NeuralStem’s results from their 15pt, open-label trial of neural stem cells (fetal source) in ALS proved to be both ambiguous and controversial. Since there was no control group, NeuralStem parsed the treatment sample into two cohorts, those who appeared to have responded, as shown by relative stabilization on the ALSFRS and grip strength measures, and those who were deemed to be non-responders, because of their decline. This was excessively self-serving, because NeuralStem then compared these two groups on their outcome measures, even though the groups had been solely devised on the basis of…those same outcomes. This produced a circular logic chain: These patients did better on the outcome measures because they were treatment responsive, and they were treatment-responsive because they did better on those measures. NeuralStem has had a tradition of heavily emphasizing anecdotal reports of improvement, thus it is not particularly surprising that they tried to earnestly, and heavily, spin these results–though the net result has been one of both decreased credibility and a shock to their valuation.
The data disclosure was sparse, but doing some back-of-the envelope math based on what was released, it would appear that the overall mean group rate of decline on the ALSFRS was 1.61 points/month. There have been reports that the measured rate of decline in ALS has been reduced in recent years; a Phase II trial, reported by Miller and Moore in 2011, utilized 249 historical control patients who had displayed a mean rate of decline of 1.01 points/month. Compared to that relatively recent benchmark, the overall rate of decline for patients receiving injections of sixteen million NSI-566 cells each, was worse than that seen in that historical comparison.
We would not dismiss the possibility of a genuine responder/non-responder subgrouping of patients, even if we do not yet have an explanatory framework for that bifurcation. Comparing the seven ‘responders’ to historical controls, their decline rate of .02 points per month is indeed better than the aforementioned historical control group mean. And the nonresponder group’s decline rate of 3.0 points per month would be about triple the historical control rate. Thus, if one is going to take seriously the responder subgroup’s improvement, and describe it as potentially due to the cell therapy, one would also have to take seriously the finding that the majority of patients (53%) not only did not respond positively, but actually did worse than their historical control group, raising the possibility that the treatment might have contributed to that worsening (no causality can be established from this fifteen patient pilot study). If one were to take this trial data at face value, which is a dubious proposition, it could be interpreted as showing that NSI-566 slows progression in some patients, but accelerates it in others. If that were the case, prospectively identifying the differentiator would be of vital importance, because hastening decline in a rapidly-progressing, fatal neurodegenerative disorder would not be acceptable in clinical trials or practice. NeuralStem believes that there may have been a differentiation between groups in terms of illness stage, based on muscle function measurements, but this remains to be confirmed as a prospective screen. While NeuralStem and its chief investigator say they plan to initiate a larger, controlled study later this year, that statement may be premature. While NeuralStem claimed that the trial showed NSI-566 to be safe, that assertion speaks only to its observable adverse event profile. An acceleration of disease progression in more than half the patients treated would be a genuinely problematic safety issue, and until NeuralStem has sorted out that question, we believe it would be folly to launch another trial–one that indeed might have considerable difficulty enrolling patients.