Early CAR T-cell expansion predicted early and late treatment response and was associated with survival. (A) Median absolute dose as documented in the certificate of analysis (tisagenlecleucel [tisa-cel]) or in the prescribing information (axicabtagene ciloleucel [axi-cel]), comparing Rs (n = 28) and NRs (n = 35). (B) Median absolute CAR T-cell expansion in the peripheral blood, comparing Rs and NRs. Only the significant P values from Mann-Whitney U tests are shown. If corrected for multiple comparisons using the Holm–Šídák method (α = 0.05), only the P values for days 7 and 14 remain. Number of Rs at specific time points: day 4, n = 22; day 7, n = 24; day 14, n = 26; day 28, n = 24; day 60, n = 18; and day 90, n = 22. Number of NRs at specific time points: day 4, n = 24; day 7, n = 32; day 14, n = 32; day 28, n = 24; day 60, n = 17; and day 90, n = 16. (C-D) Absolute CAR T-cell frequencies estimated over time as median area under the curve (AUC) (C) and median peak expansion (D) between days 0 and 90 according to treatment response on day 90 in Rs compared with those in NRs. P values from Mann-Whitney U tests are shown. AUC: R, n = 28; NR, n = 33. Peak: R, n = 28; NR, n = 35. Effector-to-target ratios were estimated as the ratio of CAR T-cell peak expansion (per μl) per tumor volume (cm³). (E) Effector-to-target ratios were estimated by dividing peak expansion by tumor volume. Median effector-to-target ratios according to the treatment response on day 90 in Rs compared with that in NRs. The P value from an unpaired t test is shown. R, n = 28; NR, n = 35. (F) Association of absolute number of CAR T cells on day 7 with treatment response on day 90. The P value from a logistic regression analysis is shown. (G) Association of absolute number of CAR T cells on day 7 and response to treatment on day 90. The P value from ROC is shown. (H) Proportion of high or low CAR T-cell expansion according to the cutoff value on day 7 among Rs and NRs. Relative frequency of patients , the P value from Fisher exact test, positive predictive values (PPV) and negative predictive values (NPV) are shown. (I-K) Association of absolute number of CAR T cells on day 7 and treatment response on day 90 in the validation cohort. P values from Mann-Whitney U tests (I), logistic regression analysis (J), and ROC analysis (K) are shown. R, n = 36; NR, n = 19. (L) Proportion of high or low CAR T-cell expansion according to the cutoff value on day 7 among Rs and NRs in the validation cohort. Relative frequency of patients and P values from Fisher exact tests are shown. (M) Corrplot of baseline parameters with CAR T-cell kinetics; n = 104. The size of the Spearman correlation coefficient r is represented by circle color and size. Only the significant correlations after adjusting P values for multiple comparisons using a false-discovery approach with the 2-stage step-up method of Benjamini, Krieger, and Yekutieli (false discovery rate, 5%) are shown. (N) Multivariate analysis of clinical characteristics of advanced-stage disease and inflammation at baseline obtained on the day of lymphodepletion, as well as administered CAR T-cell product and CAR T-cell levels on day 7 with treatment response on day 90 after infusion; n = 104. A forward stepwise logistic regression was used to identify possible predictors out of the following candidate variables: number of prior therapies, ECOG, Ann Arbor stage, lactic acid dehydrogenase (LDH) level, C-reactive protein (CRP), ferritin level, CAR T-cell product, and CAR T-cell levels on day 7. The odds ratio, 95% confidence intervals, and P values from a logistic regression analysis are shown. Range of time points is provided in supplemental Figure 1.