In October ’06, the people at ICORD (International Collaboration On Repair Discoveries) in Vancouver asked me to give a talk. I spoke on Statistical Considerations in Designing a Trial in Spinal Cord Injury with ASIA Motor Score as the Outcome, and you can download a 1.5 MB pdf copy of my slides. I tend to write everything I plan to say into my slides, so they’re less schematic and easier to read than most.
Recently there have been many potential treatments for spinal cord injury reaching the stage of development where they’re ready for human trials, and thus there’s interest in using previous data to develop design information. The largest existing data set is the one from the Sygen GM-1 multi-center trial, which recruited 760 patients in 28 centers in the US and Canada in the 1990s. I was a designer of it, and so my colleagues and I have been trying to be helpful in filling requests for information.
This talk summarizes my preliminary thoughts about using the ASIA (American Spinal Injury Association) motor score as the outcome variable. This score is highly useful and well-known and -validated as a clinical tool — for individual patients — but its properties in a research setting — for aggregate groups of patients — are poorly understood. We are in a long-term program to study it using our data.
Here’s a summary of what these slides deal with.
It’s important to avoid ceiling effects: patients with complete injuries tend to recover very little, so that it’s hard for any treatment to make headway — while patients with milder injuries almost always recover significantly, so that it’s hard for any treatment to appear better than placebo, especially at later measurements. It was hoped that the ASIA motor score would be a more sensitive measure than the ASIA Impairment Scale (AIS) and be less prone to this — but it still has difficulties.
The sample size estimates are based on the Sygen data. Recruitment was very uneven: much higher for cervical injuries than for thoracic, and much higher for complete injuries than incomplete. Further, due to the nature of the measurement process, the ASIA motor score is prone to artifacts in a research setting that would not be noticed clinically. In particular, its distribution is very asymmetrical and different in different population subgroups. Therefore the standard deviation observed in any trial (and, accordingly, the sample size needed) can vary drastically depending on the proportion in which these subroups are recruited.
Kim Anderson’s survey of spinal injury patients has shown the desirability of alternative outcome measures — better use of arms and hands, upper body/trunk strength and balance, bowel and bladder control, blood pressure regulation, sexual function, normal sensation, and pain relief. These outcomes do not necessarily accompany, nor are they necessarily accompanied by, the return of ability to walk. (See, for example, an earlier blog engtry: Reduced Bladder and Bowel Control after Severe Spinal Cord Injury — Even in Patients Able to Walk.)
It is important to model covariates, and the Sygen data show the most important ones are AIS grade at baseline, cervical vs thoracic, timing of decompressive surgery, mild injury (central cord, no fracture dislocation, or not requiring decompressive surgery), and age.