and medulloblastoma (Figure 1). Most other diagnosis groups had periods of both open and closed frontline trials between 2008 and 2015. Although overall observed-to-expected varied across the study period, the association between higher CCRN enrollment rates during periods of open COG clinical trials was consistent across all years included in the analysis (Figure 2). In multivariable linear regression models adjusting for year of diagnosis and cancer diagnosis, observed-to-expected CCRN enrollment ratios were 3.1% (95% CI=0.6-5.6%, p-value=0.02) higher during periods of open clinical trials (Table 2). The association between open clinical trials and higher enrollment ratios in CCRN remained (adjusted mean=3.5%, 95% CI=0.7-6.3%, p-value=0.01) in secondary analyses restricted to the cancer diagnoses with both open and closed COG clinical trials between 2008 and 2015.

Figure 1.

Distribution of observed-to-expected CCRN enrollment ratios across year of enrollment by presence of absence of concurrent COG therapeutic clinical trials

*Results based on number of open trials overall (n=65) and by year: 2008 (n=38), 2009 (n=43), 2010 (n=43), 2011 (n=42), 2012 (n=34), 2013 (n=28), 2014 (n=26), 2015 (n=27)

Abbreviations: CCRN, Childhood Cancer Research Network; COG, Children’s Oncology Group

Figure 2.

Cancer diagnosis specific distribution of observed-to-expected CCRN enrollment by presence of absence of concurrent COG therapeutic clinical trials

*Trials (n=65) were opened for 8 years for acute lymphoid leukemia, nephroblastoma (Wilms’ tumor), Ewing sarcoma, neuroblastoma and ganglioneuroblastoma, medulloblastoma, and retinoblastoma; 7 years for atypical teratoid/rhabdoid tumor, acute myeloid leukiemia, heptaoblastoma, and non-rhabdomyosarcoma soft tissue sarcoma; 6 years for Hodgkin lymphoma, non-Hodgkin lymphoma, ependymoma, rhabdomyosarcoma, adrenocortical carcinoma, and nasopharyngeal carcionoma; 5 years for glioma, any germ cell tumor, and melanoma; 4 years for osteosarcoma and extracranial germ cell tumor; and 2 years for primary central nervous system gerinoma.

Abbreviations: CCRN, Childhood Cancer Research Network; COG, Children’s Oncology Group

Table 2.

Association between concurrent COG therapeutic trials and observed-to-expected CCRN enrollment ratios, 2008-2015

Modela		Effect Estimate	95% CI	P-value
Model 1
	Periods of no COG trials	Ref.		0.02
	Periods of open COG trials	0.031	(0.006 – 0.056)
Model 2
	Periods of no COG trials	Ref.		0.01
	Periods of open COG trials	0.035	(0.007 – 0.063)

Abbreviations: COG, Children’s Oncology Group; CCRN, Childhood Cancer Research Network; CI, Confidence Interval

^aModel 1 includes all cancer diagnosis groups, Model 2 is restricted to cancer diagnosis groups with periods of both open and closed COG therapeutic trails. Models adjusted for year and cancer diagnosis group.

DISCUSSION

This study reports on the representativeness of CCRN as a resource for childhood cancer epidemiologic research. Overall, more than one-third of U.S. patients diagnosed with a malignancy <20 years of age between 2008 and 2015 were enrolled in CCRN. However, consistent with a previous CCRN report (13), we observed some disparities in expected-to-observed enrollment ratios for certain cancer diagnosis, age, and racial/ethnic groups. Specifically, observed-to-expected enrollment rates were highest among some of the most frequently occurring malignancies (e.g., acute lymphoblastic leukemia), non-Hispanic Whites, and patients diagnosed between one and nine years of age. Additionally, we observed that CCRN enrollment rates for many cancers was partially dependent on concurrent COG therapeutic trials, whereby open phase III trials resulted in a modest but statistically significant increase in CCRN enrollment. Ultimately, this study provides insight into factors potentially impacting the generalizability of CCRN research and may inform efforts to improve enrollment on current and future non-therapeutic registry protocols, including Project:EveryChild.

Epidemiologic studies of childhood cancer often rely on SEER and other state-based cancer registries (16-18). While these cases are representative of the underlying population, investigations are often limited by the level of information captured by these surveillance programs. For example, detailed treatment histories and refined diagnosis information may not be readily available. With an increasing awareness of the molecular heterogeneity of many childhood cancers (19, 20), etiology studies are necessary to better understand the biology of specific molecular subtypes and potentially identify novel therapeutic targets. Large-scale population-based cancer registries do not collect the biologic samples necessary to conduct molecular investigations of cancer risk and outcomes. Therefore, COG-supported registry protocols are likely the most attractive alternative for molecular epidemiologic investigations of cancer etiology and the long-term late effects of childhood cancer treatment.

The Children’s Oncology Group is a cooperative multi-site clinical trials network of more than 200 pediatric cancer treatment centers in North America, providing care for many of newly-diagnosed pediatric cancer cases in North America (21). The current analysis was restricted to U.S. cases enrolled in CCRN due to the availability of U.S. childhood cancer incidence rates in SEER. Our findings indicate that approximately 36% of eligible patients were enrolled in the non-therapeutic CCRN registry protocol. Disparities in CCRN observed-to-expected enrollment ratios generally reflect those identified in COG therapeutic protocols (22), with certain diagnosis, age, and race or ethnic minority groups experiencing lower enrollments than others. The low CCRN enrollment ratios observed for some groups likely reflect referral patterns (e.g., low-stage germ cell tumors treated with surgery by non-oncologist, uncomplicated retinoblastoma treated by ophthalmologists, adolescents referred to adult oncologists), although this information is not captured by COG member institutions or CCRN. Importantly, we identified gaps in CCRN enrollment even among groups likely treated at pediatric oncology centers. Barriers to participating in CCRN may be similar to those encountered in therapeutic clinical trials, including individual-level factors and structural barriers. It is likely that CCRN enrollment rates differ between COG member institutions. However, we did not evaluate COG site as a variable in this analysis because information on the number of potentially eligible individuals and number of eligible individuals approached for CCRN enrollment was not systematically captured at each COG institution. Additional work is needed to identify barriers to nontherapeutic research participation at specific institutions.

An early analysis of 1,990 patients eligible for CCRN found that >95% of those approached consented to the collection of personal identifiers and contact for future research (23), suggesting overall high rates of willingness to participate. Accordingly, the relatively low observed-to-expected enrollment ratios identified in this study could indicate gaps in approaching eligible patients. Some of the observed disparities in enrollment ratios likely stems from cancer-specific referral patterns, such as older patients being referred to primarily adult treatment centers (24). For other groups, it is possible that funding provided for CCRN enrollment was insufficient to incentivize CCRN enrollment at all institutions. We observed some evidence that enrollment in CCRN was at least partly dependent on concurrent enrollment on COG therapeutic trials, some of which may have required co-enrollment in CCRN and provided higher per case reimbursement to encourage patient recruitment. Of note, the five diagnosis groups with the highest observed-to-expected enrollment rates in CCRN all had concurrent therapeutic protocols open for the entire evaluation period. For diagnosis groups with periods of both open and closed COG therapeutic protocols, we observed an average CCRN enrollment increase of 3% during periods of open phase III trials. This finding indicates that factors surrounding enrollment on COG therapeutic protocols have a positive impact on institutional engagement of patients for non-therapeutic research participation.

Disparities persist across the continuum of cancer care, from access to preventive services and treatment centers to enrollment in clinical trials (25). Lack of representativeness in clinical trials may have a profound impact of treatment outcomes in the broader patient population who ultimately receive new therapies; however, the potential bias introduced by enrollment disparities in non-therapeutic registry protocols depends on the research question. Studies utilizing CCRN as a resource can obtain unbiased results, provided enrollment in CCRN was independent of the outcome and exposure of interest. Thus, for investigators interested in large-scale epidemiologic studies of childhood cancer that involve direct contact and sample collection, CCRN remains a robust resource.

As interest in identifying the molecular underpinnings of childhood cancer incidence and outcomes increases, COG provides the resources and infrastructure needed to accrue adequate numbers of rare diagnoses to develop well-powered investigations. This updated and expanded assessment of factors affecting enrollment in CCRN supports the use of this protocol as a valuable resource for epidemiologic investigations of childhood cancer. The contemporary COG-maintained registry protocol, Project:EveryChild, which began enrolling participants in October of 2015, actively collect and store clinical, demographic, epidemiologic, and biospecimens for non-therapeutic research, regardless of enrollment on COG therapeutic trials. Collectively, these resources will likely lead to improvements in our understanding of childhood cancer etiology and outcomes. In this analysis, we identified specific groups which Project:EveryChild and other future generations of COG-supported registry protocols may prioritize to improve enrollment. Identifying and enrolling participants on large-scale, multi-site registry protocols requires complex coordination across multiple levels, and barriers to participation may emerge at each step of the process. Accordingly, these barriers likely require multifaceted solutions. Potential interventions targeting individual-level awareness of research opportunities may include targeted, culturally appropriate advertising in high visibility areas within clinics or on social media. Overcoming structural barriers may require alternative solutions, such as establishing partnerships between academic centers and satellite treatment and referral centers and providing adequate staffing to approach and enroll eligible participants at these locations. Both CCRN and Project:EveryChild were supported by philanthropic funding, with a lower institutional per case reimbursement than typically provided for NCI-supported clincial trial enrollments. Federal funding to provide adequent per case reimbursement could provide a greater incentive to institutions to approach all eligible patients for enrollment. Previous studies have highlighted the potential for partnerships between COG member and nonmember institutions to improve the accrual of underrepresented populations (26). Additional research is needed to evaluate whether approaches which show promise at increasing enrollment on COG therapeutic clinical trials similarly benefit enrollment on non-therapeutic registry protocols.