Relationship of plasma and CSF arsenic levels in 9 patients with APL receiving oral As₂O₃. (A) Regression and correlation analysis of plasma and CSF arsenic levels, showing a linear positive correlation, with slope of the regression line at 0.177. (B,C) Patients 6 and 8, showing that CSF arsenic levels paralleled the changes of the plasma arsenic levels. The plots of CSF versus plasma arsenic levels of the remaining 7 patients are shown in Figure S1. (D) Significant variations in the CSF/plasma arsenic ratio existed in individual patients (values shown are mean ± 95% confidence interval [CI]; 1-way analysis of variance, P < .001). Analysis of the same set of data with a nonparametric test (Kruskal-Wallis test) also showed that the difference was significant (P < .001; data not shown). (E) Repeated intrathecal chemotherapy did not significantly affect the CSF/plasma arsenic ratios (values shown were mean ± 95% CI; 1-way analysis of variance, P = .92). The confidence intervals were wide for intrathecal chemotherapies 8 to 17 because the number of patients requiring 8 to 17 intrathecal treatments was small. The range of data can also be better appreciated by plotting the mean plus or minus SD (Figure S2). Analysis of the same set of data with a nonparametric test (Kruskal-Wallis test) also showed that there was no significant difference (P = .9; data not shown). (F) The CSF/plasma arsenic ratio was unaffected by the presence of blasts in the CSF (+ve indicates positive; −ve, negative; t test, P = .94).