We conducted several secondary analyses to examine the robustness of the detected effects. Although log transformation of the CD symptom scores has the effect of reducing the impact of outliers at the upper end of the distribution, we also conducted analyses removing the 16 males and 11 females at the upper end of the symptom count distribution (Table 1), with results largely unchanged (data not shown). In addition, results with the binary CD case status variable were consistent with the primary analysis of log-transformed CD symptom counts. Furthermore, most of the top SNPs implicated in the primary analysis showed consistent effects in the European and African-American subgroups. Accordingly, these analyses bolster our confidence in the robustness of the findings reported in this study.
