Table 2.
Severe4 independently correlates with PFS and OS in both the LC and VC
| Model | N | Covariates‡ | Main predictor | HR (95% CI) | P value | iAUC |
|---|---|---|---|---|---|---|
| PFS | ||||||
| LC | ||||||
| Univariate | 577 | N/A | Severe4 | 2.15 (1.56-2.97) | <.001 | 0.535 |
| CIRS score ≥7 | 1.26 (1.01-1.57) | .0399 | 0.529 | |||
| Multivariate | 576 | High LDH, ECOG, previous tx ≥3, GCB subtype | Severe4 | 2.06 (1.48-2.86) | <.001 | 0.620 |
| CIRS score ≥7 | 1.08 (0.86-1.36) | .498 | 0.605 | |||
| Multivariate, LT | 550 | High LDH, ECOG, previous tx ≥3, GCB subtype, steroids as bridging tx and its interaction with Severe4, stratify on CART product | Severe4∗ | 2.09 (1.42-3.09) | <.001 | 0.627 |
| Severe4† | 9.32 (4.26-20.4) | <.001 | 0.722 | |||
| CIRS score ≥7 | 1.16 (0.91-1.48) | .221 | 0.624 | |||
| TFL subgroup | 118 | High LDH, ECOG, previous tx ≥3, GCB subtype | Severe4 | 1.58 (0.69-3.59) | .280 | 0.670 |
| CIRS score ≥7 | 0.67 (0.38-1.20) | .180 | 0.672 | |||
| VC main predictor is always Severe4 because CIRS score of ≥7 was not available for VC. | ||||||
| Univariate | 218 | N/A | Severe4 | 1.71 (1.15-2.52) | .008 | 0.541 |
| Multivariate | 218 | Model not fit because IPI component variables were not measured until time of CART infusion (ie, after T-cell collection) in the VC. Moreover, unlike in the LC, all patients in the VC received CART. | ||||
| Multivariate, LT | 218 | Age, high LDH, >1 extranodal site, ECOG | Severe4 | 1.85 (1.24-2.76) | .003 | 0.659 |
| TFL subgroup | 40 | Age, ECOG, previous tx ≥3, high LDH, >1 extranodal site | Severe4 | 4.20 (1.25-14.11) | .020 | 0.701 |
| OS | ||||||
| LC | ||||||
| Univariate | 577 | N/A | Severe4 | 2.18 (1.53-3.12) | <.001 | 0.538 |
| CIRS score ≥7 | 1.35 (1.05-1.75) | .022 | 0.538 | |||
| Multivariate | 576 | High LDH, previous tx ≥3, stratify on ECOG | Severe4 | 1.89 (1.32-2.71) | <.001 | 0.574 |
| CIRS score ≥7 | 1.12 (0.85-1.46) | .421 | 0.557 | |||
| Multivariate, LT | 550 | High LDH, ECOG, previous tx ≥3, steroids as bridging therapy | Severe4 | 2.08 (1.43-3.02) | <.001 | 0.639 |
| CIRS score ≥7 | 1.14 (0.86-1.51) | .364 | 0.638 | |||
| TFL subgroup | 118 | High LDH, previous tx ≥3, ECOG | Severe4 | 1.48 (0.60-3.65) | .398 | 0.690 |
| CIRS score≥7 | 0.78 (0.40-1.51) | .462 | 0.688 | |||
| VC main predictor is always Severe4 because CIRS score of ≥7 was not available for VC. | ||||||
| Univariate | 218 | N/A | Severe4§ | 1.70 (1.11-2.61) | .015 | 0.542 |
| Multivariate | 218 | Model not fit because IPI component variables were not measured until time of CART infusion (ie, after T-cell collection). Unlike in the LC, all patients in the VC received CART. | ||||
| Multivariate, LT | 218 | Age, ECOG, previous tx ≥3, high LDH, stratify on >1 extranodal site | Severe4§ | 1.70 (1.09-2.66) | .019 | 0.665 |
| TFL subgroup | 40 | ECOG, high LDH, >1 extranodal site, CART type, stratify age > 60 | Severe4 | 2.51 (0.60-10.50) | .208 | 0.715 |
ECOG performance status was included as a categorical variable, with categories: 0 vs 1 vs 2/3/4.
iAUC, integrated Area Under the incident/dynamic ROC curve (a measure of predictive accuracy or concordance); LT, left truncation time between T-cell collection and CART infusion (denoting delayed entry into the risk set) was accounted for in the model; N/A, not applicable; TFL, transformed follicular lymphoma; tx, treatment.
Patients without steroids as bridging therapy (n = 480).
Patients receiving a steroid for bridging therapy (n = 70).
Covariate set for each multivariable Cox model was chosen in the absence of any CIRS-based predictor and includes patient and disease features with univariable Cox model P value < .20 that satisfied model assumptions and was retained upon AIC-based backward elimination.
Cox model proportional hazards assumption was violated.