Isabelle Delor (1), Sylvie Retout (2), Jean-Eric Charoin (2), Ronald Gieschke (2), Philippe Jacqmin (1) for Alzheimer’s Disease Neuroimaging Initiative
(1) SGS-Exprimo, Mechelen, Belgium, (2) F. Hoffmann-La-Roche Ltd, Basel, Switzerland
Objectives: To establish the disease onset time (DOT) and disease trajectory (DT) concepts[1] by developing an original natural history population disease progression model of Alzheimer’s disease (AD)[2,3] based on the CDR-SOB scores from the ADNI database[4].
Methods: The final data set consisted of CDR-SOB records collected from 229 control (NL), 380 mild cognitive impaired (MCI) and 180 AD subjects for up to 4 years. A total of 19 covariates were included in the database. Model development was performed in NONMEM V7.2.0.[5] using the FOCE method and guided by the OFV and GOF plots.
Results: Subjects entered the study at different stages of the disease. DOT and DT were implemented in a differential equation to describe disease progression:
dCDR/dT=(RATE+CDRxALP)xT30/(DOT30+T30)
At the time the disease was estimated to start in a subject, DOT activated the increase in CDR-SOB score (CDR) based on a disease rate (RATE) adjusted to the subject’s progression velocity by an additive term (CDRxALP). A transient drop of the score in some subjects just after study entry was captured by a ‘placebo’ effect function. In addition, it was detected that the disease progression evolved significantly more slowly in 40 to 50% of MCI subjects. To capture this difference in DTs, a mixture model was implemented allowing two different disease rates, coupled with the additive term only for the fast rate. Modelling was performed in the logit domain. CDR-SOB and ADAS-cog scores were identified as covariates of DOT and the Mini Mental Score Exam (MMSE) as covariate of ALP. Covariates for assignment to slow or fast progressing subjects were functional assessment questionnaire (FAQ), normalized hippocampus volume and CDR-SOB score. Based on these 3 prognostic factors at study entry, more than 78% of MCI subjects could correctly be assigned to the slow or fast progressing subpopulations. The final model could predict 85% of MCI-AD conversions. Finally, synchronization of biomarker time-profiles on individual DOT estimated by the CDR-SOB model virtually expanded the observation period of these biomarkers from 3 to 8 years.
Conclusions: The use of a DOT-DT model is powerful for detecting different disease progression rates in a population and identifying corresponding prognostic factors. Estimation of DOT allows the synchronisation of biomarker-time profiles on disease onset rather than on study entry and provides insights on long-term changes in biomarkers.
References:
[1] Yang, E. et al. : J Alzheimers Dis. (2011) 26: 745-753
[2] Samtani, M.N. et al. : J. Clin. Pharmacol. (2012) 52: 629-644
[3] Samtani, M.N. et al. : Br. J. Clin. Pharmacol. (2013) 75: 146-161
[4] http://www.loni.ucla.edu/ida%20/login.jsp?project=ADNI
[5] Beal SL, Sheiner LB, Boeckmann AJ & Bauer RJ (Eds.) NONMEM Users Guides. 1989-2011. Icon Development Solutions, Ellicott City, Maryland, USA
Reference: PAGE 22 (2013) Abstr 2789 [www.page-meeting.org/?abstract=2789]
Poster: New Modelling Approaches