How disease modeling and model-based meta-analyses contributed to understanding the path forward after dose discontinuation of a Phase III study in Alzheimer's disease
Sylvie Retout (1), Ronald Gieschke (1), Cornelia Weber (1), Jean-Eric Charoin (1), Dietmar Volz (2), Robert Lasser (2), Carsten Hofmann (1), Nicolas Frey (1)
(1) Roche Pharma Research and Early Development, Clinical Pharmacology, Roche Innovation Center Basel, F. Hoffmann-La Roche Ltd, Basel, Switzerland, (2) F. Hoffmann-La Roche Ltd, Basel, Switzerland
Objectives: SCarlet RoAD (SR) was designed as a pivotal Phase 3 study to evaluate the effect of monthly subcutaneous doses of 105 and 225 mg anti-amyloid compound gantenerumab (GAN) in prodromal Alzheimer’s disease (AD) patients. Dosing was discontinued based on results of a pre-planned futility analysis of the clinical endpoints (N=312, 2 years completers). At the same time, Phase 1b results of aducanumab (ADU) , a very similar compound, became available, exploring much higher doses and suggesting that GAN was under-dosed. Leveraging a previously developed disease progression model  and available ADU data, we investigated the reasons of the futility results and uncovered a way forward.
Methods: The AD progression model was used to split the SR patients into SLOW progressors (SP) or FAST progressors (FP) based on 3 covariates at baseline (FAQ, CDR-SB and hippocampal volume). The GAN effect on AD scales time courses (ADAS-Cog13, CDR-SB, MMSE and CANTAB), was graphically investigated per category of SP/FP. Relationships between drug concentration and brain amyloid plaque removal measured by PET and between drug concentration and early amyloid related imaging abnormality vascular edema (ARIA-E) were modeled, as a GAN and ADU data meta-analysis. Both models are detailed in the 2016 PAGE abstracts  and .
Results: The lack of significant drug effect detection in SR could be explained by a relatively small proportion of FP (only one third at study entry), and doses that were too low. A clear exposure–ADAS-Cog13 relationship was observed among FP with a median decrease of 3 points at 2 years in the highest exposure group compared to placebo. Similar exposure trends were observed for MMSE and CANTAB. No exposure trend was observed among SP. Both PET and ARIA-E models described internal and external data well. Using simulations, up-titration dosing regimens were found to balance ARIA-E events and plaque removal, increasing the Ph3 success likelihood.
Conclusions: Quantitative clinical pharmacology tools, including disease modeling and exposure-response meta-analyses, shed positive lights on the GAN Phase 3 program and guided up-titration regimens using higher doses for a higher chance of success. Those GAN dosing regimens are currently being investigated in ongoing studies.
 Sevigny, J et al. 67th Annual Meeting of the American Academy of Neurology, Washington DC, US (2015)
 Delor, I et al. CPT: Pharmacometrics & Systems Pharmacology (2013) 2, e78
 Hutmacher, M et al. 7th American Conference on Pharmacometrics, Fort Lauderdale, US (2013)
 Frey, N et al. Abstr PAGE 25, Lisbon, Portugal (2016)  Gieschke, R et al. Abstr PAGE 25, Lisbon, Portugal (2016)