Exposure-efficacy and exposure-safety analyses. (A) Logistic regression of 3-month response vs Cmax overlaid with observed proportions (95% CI) by quartile. The estimated OR for having a 3-month response with a twofold increase in Cmax was 1.09 (95% CI, 0.897-1.321). Kaplan-Meier plots of DOR, based on response at any time, by median Cmax (B) and by median concentration (C) at month 3. Based on the Cox regression model of DOR by log of Cmax (B), the hazard ratio for a twofold increase in Cmax was 0.79 (95% CI, 0.61-1.01). (D) Tisagenlecleucel exposure by CRS grade. Geometric mean Cmax was 3200, 6210, and 16 000 copies/µg in patients with no (n = 42), grade 1/2 (n = 37), and grade 3/4 (n = 22) CRS, respectively. Diamonds represent mean values and circles represent values outside of 1.5 × IQR. Lower and upper whiskers extend to the most extreme points within 1.5 × IQR of quartile 1 and quartile 3, respectively. (E) Logistic regression of any-grade CRS vs Cmax by qPCR, overlaid with observed proportions (95% CI) by quartile. The estimated OR for having any-grade CRS with a twofold increase in Cmax was 1.44 (95% CI, 1.159-1.794). (F) Logistic regression of any-grade neurologic events vs Cmax by qPCR, overlaid with observed proportions (95% CI) by quartile. *n/N is the number of patients with 3-month response/total number of patients in the quartile range. †Patients who had ≥1 sample with evaluable cellular kinetics data were included. NE, not estimable.