Race and Subset Analyses in Clinical Trials: Time to Get Serious About Data Integration

Abstract

Colorectal cancer mortality rates have decreased in the last two decades primarily because of improved population screening rates and enhanced multimodality treatment of the disease (1). Because the decline in mortality rates among African American patients has not been as substantial as in white patients of European descent, the disparity in stage-specific survival between African Americans and whites continues to widen (2,3). Multiple factors contribute to the lower survival of African American patients, including reduced access to quality primary care that can lead to less preventative care, higher rates of comorbidities, diagnosis at later stages of disease, and lower utilization of potentially curative treatment after diagnosis (4–7). Unfortunately, there has been a paucity of integrated data at both the individual and population levels to develop effective policies to close the gap. In this issue of the Journal, Yothers et al. (8) sought to determine whether black (African American or of African ancestry) colon cancer patients with stage II and III disease treated in an equivalent manner within the setting of clinical trials achieved similar cancer-specific and overall outcomes compared with white patients. The authors examined the overall survival, recurrencefree survival, and recurrence-free interval of 13 393 whites and 1218 blacks from the Adjuvant Colon Cancer ENdpoinTs (ACCENT) collaborative group database. The strength of this study was the uniform collection of individual patient level data from participants in 12 adjuvant phase III clinical trials in North America from 1977 to 2002. The results of the study showed that under standardized treatment conditions, black patients still had statistically significant worse overall and recurrence-free survival compared with white patients, whereas both patient groups had a similar recurrence-free interval. Based upon these findings, the authors conclude that the worse survival of black patients is likely because of factors unrelated to response to adjuvant treatment, such as racial differences in comorbidities, general life expectancy, or potentially in care for recurrent disease. It is worth noting that despite the addition of a more contemporary cohort of patients, the results of this study are almost identical to those published in this Journal by Dignam et al. (9) more than a decade ago, which included only the first five National Surgical Adjuvant Breast and Bowel Project (NSABP) colon cancer trials (C-01 through C-05). Moreover, data within this study precede the adoption of biologic therapies for the treatment of advanced colorectal cancer and include outcome data from 5-fluorouracil (FU)-based therapies, which were broadly applied to patients without consideration of the molecular mechanisms driving disease progression. We now understand that colorectal cancers are of a heterogeneous nature (10–12) and should be treated as such. Therefore, studying the poorer overall and recurrence-free survival of African Americans compared with whites among a uniformly treated population without reference to the underlying heterogeneity in tumor biology limits interpretation of the study findings. Nevertheless, it is useful to examine what additive information is gleaned from each of the endpoints in this article (8). This study demonstrates a non-statistically significant 8% increase in the haz ard of recurrence for blacks compared with whites (hazard ratio = 1.08, 95% confidence interval = 0.97 to 1.19, P = .15) (8). The authors conclude that this supports either no difference in response to therapy or at best a very small difference. This conclusion has some caveats. Given the higher competing causes of death for black patients before potential recurrence (as evidenced by the poorer recurrence-free survival), recurrences that may have manifested will never occur, which biases the results toward the null hypothesis. More importantly, despite a very large sample size, the study still only included 1218 black patients, and of these, a little less than one-third had recurrence events. Given the heterogeneity in both tumor biology of colon cancer [eg, microsatellite instability