Date of Award
Doctor of Philosophy (PhD)
In 2020, an estimated 16,850 new cases and 1,730 cancer deaths are expected to occur in U.S. children and adolescents under 20 years of age. Childhood cancer is rare compared to cancers in adults. However, it is the most common disease-related cause of death among children and adolescents, accounting for approximately 13% of all childhood deaths in 2015. Although the overall five-year survival rate for childhood and adolescent cancers has increased to 85%, largely due to advances in treatment, supportive care, and high rates of participation in clinical trials, racial/ethnic disparities in cancer diagnosis and prognosis remain a big concern. This dissertation leverages both population- and hospital-based data to further understand the underlying causes of racial/ethnic disparities in childhood and adolescent cancer diagnosis and survival.In manuscript 1, we assessed the disparities in the improvement of survival by age at diagnosis, sex, and race for U.S. children and adolescents with cancer between 1992 and 2016. Data for primary cancers diagnosed between 1992 and 2011 in individuals ≤ 19 years old were obtained from the Surveillance, Epidemiology and End Results (SEER) 13 registries. Hazard ratios (HRs) and 95% confidence intervals (CIs) for five- and ten-year cancer-specific death associated with age at diagnosis, sex, and race were calculated using Cox proportional hazard (PH) models for patients diagnosed during 1997 to 2001, 2002 to 2006, and 2007 to 2011 (not included for ten-year analyses), compared with those diagnosed at 1992-1996. HRs and 95%CIs for every 5- and 10-year increment of diagnosis year were also calculated using Cox PH models by treating the diagnosis period as a continuous variable. Possible interactions between diagnosis period and age at diagnosis, sex, and race were assessed using likelihood ratio and Wald tests in the Cox PH models. The results show that the risk of dying from cancer in five (HRper period =0.87, 95%CI: 0.85-0.89) and ten (HRper period =0.89, 95%CI: 0.86-0.91) years steadily decreased throughout the observation period for all cancers combined and most cancer types. No difference in five- (Pinteraction=0.198) and ten-year (Pinteraction=0.364) survival improvement was found among different age groups for cancers overall and most cancers except for leukemias, myeloproliferative diseases, and myelodysplastic diseases, in which adolescents (HRper period=0.73, 95%CI: 0.68-0.79) have a greater improvement in five-year cancer-specific survival than children (HRper period =0.81, 95%CI: 0.77-0.85; Pinteraction=0.032). Sex-associated survival improvements were comparable between males and females (Pinteraction=0.312 for five-year survival; Pinteraction=0.151 for ten-year survival). There were potential racial disparities in five-year survival improvement (Pinteraction=0.027). Specifically, Whites experienced a steady reduction in mortality (HRper period =0.86, 95%CI: 0.83-0.88); Asian/Pacific Islanders had a decreased mortality first (HR1997-2001=0.91, 95%CI: 0.74-1.12; HR2002-2006=0.69, 95%CI: 0.55-0.86) and then reached a plateau (HR2007-2011=0.72, 95%CI: 0.59-0.89); while Blacks experienced no improvements first (HR1997-2001=1.07, 95%CI: 0.88-1.29; HR2002-2006=0.99, 95%CI: 0.82-1.19) but quickly caught up with other races (HR2007-2011=0.65, 95%CI: 0.53-0.79). Stratification analyses by age and cancer types also found a similar pattern among children (Pinteraction =0.031) and those with leukemia, myeloproliferative diseases, and myelodysplastic diseases (Pinteraction =0.001). For ten-year survival, no improvement was found among Blacks (HR1997-2001=1.06, 95%CI: 0.88-1.27; HR2002-2006=1.02, 95%CI: 0.85-1.22). To conclude, we observed a closing gap in historical survival improvement for adolescents, particularly those with leukemia. Additionally, although Blacks experienced a similar substantial improvement in five-year cancer-specific survival on average compared with other races, the persistent improvement gap in ten-year survival is concerning. In manuscript 2, we evaluated the effect modifications of racial/ethnic disparities in childhood and adolescent cancer survival according to health insurance coverage. Data from individuals aged ≤ 19 years with cancer diagnoses between 2004 and 2010 were obtained from the National Cancer Database (NCDB). HRs and 95%CIs for the association between race/ethnicity and overall cancer death were estimated using Cox PH regression for all cancers, by cancer type, age group and cancer metastasis status. An interaction term between race/ethnicity and health insurance status was further included in the Cox PH model to examine the racial/ethnic disparities in survival by each insurance status category. Subgroup-specific HRs and 95%CIs for different race/ethnicity groups within each insurance status were derived from the Cox PH model using linear contrasts. We observed that the hazard of death for racial and ethnic minorities was 14-43% higher compared with non-Hispanic Whites. However, the magnitude of racial and ethnic differences in overall cancer survival varies by their health insurance coverage at diagnosis or initial treatment (Pinteraction < 0.001). In the private insurance group, the hazards of death for non-Hispanic Blacks (HR=1.48, 95%CI: 1.35-1.62), Asians (HR=1.30, 95%CI: 1.13-1.50), American Indian/Alaska Natives (HR=2.00, 95%CI: 1.37-2.93) and Hispanics (HR=1.29, 95%CI: 1.18-1.41) were higher compared with non-Hispanic Whites. Among those with Medicaid, the racial/ethnic disparities in survival were slightly reduced for non-Hispanic Blacks (HR=1.30, 95%CI: 1.19-1.43) but eliminated for other racial/ethnicity minorities (HR=0.98~1.00). Among those without insurance, the hazard of death for non-Hispanic Blacks (HR=1.69, 95%CI: 1.27-2.24) and Hispanics (HR=1.25, 95%CI: 0.99-1.58) remained higher vs. Whites. When stratified by age group and cancer metastasis status at diagnosis, minority children and adolescents without distant metastasis had larger racial and ethnic disparities in cancer survival overall vs. those without metastasis across health insurance groups; similar findings were only observed among non-Hispanic Black children vs. adolescents. Overall, we observed that although the racial/ethnic survival disparities remained present among children and adolescents with private insurance, the survival gap narrowed among those enrolled in Medicaid. This finding suggests that non-Hispanic White children and adolescents derive greater benefits from private insurance than Medicaid. In manuscript 3, we estimated whether health insurance coverage mediates racial/ethnic disparities in diagnosis stage among children and adolescents with cancer. Data for primary cancers diagnosed in individuals ≤ 19 years old from 2007 to 2016 were obtained from the SEER 18 database. Odds ratios (ORs) and 95% CIs for cancer diagnosis stage associated with race/ethnicity were calculated using logistic regression. Possible interactions between diagnosis between race/ethnicity and age group were assessed using likelihood ratio and Wald tests in logistic regression. Mediation analyses assessing the total, direct, and indirect effects of health insurance coverage on racial and ethnic disparities in stage at diagnosis were further performed. In the multivariate models, racial/ethnic minority children and adolescents were more likely to be diagnosed at distant cancer stage. These associations were slightly attenuated when further adjusting for health insurance coverage status, with higher odds of 33%, 15%, and 9% for non-Hispanic Blacks (OR=1.33, 95%CI: 1.21-1.46), Asian/Pacific Islanders (OR=1.15, 95%CI: 1.03-1.28) and Hispanics (OR=1.09, 95%CI: 1.01-1.17) respectively, vs. non-Hispanic Whites. The observed racial/ethnic disparities in cancer diagnosis stage also differ by age (Pinteraction <0.001), with stronger odds among non-Hispanic Black (ORadolescents=1.52, 95%CI: 1.29-1.77; ORchildren=1.24, 95%CI: 1.10-1.38) and Hispanic (ORadolescents =1.14, 95%CI: 1.01-1.29; ORchildren=1.05, 95%CI: 0.96-1.16) adolescents but comparable risk in non-Hispanic Asian/Pacific Islanders (ORadolescents =1.11, 95%CI: 0.92-1.34; ORchildren=1.15, 95%CI: 1.00-1.32). Approximately half of the observed association with distant stage was explained by receiving Medicaid or having no insurance at diagnosis or initial treatment in Hispanic children and adolescents, 20% in non-Hispanic Blacks, and 10% in non-Hispanic Asian/Pacific Islanders. Overall, our study observed widespread racial/ethnic disparities in childhood and adolescent cancer diagnosis stage across cancer types, with racial/ethnic minorities more likely to have distant stage cancer diagnoses vs. non-Hispanic Whites. Adolescents experienced stronger disparities than children in non-Hispanic Black and Hispanic. Health insurance coverage partially explains the observed racial/ethnic disparities in cancer diagnosis stage.
Chair and Committee
Christine Ekenga, Derek Brown, Shenyang Guo, Yin Cao,
Wang, Xiaoyan, "Racial Disparities in U.S. Childhood and Adolescent Cancer Diagnosis and Survival" (2021). Arts & Sciences Electronic Theses and Dissertations. 2471.