Inferior survival among Aboriginal children with cancer in Ontario

Authors


  • We thank Dr. Loraine Marrett and Diane Nishri (Population Studies and Surveillance, Cancer Care Ontario) for methodological assistance and Alethea Kewayosh and Usman Aslam (Aboriginal Cancer Control Unit, Cancer Care Ontario) for aid with protocol review and knowledge translation.

Abstract

BACKGROUND

Pediatric cancer distribution and outcomes have not been examined in Canadian Aboriginal children. The objective of this study was to describe the distribution, event-free survival, and overall survival of Aboriginal children with malignancies who reside in Ontario compared with non-Aboriginal children.

METHODS

This population-based study included 10,520 Ontario children (aged <18 years) who were diagnosed with cancer between 1985 and 2011. Patients were identified from the Pediatric Oncology Group of Ontario Networked Information System database. Aboriginal children were identified by self-reported ethnicity or postal code on a Native reserve at diagnosis. Descriptive statistics of the patients were presented and compared using the Fisher exact test. Event-free and overall survival probabilities were calculated for Aboriginal and non-Aboriginal children, described using Kaplan-Meier curves, and compared using log-rank tests.

RESULTS

In total, 65 Aboriginal children and 10,364 non-Aboriginal children with malignancy were identified. The distribution of malignancy type was similar between the 2 groups. There were no significant differences in baseline characteristics, presence of metastatic disease, or treatment approach (clinical trial, standard of care, or individualized protocol) between the groups. The 5e-year event-free survival rate (± standard error) was 56.3% ± 6.2% among Aboriginal children versus 72.8% ± 0.4% among non-Aboriginal children (P = .0042), and the 5-year overall survival rate was 64% ± 6.0% versus 79.3 ± 0.4% (P = .0017), respectively. The cause of death did not vary according to Aboriginal ethnicity.

CONCLUSIONS

Survival was significantly inferior among Aboriginal children who had cancer compared with non-Aboriginal children who had cancer in Ontario. Future studies are required to define the etiology of this disparity, evaluate the issue nationally, and create interventions to improve outcomes for Aboriginal children. Cancer 2014;120:2751–2759. © 2014 American Cancer Society.

INTRODUCTION

In Canada, Aboriginal peoples comprise those of First Nations, Inuit, or Métis heritage.[1] Several studies have described poor health outcomes among Aboriginal Canadians, and Aboriginal children in particular have higher rates of infant mortality, infections (respiratory tract infections and hepatitis A), chronic illnesses (diabetes), and injuries compared with non-Aboriginal children.[2-9] Although multifactorial, the reasons behind these poorer health outcomes for Aboriginal children are thought to be grounded in higher rates of poverty, household food insecurity, and substandard housing as well as over-representation in the child welfare and youth criminal justice systems.[2, 9]

In the United States, Native American (American Indian and Alaskan Native) children share many of the same characteristics and challenges as Canadian Aboriginal children. Demographically, these populations have in common similar ethnic backgrounds,[10] a proportion of their population living on self-governed lands (Native reserve/American Indian reservation) and a higher than average percentage of their population living below the poverty line.[11] In contrast, Native Americans make up a smaller proportion of the general population than Aboriginal Canadians (1% vs 4%)[12] and have higher rates of high school graduation (percentage gap compared with the general population, 28% vs 9.5%).[13] From a child health perspective, Native American children, like Canadian Aboriginal children, have higher rates of infant mortality, infections, and injuries than the general population.[11] Native American children with cancer from the United States have generally worse survival compared with non-Native American children.[14-21] This association has been observed most clearly in acute lymphoblastic leukemia (ALL)[16, 18, 19] and neuroblastoma.[15] Proposed reasons include advanced presentation at diagnosis[15, 18] and poorer adherence to therapy.[18] Genomic differences in susceptibility and response to therapy may also play a role, as recently demonstrated in ALL.[22-24] However, some studies have reported the persistence of these survival differences when adjusting for known risk factors[16, 19] and have suggested that the issue may be a combination of biologic factors and socioeconomic lack of access to care. Isolating these factors to examine their relative contribution is difficult in the US health system.

In contrast to the Native American literature from the United States, cancer incidence and outcomes have not been explored in Canadian Aboriginal children. There are some important systemic differences between Canada and the United States. All children with cancer in Canada have universal access to health care. Furthermore, the Pediatric Oncology Group of Ontario (POGO), which was established in 1983, ensures equitable access to care across the province and provides standardized guidelines for cancer treatment and management. Care is provided by 5 pediatric oncology centers, 7 satellite centers, and the Pediatric Interlink Community Cancer Nursing Program, which provides single point access to care in the community for every child with a malignancy in the province. Thus, challenges with access to care described in the United States should not be a major factor in Ontario.

Given the description of generally poor health outcomes for Canadian Aboriginal children, there is an urgent need for studies of cancer incidence and outcomes in this group. Conduct of the current study in Ontario allowed for evaluation within a system in which access to care and ability to pay for health care costs should not be issues and allowed us to explore some of the reasons proposed to explain disparity in survival in the US literature. Our objectives were to describe the distribution, event-free survival (EFS), and overall survival (OS) of Aboriginal children with malignancies who reside in Ontario compared with non-Aboriginal children.

MATERIALS AND METHODS

Population and Setting

We examined all malignancies among Aboriginal children in Ontario compared with non-Aboriginal children. We included children with de novo cancer diagnosed between January 1, 1985 and January 1, 2011 who were aged <18 years at diagnosis.

Data Source

The data source was the POGO Networked Information Systems (POGONIS) database. POGONIS is a population-based, active registry that has been collecting information prospectively on all cases of pediatric cancer diagnosed and treated in the province of Ontario since 1985. Children who are diagnosed and treated at 1 of the 5 tertiary pediatric oncology centers located in the province are included in the database. The 5 tertiary centers are The Children's Hospital of Western Ontario (London), McMaster Children's Hospital (Hamilton), The Hospital for Sick Children (Toronto), The Kingston General Hospital (Kingston), and The Children's Hospital of Eastern Ontario (Ottawa). POGONIS captures from 96% to 98% of children ages birth to 14 years who are diagnosed with cancer in Ontario compared with the Ontario Cancer Registry, although the capture rate (46%) is much poorer for adolescents ages 15 to 17 years, largely because of referral patterns of adolescent and young adults, who attend adult treatment centers outside the POGO network.[25] POGONIS contains detailed information on each patient registered, including demographics (including postal code at the time of diagnosis), timing of the cancer presentation, diagnostic details (such as staging), treatment information, and outcomes, including relapse and death. Death information contained in POGONIS is derived from information collected at each site as well as through a linkage with the Ontario Cancer Registry maintained by Cancer Care Ontario, which obtains quarterly death information from the Registrar General and from the Statistics Canada Vital Statistics Death Database, ensuring virtually complete death data collection at the national level and in the Ontario Cancer Registry-linked data in POGONIS. It is noteworthy that POGONIS also captures the ethnic background of each patient as determined by the local data managers based on self-reported ethnicity documented in the patient's hospital chart. Research ethics approval for this study was obtained from The Hospital for Sick Children, Toronto, Ontario, Canada.

Outcome, Exposure, and Covariate Variables

Our primary outcomes were EFS and OS. EFS was defined as the time from diagnosis to either the first event (relapse, second malignancy, or death, whichever occurred first) or the end of the study period (January 1, 2011). OS was calculated as the time from diagnosis to either death or the end of the study period (January 1, 2011).

Aboriginal status was identified using 2 approaches, namely, by self-reported ethnicity and by postal code of residence at diagnosis. First, we included any child whose reported ethnicity was categorized as Aboriginal (First Nations, Métis, Inuit, or North American Indian) in the POGONIS database. Second, we used a previously developed approach that used postal code to describe Aboriginal status in adult cancer.[26, 27] Based on postal code, patients were classified as Aboriginal if the postal code at diagnosis was associated with >95% of inhabitants residing on an Aboriginal reserve. Patients were classified as potentially Aboriginal if the postal code at diagnosis was associated with 1% to 95% of residents living on an Aboriginal reserve. Patients were classified as not Aboriginal if the postal code at diagnosis was not associated with inhabitants living on a reserve. The overall Aboriginal classification considered the child Aboriginal if either the ethnicity or the postal code algorithm classified the child as Aboriginal. Children were classified as not Aboriginal if both the ethnicity and the postal code information indicated that the child was not Aboriginal. Children who were considered potentially Aboriginal by postal code and not Aboriginal by ethnicity were excluded from the primary analysis but were described.

Malignancy of all types were included and were categorized according to the International Classification of Childhood Cancer, third edition.[28] Potential covariates included sex, age at diagnosis (<10 years vs ≥10 years), body mass index (BMI) percentile below or above the fifth to 95th percentiles for age and sex according to the Centers for Disease Control and Prevention,[29] diagnostic era (1985-1998 or 1999-2011), the presence of metastatic disease, treatment approach (clinical trial, standard of care, or individualized plan), and times to diagnosis and treatment. Times to diagnosis and treatment were calculated from the first encounter with a health care professional and from the first visit to the tertiary care pediatric oncology center, respectively. Causes of death were categorized as progressive cancer (including unresponsive disease, relapsed disease, second malignancy), treatment-related causes (infection, hemorrhage, organ failure), other causes, and unknown cause of death.

Statistical Analysis

The characteristics of Aboriginal and non-Aboriginal children were presented with descriptive statistics and were compared using the Fisher exact test. EFS and OS were calculated for each International Classification of Childhood Cancer disease category, for ALL and acute myeloid leukemia (AML) subtypes, and for a composite solid tumor group that included all nonleukemia or lymphoma malignancies. Survival curves were described using the Kaplan-Meier method, and Aboriginal and non-Aboriginal groups were compared using the log-rank test. Causes of death were presented with descriptive statistics and were compared using the Fisher exact test.

Sensitivity analyses were performed on the overall EFS and OS data. The children who were identified as potentially Aboriginal by postal code were included with the Aboriginal children and then subsequently with the non-Aboriginal children to determine the impact of misidentifying those children who were potentially Aboriginal on the significance of the survival difference observed. A second sensitivity analysis restricted the age of the children included to those aged <15 years to determine the effect of poorer capture of those ages 15 to 17 years.

All statistical analyses were performed using the SAS statistical program (SAS-PC, version 9.2; SAS Institute, Cary, NC). All tests of significance were 2-sided, and statistical significance was defined as P < .05.

RESULTS

During the study timeframe, 10,520 Ontario children with malignancies were identified. Of these, 65 were determined to be Aboriginal, 10,364 were determined to be non-Aboriginal, and 91 were determined to be potentially Aboriginal. Of the 65 children who were identified as Aboriginal, 63 were identified by self-reported ethnicity, demonstrating a sensitivity of 96.9% (95% confidence interval [CI], 88.4%-99.5%) using the combined criteria as the gold standard. Postal code data, however, only identified 19 Aboriginal children, demonstrating a sensitivity of 29.2% (95% CI, 18.9%-42%). There were 17 children identified by stated ethnicity as Aboriginal whose postal code was in the 1% to 95% on-reserve group.

Sex, age distribution, BMI percentile, and diagnostic era did not differ by Aboriginal ethnicity (Table 1). There were no significant differences in malignancy distribution, metastatic disease, or treatment plan between the Aboriginal and non-Aboriginal children. Times to diagnosis and treatment also were similar between both groups, although these data were missing for many of the children. The potentially Aboriginal children were generally no different from the Aboriginal and non-Aboriginal children with regard to these measures (Supporting Table 1) (see online supporting information), with the exception that there were more potentially Aboriginal children diagnosed in the earlier period.

Table 1. Characteristics of the Study Population of 10,429 Ontario Children With Malignancies by Aboriginal Status, 1985 to 2011
 No. of Patients (%) 
CharacteristicAboriginal, N = 65Non-Aboriginal, N = 10,365Pa
  1. Abbreviations: BMI, body mass index; IQR, interquartile range; HCP, first seen by any health care professional; IQR, interquartile range; TCC, first seen at a tertiary care center.

  2. a

    P values were determined using the Fisher exact test except as indicated otherwise.

  3. b

    These P values were determined using the Wilcoxon rank-sum test.

  4. c

    The number of incident cases was suppressed because of privacy concerns (n < 5).

  5. d

    Data were missing for 53 patients in the Aboriginal group and 7944 patients in the non-Aboriginal group.

  6. e

    Data were missing for 53 patients in the Aboriginal group and 7886 patients in the non-Aboriginal group.

  7. f

    Data were missing for 0 patients in the Aboriginal group and 222 patients in the non-Aboriginal group.

  8. g

    Data were missing for 6 patients in the Aboriginal group and 786 patients in the non-Aboriginal group.

  9. h

    These include observation, death before treatment allocation, surgery alone, radiation alone, reduction of immunosuppression, therapy at an outside center, and missing protocol data.

Males38 (58.5)5691 (54.9).62
Age at diagnosis   
Median [IQR], years5.3 [2.8-11.7]6.6 [2.8-12.5].35b
<10 years42 (64.6)6672 (64.4)1.00
≥10 years23 (35.4)3692 (35.6) 
BMI percentile at diagnosis   
<5%2 (9.1)278 (9.5)1.00
5%-95%18 (81.8)2378 (81.7)1.00
>95%2 (9.1)254 (8.7)1.00
Missing437459 
Diagnosis era   
1985-199831 (47.7)4805 (46.4).90
1999-201134 (52.3)5559 (53.6) 
Diagnosis   
Leukemia26 (40)3123 (30.1).10
Acute lymphoblastic17 (26.2)2446 (23.6).66
Acute myeloid6 (9.2)490 (4.7).13
Lymphoma7 (10.8)1682 (16.2).31
Solid tumors32 (49.2)5585 (53.6).48
Central nervous system11 (16.9)2097 (20.2).64
Neuroblastoma8 (12.3)634 (6.1).06
Retinoblastomac234 (2.3).66
Renal0 (0)499 (4.8).08
Hepaticc143 (1.4).60
Bone6 (9.23)483 (4.7).13
Soft tissuec704 (6.8)1.00
Germ cell0 (0)361 (3.5).17
Epithelial0 (0)320 (3.1).27
Other0 (0)84 (0.8)1.00
Metastatic disease8 (15.1)1873 (21.1).40
Missing121504 
Time to diagnosis and treatment: Median [IQR], days   
HCP to treatment16 [2-40]17 [5-44].30b, d
HCP to TCC2 [1-15.5]1.5 [0-12].46b, e
TCC to diagnosis2 [0-6]2 [1-7].52b, f
Diagnosis to treatment2 [0-19]3 [0-15].29b, g
Treatment plan  .21
Clinical trial15 (23.1)3201 (30.9).22
Standard of care23 (35.4)3306 (31.9).59
Individualized protocol18 (27.7)2087 (20.1).16
Otherh3 (4.6)872 (8.4).37

Table 2 illustrates the comparison of EFS and OS by Aboriginal status. Significantly inferior EFS and OS rates were observed for Aboriginal children when all cancers were combined (Figs. 1, 2). More specifically, the 5-year EFS rate (± standard error) was 56.3% ± 6.2% among Aboriginal children versus 72.8% ± 0.4% among non-Aboriginal children, whereas the 5-year OS rate was 64% ± 6% versus 79.3% ± 0.4% among Aboriginal and non-Aboriginal children, respectively. This significant difference in survival was maintained in the sensitivity analyses (Supporting Tables 2 and 3) (see online supporting information). When considering single diseases, including ALL and AML separately, there was no single disease category with a statistically significant reduced EFS or OS among Aboriginal children. However, for each disease group in which there were at least 10 Aboriginal children (ALL, leukemia, solid tumor, central nervous system tumor), survival was always qualitatively worse for Aboriginal children (Table 2). There were no patients with multiple primary diagnoses among the Aboriginal children, but there were 224 such patients (2.14%) in the non-Aboriginal population. Qualitatively, the survival among the potentially Aboriginal children was generally between that of the Aboriginal and non-Aboriginal groups, although the difference rarely reached statistical significance (Supporting Table 4). The cause of death did not differ significantly between the Aboriginal and non-Aboriginal children (Table 3), with the potentially Aboriginal children qualitatively distributed between these groups (Supporting Table 5) (see online supporting information). Notably, cause of death was unknown in approximately 33% of Aboriginal children and in just over 20% of non-Aboriginal children (Table 3).

Table 2. Five-Year Event-Free and Overall Survival Probabilities for Ontario Children With Malignancies by Aboriginal Status, 1985 to 2011
 EFS: Mean ± SE, %OS: Mean ± SE, %
Cancer TypeAboriginalNon-AboriginalPaAboriginalNon-AboriginalPa
  1. Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; EFS, event-free survival; CNS, central nervous system; NA, not available (because no Aboriginal children had these diagnoses); OS, overall survival; SE, standard error.

  2. a

    P values were determined using the log-rank test. Values in boldface indicate a statistically significant difference.

  3. b

    There were fewer than 10 children with this diagnosis.

All cancers56.3 ±6.272.8 ± 0.4.00464.0 ± 6.079.3 ± 0.4.002
Leukemia47.3 ± 10.273.3 ± 0.8.00654.6 ± 10.380.8 ± 0.7.004
ALL60.5 ± 12.679.9 ± 0.80.1672.8 ± 11.787.1 ± 0.7.32
AML16.7 ± 15.2b46.2 ± 2.3.1516.7 ± 15.2b57.2 ± 2.3.05
Lymphoma85.7 ± 13.2b80 ± 1.8885.7 ± 13.2b87.1 ± 0.8.38
Solid tumors56.3 ± 8.870.7 ± 0.6.09765.6 ± 8.476.1 ± 0.6.14
CNS45.5 ± 1567.1 ± 1.1.0954.5 ± 1572.5 ± 0.9.13
Neuroblastoma62.5 ± 17.1b63.8 ± 1.9.9375 ± 15.3b68.6 ± 1.9.97
Retinoblastoma100b92.9 ± 1.7.70100b97 ± 1.1.80
RenalNA81.9 ± 1.8NANA89.8 ± 1.4NA
Hepatic100b69.2 ± 3.9.54100b70.3 ± 3.9.54
Bone50 ± 20.4b58.5 ± 2.3.7750 ± 20.4b63.9 ± 2.3.51
Soft tissue50 ± 25b67.7 ± 1.8.3575 ± 21.7b72.6 ± 1.7.90
Germ cellNA86.2 ± 1.9NANA90.7 ± 1.6NA
Other epithelialNA84.7 ± 2.1NANA91.7 ± 1.6NA
OtherNA65.9 ± 5.3NANA69.6 ± 5.1NA
Figure 1.

Kaplan-Meier event-free survival was estimated for Aboriginal children compared with non-Aboriginal children during the study period from 1985 to 2011. Event-free survival was inferior for the Aboriginal children (log-rank test at 5 years, P = .004). The number of patients still at risk is indicated below the graph.

Figure 2.

Kaplan-Meier overall survival was estimated for Aboriginal children compared with non-Aboriginal children during study period from 1985 to 2011. Overall survival was inferior for the Aboriginal children (log-rank test at 5 years, P = .002). The number of patients still at risk is indicated below the graph.

Table 3. Causes of Death for Ontario Children With Malignancies by Aboriginal Status, 1985 to 2011
 No. of Patients (%) 
Cause of DeathAboriginal, N = 25Non-Aboriginal, N = 2333Pa
  1. a

    P values were determined using the Fisher exact test.

Progressive cancer14 (58.3)1495 (64.25).41
Treatment-related2 (8.4)250 (10.7)1.00
Infection0 (0)57 (2.5)1.00
Hemorrhage1 (4.2)32 (1.4).30
Organ failure0 (0)92 (4).62
Other treatment-related1 (4.2)69 (3).53
Other0 (0)80 (3.4)1.00
Unknown cause of death8 (33.3)502 (21.6).22

DISCUSSION

The current results demonstrate that, although the distribution of cancers is similar, survival is significantly inferior among Aboriginal children with malignancies in Ontario compared with non-Aboriginal children when all cancers are considered. This finding highlights a major disparity in outcome for Aboriginal children with cancer despite universal access to health care and suggests that these children should be the focus of future study and intervention to improve their outcomes.

We were not able to identify a clear etiology for the survival discrepancy. Many potential explanations for poorer outcomes that could be driven by lower socioeconomic status were not supported by our data. The observed difference in outcome was not explained by baseline characteristics, nutritional status (BMI at diagnosis), or type of malignancy. With no significant differences in the times to diagnosis and treatment or in the presence of metastatic disease at diagnosis, there was no evidence that poor outcomes were related to delays in diagnosis or treatment. Furthermore, the local health care system should reduce the impact of socioeconomic disparities. While on therapy, the Interlink Program ensures that all children have equal access to care as well as standardized treatment approaches and guidelines irrespective of rural or urban residence. Abandonment of therapy (discontinuing treatment against medical advice), in which a child with cancer has a reasonable likelihood of long-term cure and cannot consent to therapy cessation, mandates referral to the legal system in the Canadian and American contexts. In this case, the court may apprehend the child and mandate continuation of treatment.[30] Thus, abandonment of therapy cannot explain our findings.

Previous studies have suggested that children from ethnic minority groups are registered less often on clinical trials[31, 32] and thus lack the potential survival advantage of state-of-the-art therapy. However, there was no significant difference in clinical trial enrollment, provision of standard of care, or individualized treatment plans by Aboriginal ethnicity in our results. Furthermore, centralized care in Ontario to 5 pediatric centers makes a differential approach to treatment less likely.

The survival discrepancy was not limited to a single disease but, rather, appeared consistently among different diseases. This observation suggests that adherence to chemotherapy is unlikely to be a major factor, although we did not have any specific data on treatment adherence. If adherence was the primary reason for poor outcomes, then we should have observed a greater difference in survival for diseases based on at-home, self-administered oral chemotherapy (such as ALL) compared with inpatient treatment (such as AML).

Host biology is another plausible explanation for our findings. Differences in pharmacogenomics are less likely to be the primary explanation given the wide variety of chemotherapy regimens for these diverse cancers. However, other host biology features may be important. Despite similar mortality from treatment-related complications, it is unknown whether Aboriginal children experience higher rates of treatment-related toxicity, resulting in chemotherapy dose reduction. Another explanation may be related to the presence of comorbidities and whether Aboriginal children with cancer are more likely to have other chronic health conditions compared with non-Aboriginal children.

Our data build on what is known about Native American childhood cancer outcomes. We demonstrated the presence of a survival discrepancy among Canadian Aboriginal children similar to that observed in the United States. Our unique contribution, however, is that our results indicate that this disparity is not because of therapeutic approach, nutritional status (BMI), or metastatic disease at diagnosis, allowing us to speculate that it is not driven by differences in adherence or access to care in the community. Combined, these observations suggest that there are biologic features that need further delineation and that the disparity in outcomes in the United States is unlikely to be explained by access to care alone. Future studies should be broad in their conceptualization of reasons for inferior cancer outcomes for Aboriginal Canadian/Native American children, including detailed evaluation of socioeconomic, biologic, treatment selection, and treatment response variables.

We also were interested in comparing our current results with those in the adult literature. Historically, cancer incidence and mortality rates among Canadian Aboriginal adults have been lower than those among the general Canadian population.[33-36] Recent studies, however, have demonstrated that incidence and mortality rates in Canadian Aboriginal adults are now equivalent to or higher than the rates in the general population.[27, 36-40] It is believed that the worse outcomes in Aboriginal adults are caused by a combination of environmental exposures, lifestyle factors, and access to care, although the questions of genetic predisposition and response to therapy remain unanswered.[27, 36-40]

According to Ontario census data, Aboriginal children comprise 2.77% (86,665 of 3127,700) of the population aged ≤19 years.[41] However, in our study, the proportion of children with cancer who were identified as Aboriginal was only 0.62%. This suggests that our definition of Aboriginal failed to capture all Aboriginal children, and the major question is whether those identified were somehow systematically different from children who were not identified as Aboriginal. Alternatively, there may be a lower incidence of pediatric cancer across diagnoses. This hypothesis is supported by data from the US National Cancer Institute's Surveillance, Epidemiology, and End Results Program demonstrating a lower incidence of cancers among Native American children compared with other ethnic groups, although only 42% of this population is captured and, thus, the sample may not be entirely representative.[20, 42] Poorer capture of children ages 15 to 17 years in the POGONIS database also may have contributed to this finding.[25]

Our study also highlights a methodological finding. In defining Canadian Aboriginal pediatric cancer patients for future studies, postal code information should not be used alone, because it is not a sensitive method of ascertaining Aboriginal status. Conversely, chart review appears to be a sensitive approach to patient identification. However, we would recommend the combined approach for future similar studies.

Our report has several important strengths. To our knowledge, it is the first to evaluate the cancer outcomes of Canadian Aboriginal children, a population known to have generally poor health outcomes. We used a population-based database that allowed capture of virtually all children ages birth to 14 years with cancer and all deaths, regardless of location of death or enrolment on a clinical trial. This database also allowed us to examine a breadth of factors implicit in cancer outcomes. We were able to perform sensitivity analyses that demonstrated the robustness of our findings to the risk of ethnic misclassification and to poorer capture of those ages 15 to 17 years. Finally, we used a novel approach to identify Aboriginal children by combining self-reported ethnicity, as abstracted by chart review, and postal code as an indication of living on an Aboriginal reserve.

However, our report must be interpreted in light of its limitations. The number of Aboriginal children with cancer identified was small; consequently, survival analyses within specific diagnoses were underpowered. Furthermore, given the information available, we had limited ability to understand the reasons behind worse outcomes in Aboriginal children. Although there were no differences between Aboriginal and non-Aboriginal children regarding covariates, our small sample size precluded adjusting for these potential confounders in the analysis. Another limitation of our study is that there is an important difference between access to care and utilization of care. We were not able to disentangle this issue, and previous research has suggested that, even in Ontario, there may be barriers to utilization of care for Aboriginal patients.[37, 43] Future research into utilization of pediatric oncology care among minority and marginalized populations in Canada is required. Finally, our data set did not contain information regarding adherence to therapy. Although we did not observe a difference in outcome between the inpatient setting and the outpatient setting, it has been demonstrated that this is an important predictor of outcome in pediatric cancer and 1 that may vary by ethnicity[44]; thus, specifically assessing this factor in future studies will be important.

There are several important avenues for future research. A Canada-wide analysis would increase the sample size considerably and would indicate whether our findings are limited to Ontario or are generalizable outside the province. Second and most important, we need to understand the mechanisms behind worse outcomes for Aboriginal children. These studies will be facilitated by the larger sample size of a pan-Canadian or North American-wide study. Furthermore, detailed data on genomics, socioeconomic status, treatment trajectories, and decision making may shed insight into the disparity in outcome. Studies within single diseases will likely be more informative.

In summary, this first study of malignancy outcomes among Canadian Aboriginal children demonstrated that they have inferior survival compared with non-Aboriginal children. This difference in survival was not explained by demographic features, cancer type, metastatic disease at presentation, or treatment plan. Further studies are necessary to determine the etiology of this difference and to quantify the issue nationally.

FUNDING SUPPORT

Dr. Sung was supported by a New Investigator Award from the Canadian Institute of Health Research.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary