You get what you measure: evaluating endpoints in MS clinical trials.

Neurology 2008;71:620–621 Since the emergence of the modern therapeutic era in multiple sclerosis (MS), the field has focused on identifying agents to prevent long-term disability progression. In lieu of measuring the true endpoint of interest, long-term disability progression, clinical trials typically measure surrogate endpoints, such as short-term disability progression or changes in lesion burden on MRI.1 Because these surrogates can be measured earlier and more easily than long-term disability progression, clinical trials can be shorter and smaller. The critical issue, however, is whether changes induced by a therapy on a surrogate endpoint reliably predict the clinical benefit on the true endpoint. Surrogate endpoints are measures which correlate with clinically important (true) endpoints of interest; however, this criterion alone is not sufficient to define a surrogate endpoint.2 Ideally, surrogates are in the only causal pathway for disease progression. That is, for short-term disability progression to be a good surrogate for long-term disability progression, longterm progression should be solely mediated by shortterm progression. Also, the treatment under study must affect the true endpoint via its effects on the surrogate, and the entire treatment effect should be fully captured by the surrogate; this can be difficult to establish.2 Treatments developed using surrogates without these attributes may have undesired, unexpected, or unrecognized effects. Effects on the true endpoint may also be incorrectly estimated. Examples in other fields illustrate problems with using surrogate endpoints. Arrhythmia is associated with cardiovascular mortality, and was used as the surrogate endpoint in the Cardiac Arrhythmia Suppression Trial.3 In that study, three agents effectively suppressed arrhythmias, but were associated with an increased risk of sudden death. Diabetes is associated with microvascular and macrovascular complications. In diabetes trials, glycated hemoglobin is often used as a surrogate endpoint for new therapies; however, improvements in glycated hemoglobin may occur without demonstrable benefit on vascular complications, or even evidence of harm.4 The virtual absence of thoroughly validated surrogates leads some experts to suggest that surrogates be used for screening new treatments in phase 2 trials, but not in phase 3 trials.2 In this issue, Ebers et al.5 investigate the validity of commonly used surrogate endpoints in MS trials. Using pooled data from the placebo arms of 31 randomized clinical trials, they defined sustained worsening as either a minimum 0.5 or 1.0 increase in the Expanded Disability Status Scale (EDSS), sustained for 90 or 180 days. Similarly, sustained improvement was defined based on decreases in the EDSS. Participants with secondary progressive MS were more likely to worsen than to improve. Participants with relapsing-remitting MS, however, were equally likely to have sustained improvement as sustained worsening. Even when the authors defined worsening as a 1.0 point increase in the EDSS sustained for 6 months, there was no significant difference in the frequency of worsening vs improvement (p 0.184). The authors suggest that definitions of disability progression used in clinical trials are inadequate. Ebers et al. pooled data from many studies which were heterogeneous in study population, duration, frequency of assessments, and nature of missing data. They did not address the question of whether individuals with transient worsening of disability are at greater risk of long-term disability progression than those who remain stable. Beyond clinical course subgroups, they also did not address whether these endpoints function better in some patient populations than others. Nonetheless, this study suggests that there are significant limitations to using commonly applied definitions of short-term disability progression as endpoints in clinical trials for relapsingremitting MS. Disputes persist regarding the