Abstract 1089
Background
We validated 2 models (the 7thAmerican joint committee on cancer (AJCC) and the Helsinki university central hospital (HUCH) staging) and 1 nomogram; the Padova-Mayo (PMN), for progression free (PFS) and overall survival (OS) using patient (pt) level data from the PUMMA meta-analysis.
Methods
29 prospective trials’ (1988-2015) pt data was analysed. Models were validated with cox regression analysis for survival in months (m). Concordance index (CCI) was used to test predictive value.
Results
Comparable data was available for 463 pt; see table for variables used in each system. Models were prognostic differentiating into M1a, M1b and M1c groups. Median PFS for AJCC was 4m for M1a, 3 for M1b and 2 for M1c. Median PFS for HUCH was 3.5m for M1a, 2.5 for M1b and 1 for M1c. CCI for PFS using AJCC was 0.69 (SE 0.02, 95%CI 0.65-0.73), for HUCH it was 0.79 (SE 0.02, 95%CI 0.74-0.83). Median OS for AJCC was 15m for M1a, 9 for M1b and 5 for M1c. Median OS for HUCH was 13m for M1a, 6 for M1b and 2 for M1c. CCI for OS for AJCC was 0.69 (SE 0.02, 95%CI 0.65-0.73). For HUCH it was 0.79 (SE 0.02, 95%CI 0.74-0.83). Using ECOG and LDH (available variables used in PMN) median PFS was 4m (95% CI 4-5) for normal LDH and ECOG 0, 7 (3-9) for normal LDH and ECOG > 0, 2.6 (2-3) for elevated LDH and ECOG 0 and 2.5 (2-3) for elevated LDH and ECOG > 0. Corresponding median OS was 17m (95%CI 15-18), 12.7 (95%CI 10-19), 7.4 (95%CI 6.3-8.9) and 5.3 (95%CI 3.8-6.1). CCI were PFS 0.72 (SE 0.02, 95% CI 0.69-0.75), OS 0.73 (SE 0.02, 95% CI 0.7-0.76).Table:
1239P
| Variable (n = 463) (n (%), median, range) | 7thAJCC | HUCH | PMN | |
|---|---|---|---|---|
| ECOG | 0 | 296 (64) 156 (34) 11 (2) | ||
| 1 | ||||
| > 2 | ||||
| Diameter in cm of largest metastasis | < 3 cm | 1.9 (0-2.9) | ||
| 3-8 cm | 4.4 (3-8) | |||
| > 8 cm | 10.8 (8.1-22.5) | |||
| Diameter of largest liver lesion | 3.8 (0-22.5) | |||
| % liver involvement | < 20 20-50 > 50 Missing | 3 (0-65) 5 (1) 3 (1) 452 (98) | ||
| LDH | 344 (39-8198) | |||
| ALP | 89 (24-1178) | |||
| Disease free interval | Missing | |||
| OS (%, 95% CI) | ||||
| M1a | 6 | 89 (83-93) | 84 (80-88) | |
| 12 | 60 (52-68) | 55 (49-60) | ||
| 24 | 25 (18-33) | 22 (18-27) | ||
| M1b | 6 | 68 (62-74) | 48 (39-57) | |
| 12 | 38 (31-44) | 17 (11-24) | ||
| 24 | 14 (94-19) | 5 (2-1) | ||
| M1c | 6 | 45 (33-57) | 8 (0-29) | |
| 12 | 19 (10-29) | 0 | ||
| 24 | 78 (27-17) | 0 | ||
Conclusions
Prognostic models in MUM remain imprecise in an externally validated dataset. Further validation is needed to find clinical utility
Clinical trial identification
not applicable
Legal entity responsible for the study
Princess Margaret Cancer Centre
Funding
None
Disclosure
L. Khoja: Employed by Astrazeneca plc. All other authors have declared no conflicts of interest.