Model-based meta-analysis of aggregate data on Parkinson’s disease clinical scores to inform future trial design
Usman Arshad, Fatima Rahman, Nathan Hanan, Chao Chen
Clinical Pharmacology Modelling and Simulation, GSK
Objectives: There is a dire need to develop drugs that modify the progression of Parkinson’s disease (PD). Differentiation of disease modifying drugs from symptomatics requires accurate characterization of natural disease progression, placebo response dynamics, and symptomatic drug effects. A strong placebo effect can lead to high uncertainty in the outcome of a clinical trial. The objective of this analysis was to provide a robust quantitative description of longitudinal changes in total Unified Parkinson's Disease Rating Scale (UPDRS), with the aim of improving the probability of success of future PD trials.
Methods: Summary-level data were curated from literature published between 1996 and 2020 reported from observational and interventional clinical PD studies. Model-based meta-analysis (MBMA) was performed to describe the time course of summary UPDRS score of Parts 1, 2 and 3. Underlying disease progression was modelled by a linear function, while placebo response was modelled with the inverse Bateman function and drug effect was modelled with an onset rate and amplitude of symptomatic response [1, 2]. Investigational agents proven ineffective were treated as placebo arms. Inter-study and inter-arm variabilities of model parameters were estimated as permitted by the data, whereas the residuals were weighted by study arm size. Model performance was evaluated by simulating placebo-controlled trials over a period of 2-years and assessing the agreement between observed and simulated data. Active-placebo treatment difference over time was graphically compared between observed and simulated data. Maximal placebo and symptomatic treatment response, and the time to attain the maximal placebo and symptomatic treatment response, were derived using the simulation output.
Results: Summary-level total UPDRS scores from 62 arms [observational (4), placebo-treated (27) and investigational-drug-treated (31)] from 4 observational studies and 16 interventional trials were included in this analysis. Total UPDRS was estimated to progress by 3.80 points/year. Arms with lower baseline severity showed a faster disease progression. The half-lives for treatment response onset were comparable between placebo (5.25 weeks) and active (4.37 weeks). Placebo response amplitude was estimated at -3.28 points, whereas further symptom relief by active treatments was estimated to be -2.93 points. Between-trial variability of progression rate, offset of placebo response, and the symptomatic drug effect amplitude varied by 108.4%, 61.9% and 76.0%, respectively. Maximal placebo and drug treatment response was predicted within 2 months; however, a period of 1 year was required for a complete reflection of the underlying treatment difference. Both observed and simulated profiles of placebo adjusted scores indicated a lack of disease modification.
Conclusions: Longitudinal UPDRS progression was described through quantification of placebo response dynamics, symptomatic drug effects and underlying disease progression using MBMA. Although the maximal symptom relief by placebo and active treatments could be observed within 2 months, a much longer duration would be required to capture the full treatment difference. The estimates of the fixed-effect parameters and between- and within-trial variabilities from the current analysis can be of great value for designing clinical trial simulations thereby enhancing the rigour and success of future trials of potential disease modifying drugs.
References:
[1] Pilla Reddy V, Kozielska M, Johnson M, et al. Structural models describing placebo treatment effects in schizophrenia and other neuropsychiatric disorders. Clin Pharmacokinet 2011; 50(7): 429–50.
[2] Chen C, Jönsson S, Yang S, et al. Detecting placebo and drug effects on Parkinson’s disease symptoms by longitudinal item‐score models. CPT Pharmacometrics Syst Pharmacol 2021;10(4):309–17.