Though deemed “rare”, approximately 25—30 million Americans, and 300 million people worldwide, have been diagnosed with one of >6,800 known rare diseases. An estimated 50% of those affected are children. The prevalence of individual rare diseases can range from dozens in the US (progeria, hypophosphatasia, Niemann-Pick, for example), to thousands globally (such as, Crohn’s disease and cystic fibrosis). Although traditionally considered impractical for big pharma development, innovation and development of therapeutics for rare diseases have led to promising treatments for some diseases in recent decades. In the period of time from the passage of the Orphan Drug Act of 1983 until May 2010, the FDA approved 353 orphan drugs and granted orphan designations to 2,116 compounds. As of 2010, at 200 orphan diseases have become treatable (Armstrong). Still, rare diseases pose a critical unmet medical need.
Clinical trials for orphan drugs in rare disease populations present unique challenges to investigators. These challenges include small patient populations and their widespread geographical location, ethical decisions about placebo-treatments where no other treatment options are available, challenges in recruiting and treating pediatric patients, lack of information on disease progression, and limited data on reliability and utility of clinical endpoints.
Algorithms aimed at addressing the challenge of designing small randomized clinical trials have been proposed by Cornu et al and Gupta et al. In these literature reviews, the authors described alternatives to the “gold standard” randomized, parallel, placebo controlled design and provide examples of their applications to diseases such as Huntington disease, fibromyalgia, polyarticular onset juvenile idiopathic arthritis, pulmonary arterial hypertension, and many others. Decision trees were presented and used to guide the selection of clinical trial design. The authors noted that the choice between methods should be based on intervention, disease, anticipated recruitment duration and success, and current state of knowledge about the treatment. Designs which minimize the time on placebo, such as the randomized placebo-phase design, stepped wedge design, and randomized withdrawal designs were presented with respect to scientific and statistical merit and examples.
The application of population nonlinear-mixed effects modeling methods, to repeated-measures pharmacokinetic, pharmacodynamic, adverse event and clinical outcome data, is one solution, in part, to the analysis and extraction of information from such sparse data sets. These and other model-based strategies including, adaptive designs, meta-analyses, D-optimality, and simulation, have also been proposed to address the challenge of trial design. Bayesian model-based methods with or without the inclusion of informative prior distributions are one such strategy (as described in Johnson et al for the analysis of scleroderma clinical trials). Adaptive proof-of-concept/dose ranging trials, may also be useful when minimal prior information exists, and learning must be maximized within the context of an ongoing trial.
As more innovators develop therapies for rare disorders, experience is gained and shared regarding trial design, execution (video), and analysis. Informed decision trees, model-based data analyses, and simulation methods are likely to continue serving a critical role in improving clinical trial designs and the quality of the information gained in such precious populations. Given the obvious need and the ready applicability of tools and methods, the discipline of pharmacometrics has a unique responsibility to focus talents and resources on this important challenge.