Summary of comparative performance of the new PIRT score for patients with R/R MTCL and MNKCL (0-6) against conventional IPI and PIT score across the global (training and test data sets) and the combined TCP 2.0 and MGH cohorts (external validation data set)
| Prognostic index . | IPI . | PIT . | PIRT . | ||||
|---|---|---|---|---|---|---|---|
| Risk group (score) | LR (0-2) | HR (3-5) | LR (0-1) | HR (2-4) | LR (0-1) | IR (2-3) | HR (4-6) |
| Training set | |||||||
| n (%) | 191 (48) | 208 (52) | 204 (47) | 229 (53) | 14 (7) | 106 (54) | 76 (39) |
| Median OS since second-line Tx (95% CI)∗ | 2.53 (1.71-4.22) | 1.01 (0.65-1.21) | 2.62 (1.65-4.30) | 1.08 (0.87-1.30) | inf (2.31 to inf) | 3.05 (1.65-6.08) | 0.65 (0.40-1.07) |
| P value | <.005 | <.005 | <.005 | ||||
| AIC | 2437.55 | 2628.41 | 1057.53 | ||||
| Training C-index | 0.58 | 0.57 | 0.65 | ||||
| Test set | |||||||
| n (%) | 40 (43) | 54 (57) | 50 (48) | 55 (52) | 7 (13) | 27 (52) | 18 (35) |
| Median OS since second-line Tx (95% CI)∗ | 1.13 (0.77-3.03) | 0.64 (0.32-1.41) | 1.46 (0.96-3.40) | 0.59 (0.28-1.20) | inf (0.96 to inf) | 1.30 (0.54-2.36) | 0.28 (0.16-0.68) |
| Bootstrap testing, C-index average (95% CI)† | 0.56 (0.50-0.63) | 0.59 (0.52-0.65) | 0.70 (0.63-0.76) | ||||
| C-indextest, IPI < C-indextest, PIRT (paired t test P value <.001) C-indextest, PIT < C-indextest, PIRT (paired t test P value <.001) | |||||||
| External validation set | |||||||
| n (%) | 95 (65) | 52 (35) | 80 (54) | 67 (46) | 7 (5) | 60 (41) | 80 (54) |
| Median OS since second-line Tx (95% CI)∗ | 1.06 (0.66-2.35) | 0.67 (0.48-1.08) | 1.90 (0.77 to inf) | 0.65 (0.38- 0.94) | NR (0.07 to inf) | 1.59 (0.70 to inf) | 0.50 (0.23-0.83) |
| Bootstrap external validation C-index average (95% CI)† | 0.51 (0.46-0.57) | 0.55 (0.50-0.61) | 0.62 (0.57-0.67) | ||||
| C- indexexternal validation, IPI < C- indexexternal validation PIRT (paired t test P value <.001) C- indexexternal validation, PIT < C- indexexternal validation, PIRT (paired t test P value <.001) | |||||||
| Prognostic index . | IPI . | PIT . | PIRT . | ||||
|---|---|---|---|---|---|---|---|
| Risk group (score) | LR (0-2) | HR (3-5) | LR (0-1) | HR (2-4) | LR (0-1) | IR (2-3) | HR (4-6) |
| Training set | |||||||
| n (%) | 191 (48) | 208 (52) | 204 (47) | 229 (53) | 14 (7) | 106 (54) | 76 (39) |
| Median OS since second-line Tx (95% CI)∗ | 2.53 (1.71-4.22) | 1.01 (0.65-1.21) | 2.62 (1.65-4.30) | 1.08 (0.87-1.30) | inf (2.31 to inf) | 3.05 (1.65-6.08) | 0.65 (0.40-1.07) |
| P value | <.005 | <.005 | <.005 | ||||
| AIC | 2437.55 | 2628.41 | 1057.53 | ||||
| Training C-index | 0.58 | 0.57 | 0.65 | ||||
| Test set | |||||||
| n (%) | 40 (43) | 54 (57) | 50 (48) | 55 (52) | 7 (13) | 27 (52) | 18 (35) |
| Median OS since second-line Tx (95% CI)∗ | 1.13 (0.77-3.03) | 0.64 (0.32-1.41) | 1.46 (0.96-3.40) | 0.59 (0.28-1.20) | inf (0.96 to inf) | 1.30 (0.54-2.36) | 0.28 (0.16-0.68) |
| Bootstrap testing, C-index average (95% CI)† | 0.56 (0.50-0.63) | 0.59 (0.52-0.65) | 0.70 (0.63-0.76) | ||||
| C-indextest, IPI < C-indextest, PIRT (paired t test P value <.001) C-indextest, PIT < C-indextest, PIRT (paired t test P value <.001) | |||||||
| External validation set | |||||||
| n (%) | 95 (65) | 52 (35) | 80 (54) | 67 (46) | 7 (5) | 60 (41) | 80 (54) |
| Median OS since second-line Tx (95% CI)∗ | 1.06 (0.66-2.35) | 0.67 (0.48-1.08) | 1.90 (0.77 to inf) | 0.65 (0.38- 0.94) | NR (0.07 to inf) | 1.59 (0.70 to inf) | 0.50 (0.23-0.83) |
| Bootstrap external validation C-index average (95% CI)† | 0.51 (0.46-0.57) | 0.55 (0.50-0.61) | 0.62 (0.57-0.67) | ||||
| C- indexexternal validation, IPI < C- indexexternal validation PIRT (paired t test P value <.001) C- indexexternal validation, PIT < C- indexexternal validation, PIRT (paired t test P value <.001) | |||||||
PIRT score, for the number of selected adverse factors, ranged from 0 to 6. PIRT score was developed after selecting the final MVC model based on 80% (training data set) of 763 patients for whom information on the start of second-line treatment and complete data for the selected risk factors of the final MVC model was available. We tested the final model on 1000 bootstraps of 20% (test data set) of the global data set and used an external validation/independent cohort of 147 patients (T-Cell Project (TCP) 2.0 and Massachusetts General Hospital (MGH) cohorts pooled) to confirm the performance of the PIRT score. The C-index is a widely used metric for the global evaluation of prognostic models in survival analysis.
AIC, Akaike information criterion; MVC, multivariable Cox; NR, not reached; Tx, treatment.
OS estimates calculated from the start of second-line treatment to death or loss of follow-up.
Average = 1000 times bootstrapped.