Methodological Considerations in the Study of Leukemia among Genetically and Culturally Heterogeneous Populations.

摘要

Incidence of childhood acute lymphoblastic leukemia (ALL) differs by race and ethnicity, with lowest and highest rates observed among African American and Latino children, respectively, in the United States. Further, rates of adverse treatment outcomes are higher among both African American and Latino children. Reasons for these disparities are unclear, but may be attributable to factors such as misclassification of race and ethnicity in cancer registry records and differences in early response to treatment. Further, information regarding the genetic contribution to observed differences is limited, particularly for African Americans and Latinos. It has been suggested that genetic ancestry may serve as an alternative measure of genetic risk among such "admixed" populations, but to date, no studies have reviewed the validity of this approach.;A comparison of self-reported race and ethnicity among participants of the California Childhood Leukemia Study (CCLS) to that reported in the California Cancer Registry (CCR) found that agreement was good for Asian / Pacific Islander (API), Black, Hispanic/Latino, and White CCLS participants (kappa > 0.7), but was comparatively poor for Native Americans and multiracial participants (kappa <0.2). The main predictor for ethnic misclassification identified was a difference in race and/or ethnicity between the child's parents. Other contributing factors include the father's education and whether the mother was born outside of the United States. Though the rate of childhood ALL prior to adjustment for misclassification was significantly higher among Hispanics/Latinos compared to non-Hispanic whites (risk ratio = 1.21 95%CI = 1.09, 1.36), the ratio after adjustment was even greater (RR = 1.57, 95%CI = 1.40, 1.75) due to the prevailing directionality of misclassification toward non-Hispanic White in the CCR.;Baseline characteristics and response to induction therapy among Latino, non-Hispanic White, and non-Hispanic non-White CCLS participants were assessed. Clinical characteristics at diagnosis did not differ by ethnicity, with the exception of risk level and age at diagnosis, which were both lower among non-Hispanic Whites (p-values = 0.05 and 0.04, respectively), and platelet level which was higher among non-Hispanic Whites (p-value = 0.04). Overall response to treatment, proportion responding at days 8 and 15, and blast counts among those not responding at day 8 did not differ by ethnicity for pre-B cell ALL or for any subgroups. For all ethnic groups combined, response to treatment at day 8 was greater among participants with the ETV6-RUNX1 translocation (difference = 7.6%, 95% CI = -2.6 to 17.0%), and lower among those with high hyperdiploidy (difference = -15.0 %, 95% CI = -27.0 to -3.6%), similar to the findings of previous studies. Response by day 8 was substantially higher for high-risk non-Hispanic Whites compared to Hispanics/Latinos (difference = 25.0 %, 95% CI = -1.4 to 50.0 %).;Finally, a review of 39 studies using genetic ancestry markers identified through the Pubmed database found no trends in error of ancestry estimation. Substantial confounding of associations between genetic ancestry and disease outcomes by both self-reported race, as well as factors that are associated with race, including socioeconomic status, racial discrimination, and place of residence, were observed. The findings of this review suggest that the associations between genetic ancestry and disease outcomes observed in studies that utilize genetic ancestry as a proxy for continental differences in genetic risk may be confounded by social or environmental factors. As a result, genetic contributions to observed differences in incidence of disease and frequency of health measures by race and ethnicity may be overestimated in such studies.