base model structure on covariate model building Janet
base model structure on covariate model building Janet R Wade 1, Rik Schoemaker 1 and Lene Alifrangis 2 1 Occams, The Netherlands; 2 Symphogen A/S, Denmark Background and Objectives BLOCK(3) Diagonal Sym 004 is a mixture of two synergistic full-length anti-EGFR antibodies (futuximab & modotuximab) that bind to 2 separate non-overlapping epitopes and inhibit the sustained growth of cancer cells. The objectives of this work were; 1. Develop a population (Pop) pharmacokinetic (PK) model for Sym 004 and evaluate the potential for covariates to explain the inter-individual variability (IIV) in the model. 2. Evaluate if the Sym 004 covariate model depended on the presence/absence of correlations between the IIV parameters. 3. Evaluate if the Sym 004 Pop PK model could also describe the PK of the two constituent antibodies. Data PK data were available from 136 patients from 2 trials in advanced solid tumours, metastatic colorectal cancer and squamous-cell carcinomas of the head and neck (Sym 004 -01 and -02), and who contributed 3948 concentration measurements (Figure 1). Figure 2 Forest plots showing single screening covariate effects for Sym 004 using diagonal and BLOCK(3) omega structures. Sym 004 (0. 4 -18 mg/kg) was dosed by IV infusion weekly or every 2 nd week or as a 9 mg/kg loading dose followed by 6 mg/kg weekly. Vmax (mcg/h) Km (mcg/m. L) V 1 (m. L) Q (m. L/hr) V 2 (m. L) CL (m. L/hr) Prop error (fraction) Add error (mcg/m. L) Weight on Vmax Weight on CL Weight on V 1 Weight on V 2 log Albumin on CL ECOG 3 or greater on CL log dose on VMAX ECOG 3 or greater on VMAX log tumour type on Vmax (Other solid tumours) log Male Sex on V 2 log Male Sex on CL log age on Vmax log tumour size at baseline on Vmax Block(3) 1371 2. 91 3564 36. 6 2856 13. 7 0. 181 1. 79 0. 592 0. 525 0. 634 0 -0. 344 -0. 261 0. 186 -0. 182 -0. 139 -0. 283 - Diagonal 1340 2. 86 3570 36. 8 2860 14. 5 0. 181 1. 79 0. 645 0. 309 0. 639 0 -0. 368 -0. 373 0. 190 -0. 090 -0. 197 -0. 282 -0. 139 -0. 183 0. 264 Table 1 Parameter estimates for the Sym 004, covariate influences for the final models with diagonal and BLOCK(3) omega structures. Figure 1 Observed Sym 004 concentrations versus time, by dose level. Red points and lines are the observed data. Blue lines are smoothes. Methods Modelling was done in NONMEM v 7. 3 (FOCEI). A two compartment model with target mediated drug disposition (TMDD) as the Michaelis-Menten implementation and including an a priori influence of weight was fit to the data. Inter-individual variability was included on clearance (CL), volume of the central compartment (V 1) and maximum velocity (Vmax). A proportional + additive error model was used. Covariate model building was performed by evaluating each covariate one by one (on preselected parameters) and then building a full final model with all covariates whose point estimates were outside the arbitrary range of 0. 8 to 1. 25 and whose 90% confidence intervals did not overlap the null value [1]. Covariate model building for Sym 004 was done with both diagonal and a BLOCK(3) omega structures. Sym 004 Vmax (mcg/h) Km (mcg/m. L) V 1 (m. L) Q (m. L/hr) V 2 (m. L) CL (m. L/hr) Prop error (fraction) Add error (mcg/m. L) Weight on Vmax Weight on CL Weight on V 1 Weight on V 2 log Albumin on CL ECOG 3 or greater on CL log dose on VMAX ECOG 3 or greater on VMAX log tumour type on Vmax (Other solid tumours) log Male Sex on V 2 1371 2. 91 3564 36. 6 2856 13. 7 0. 181 1. 79 0. 592 0. 525 0. 634 0 -0. 344 -0. 261 0. 186 -0. 182 Futuximab (m. Ab 992) 612 1. 19 3893 33. 4 3051 17. 0 0. 207 0. 68 0. 574 0. 547 0. 629 0 -0. 296 -0. 231 0. 130 0. 112 Modotuximab (m. Ab 1024) 724 1. 96 3247 36. 2 2641 11. 7 0. 206 1. 11 0. 745 0. 373 0. 647 0 -0. 382 -0. 860 0. 162 -0. 139 -0. 132 -0. 129 -0. 283 -0. 228 -0. 304 Table 2 Parameter estimates for Sym 004, futuximab & modotuximab final models, using the BLOCK(3) omega structure. Conclusions The final Sym 004 Pop PK covariate model structure depended upon the underlying statistical model structure, despite low correlations between the IIV parameters [2]. The extra covariates present in the final model with the diagonal omega structure were scientifically plausible. The model structure of the base and final Sym 004 Pop PK BLOCK(3) models were applied separately to each Sym 004 constituent antibody. The time gained by using the more simple omega structure (more stable model, faster run times) could be lost by having to explain/explore additional covariates that may be present in final models with a more simple statistical structure. Results Effort should be made to consider the magnitude of included covariate effects in relation to their clinical importance when building Pop PK models. If the Pop PK analysis is supporting submission documents and labelling then effort should be made to a priori define when a covariate effect is clinically meaningless (no effect), clinically irrelevant (small effect) or clinically important (larger effect) during the planning of analyses. The two compartment model with TMDD fit the Sym 004 data well. Correlations between the 3 IIV parameters were -0. 277, 0. 334 and 0. 396. Forest plots showing the one by one covariate screening results for Sym 004, using both the diagonal and BLOCK(3) omega structures are shown in Figure 2. Differences can be seen in the magnitude and precision of the covariates effects for the diagonal and BLOCK(3) omega structures. The minor differences in the parameter estimates for the two Sym 004 constituent antibodies for both base and final Pop PK models supports analysing the combination, Sym 004. The final models for Sym 004 BLOCK(3) and diagonal models included 6 and 9 covariates, respectively, above the influence of weight. References The parameter estimates for the covariate influences for Sym 004, for the diagonal and BLOCK(3) omega structures are presented in Table 1. Gastonguay MR. Full Covariate Models as an Alternative to Methods Relying on Statistical Significance for Inferences about Covariate Effects: A Review of Methodology and 42 Case Studies. PAGE 20 (2011) Abstr 2229 [www. page-meeting. org/? abstract=2229]. The parameter estimates for the base structural model for Sym 004, futuximab & modotuximab were very similar (results not shown). The parameter estimates for the Sym 004, futuximab & modotuximab final models, using the BLOCK(3) omega structure are presented in Table 2. JR Wade, SL Beal and NC Sambol (1994). Interaction between the choice of structural, statistical and covariate models in population pharmacokinetic analysis. J. Pharmacokin Biopharm, 22, 165177.
- Slides: 1