Limitations on the number of eligible patients, and the stringent requirements for demonstrating therapeutic efficacy mean these trials must be undertaken in a highly efficient and effective manner. However, this is a particularly challenging task, especially when people living with incomplete SCI (iSCI) are likely to improve spontaneously over the first year after injury and represent a very heterogeneous population in terms of recovery patterns.
Nonetheless, their recruitment is essential since it would more closely reflect the preclinical situation, where iSCI models are commonly used to test potential therapies, and speed the completion of a trial. Finally, it would target a therapeutic to a population where it is more likely to produce a benefit, because in iSCI there is some preserved function to enhance, which is not available in sensorimotor complete SCI. Current SCI trials have either not primarily targeted incomplete subjects or have not fully taken into account the consequences of their inclusion in terms of clinical trial design. Therefore, more rigorous and statistically sound approaches for the development of SCI clinical trial protocols are needed.
We recognize that a key prerequisite for significant advancements in the translational process is the development of accurate inclusion/exclusion criteria for SCI trial participants, whether they be complete (ASIA Impairment Scale A or AIS-A) or incomplete, (AIS-B to AIS-D). In addition, reasonable cohort-specific outcomes for those subjects enrolled in a trial are needed. Based on encouraging preliminary analyses, we will refine our findings and provide a sound rationale for homogeneous cohort identification by employing a recently developed statistical method - unbiased recursive partitioning. We will rely on the European Multicentre study about Spinal Cord Injury (EMSCI) database to study cohort-specific recovery patterns and clinical endpoints. We will also model recovery patterns over time using a multivariate approach that is not based on sum scores alone.
In this context, we plan to extend conditional transformation models, so as to be able to model the joint temporal distribution of ordinal items, and overcome current weaknesses in the analysis of clinical trials outcomes. We will extend and validate our findings using resampling techniques and independent data sets (Sygen® clinical trial).