A meta-analysis of randomized double-blind clinical trials in CMT1A to assess the change from baseline in CMTNS and ONLS scales after one year of treatment

The present meta-analysis supports the conclusions made independently within each clinical trial as regards efficacy of treatments and Placebo [10, 12, 14]. First, the CMT1A patients of the Placebo groups from studies conducted from 2006 to 2014 deteriorate rather slowly compared to the estimated natural progression of 0.686 point/year in CMTNS reported by Shy and colleagues in 2008 [18]. These findings are consistent with the positive placebo effects observed in diabetic neuropathy [19] or patient-reported pain outcomes [20], although the factors accounting for such a difference remain unclear. Lewis et al. [12] considered that systematic differences between participants of the different studies may be partially responsible; for instance the mean age and CMTNS are slightly higher in the four clinical trials considered here than in the natural progression study by Shy et al. [18]. Pareyson and colleagues [11] also pointed out that the natural progression study was partly retrospective, and therefore might not be directly comparable with clinical trials. Consequently, we believe that the progression of CMTNS and ONLS under Placebo reported here is more valuable than natural progression estimates for the design of future clinical trials in CMT1A, and less prone to sampling bias that might occur in single independent studies.

Second, the progression of patients under different dosages of AA appears quite stable, and does not reach statistical significance versus Placebo. The difference between AA and Placebo is far below the order of magnitude expected for sample size calculation in the three AA clinical trials. As it happens, the a posteriori power to detect this difference as significant does not exceed 15 % (assuming an SD in CMTNS of 5, a correlation between baseline and final values of 0.8, and an ANCOVA analysis at a two-sided 5 % level). In this context, designing a confirmatory Phase 3 study for a treatment showing such stabilization in CMT1A would require a much larger sample size and longer study duration, making it clearly unrealizable. It confirms the idea that an effective treatment for this disease should bring an improvement, rather than the mere ability to slow or stabilize the disease progression [12, 14]. Even if this effect seems quite marginal, a standardized re-analysis of all AA patient-level data would be of great interest.

Lastly, this meta-analysis supports an improvement in both CMTNS and ONLS with PXT3003 treatment, statistically significant when compared to Placebo. This improvement could herald an early, meaningful change in the disease course.

Conducting a meta-analysis of clinical trials in CMT1A is challenging because of the small number of studies and of the heterogeneity of study protocols in terms of recruitment criteria, study duration, balance of groups, and statistical analysis. In addition, our study evaluates CMTNS in a context where a second version (CMTNSv2) has been proposed to reduce floor/ceiling effects and eventually to improve the scale’s sensitivity to change [21]. The current version of the CMTNSv2 has also been questioned recently through a Rasch analysis by Sadjadi et al. [22] and a ‘weighted’ alternative has been suggested. In parallel, Mannil et al. [23] proposed a CMTNSMod by adding three functional measures (9-hole peg test, foot dorsiflexion and walk test) while removing Ulnar SNAP, Pin Sensibility, Vibration and Strength of Arms. None of these modified versions has been evaluated yet in natural history or therapeutic trials. Despite these limitations, the present study provides a set of relevant observations, consistently obtained on both CMTNS and ONLS, to be used for the design of future clinical trials in CMT1A.