Change in tumor size vs. change in EIR rash score. This is a plot of each individual, evaluable patient from this trial. The x-axis represents the change in EIR from baseline to C1D8. The y-axis represents the change in tumor size by log ratio of tumor burden at week 8 to baseline. Each patient is further represented by which arm to which they were randomized (red for Arm A, blue for Arm B) and serum predictor (x for poor, circle for good). The plot reveals potentially informative trends: 1) the top 5 tumor responses (the most negative log ratios) were among subjects randomized to Arm B, all of whom had “good” predictor status, but had rash changes on the EIR of 0, 1, 2, 4, and 5; 2) Of the early progressors (11 subjects with log ratio > 0.18 at or before the first imaging session), 8 had some evidence of rash and 5 had among the most severe rashes with an EIR rating change of 4 or 5; 3) No patient with a “poor” serum proteome predictor had any tumor shrinkage at all, but only 2 of these subjects were randomized to receive pemetrexed at the end of the 2-week run-in.