Pascal Chanu (1), Laurent Claret (1), Rene Bruno (1), Chunze Li (2), Priya Agarwal (2), Sandhya Girish (2), Angelica Quartino (2), Jin Yan Jin (2), Dan Lu (2)
(1) Certara Strategic Consulting, France, (2) Clinical Pharmacology, Genentech
Objectives: To assess the link between tumor growth inhibition metrics (TGI) and progression free survival (PFS) based on previously untreated HER2-positive progressive or recurrent locally advanced or metastatic breast cancer (BC) data from the MARIANNE study [1].
Methods: TGI data from 868 patients who received trastuzumab emtansine (T-DM1), T-DM1 plus pertuzumab or trastuzumab plus a taxane in the Phase III study Marianne [1] were fit using tumor growth inhibition models [2, 3] by NONMEM version 7.3.0 [4]. The relationship between model-based estimates of TGI metrics [5]: growth rate (KG) estimated with the bi-exponential TGI model [2], early tumor shrinkage at week 8 (ETS) and time to growth (TTG) estimated with the simplified TGI model [3], as well as baseline covariates (~35) and PFS were tested in a multivariate parametric distribution survival model to select the best predictors of PFS in R version 3.1.2. Different distributions were tested. PFS model performance was evaluated by comparing the simulated distributions of PFS by quartile of TGI metrics or other significant covariates with the observed distributions in the MARIANNE study.
Results: The simplified TGI model [3] provided a better fit of the tumor size data compared to the bi-exponential TGI model [2]. Log(KG) estimated from the bi-exponential TGI model was identified as a much better predictor of PFS than TTG or ETS (based on difference in log-likelihood to the null model in a Cox univariate analysis). PFS (month) was best described by a lognormal distribution with a linear log(KG)-PFS link, and LDH, SGOT levels and the number of disease sites at baseline identified as statistically significant prognostic factors:
|
Estimate |
Std. Error |
P Wald test |
|
|
(Intercept) |
-1.18 |
0.19 |
8.99e-10 |
|
logKG |
-0.799 |
0.033 |
4.65e-127 |
|
LDH (U/L) |
-0.000481 |
0.000119 |
5.37e-05 |
|
>2 disease sites |
-0.216 |
0.074 |
0.00343 |
|
SGOT (U/L) |
-0.00459 |
0.00151 |
0.00245 |
|
Log(standard deviation) |
-0.193 |
0.030 |
1.63e-10 |
The model was qualified to predict PFS distribution by Log(KG) quartiles as well as by levels of the prognostic factors:
|
Median PFS (months) |
||||
|
Log(KG) quartiles |
Observed |
Predicted |
95% PI |
|
|
Q1 |
not reached |
35.0 |
27.7 |
45.1 |
|
Q2 |
14.6 |
16.4 |
13.7 |
19.6 |
|
Q3 |
10.4 |
10.9 |
9.2 |
13.0 |
|
Q4 |
5.3 |
4.9 |
4.1 |
5.8 |
Conclusions: A survival model that uses a model-based estimate of tumor growth rate to predict the PFS for previously untreated HER2-positive BC has been developed. This model can be used to support early decision making of investigational agents in development for this indication based on the TGI metrics.
References:
[1] Ellis PA, Barrios CH, Eiermann W et al. Phase III, randomized study of trastuzumab emtansine ± pertuzumab vs trastuzumab + taxane for first-line treatment of HER2-positive MBC: Primary results from the MARIANNE study. J Clin Oncol (2015) 33 (suppl; abstr 507).
[2] Stein WD, Gulley JL, Schlom J, et al: Tumor regression and growth rates determined in ?ve intramural NCI prostate cancer trials: The growth rate constant as an indicator of therapeutic ef?cacy. Clin Cancer Res (2010) 17: 907-917.
[3] Claret L, Gupta M, Han K, et al. Evaluation of tumor size response metrics to predict overall survival in western and Chinese patients with first line metastatic colorectal cancer. J Clin Oncol (2013) 31: 2110-2114.
[4] Beal SL, Sheiner LB, Boeckmann AJ & Bauer RJ (Eds.) NONMEM Users Guides. 1989-2011. Icon Development Solutions, Ellicott City, Maryland, USA.
[5] Bruno R, Mercier F, Claret L. Evaluation of tumor-size response metrics to predict survival in oncology clinical trials. Clin Pharmacol Ther (2014) 95: 386-393.
Reference: PAGE 25 (2016) Abstr 5922 [www.page-meeting.org/?abstract=5922]
Poster: Drug/Disease modeling - Oncology