I believe you have to read the CNSI Phase II Press release on the results of their stroke trial very carefully and analyze it to see that their so called "positive" result is far from impressive and maybe an artifact. I've also read the papers they've published on the matter, but the PR release will do for this analysis.
I'll annotate portions of their announcement to illustrate my point.
>> The double-blind, placebo-controlled study involved 132
>> patients enrolled at 30 centers in the United States and Europe.
>> The primary measure of outcome was the National Institutes of
>> Health Stroke (NIHS) Scale, a measure of the severity of
>> neurological impairment. Secondary measures included
>> assessments of function, such as the Rankin Scale and the
>> Barthel Index of Activities of Daily Living. Patients were
>> randomized to either placebo, or one of three dose regimens: 30,
>> 70 or 110 micrograms/kg of CERESTAT(R) by bolus and
>> infusion over a four-hour period.
>>
>> Among the findings were:
>>
>> -- At Day 30, patients in the top two CERESTAT(R) dose
>> groups improved on average by 5.2 and 4.3 points on the NIHS
>> Scale, in contrast to control patients whose average improvement
>> was 3.6 points. This effect did not reach statistical significance.
It is important to note that the NIH Stroke Scale is a 40 point
scale. Since the low dose group was only 1.6 scale points above placebo and the high dose group only 0.7 points above placebo, that the effect was only a 4% and and 2% improvement respectively. It is no
wonder that this was not significant.
>>
>> -- At Day 90, patients in the top two CERESTAT(R) dose
>> groups had improved on average by 5.2 and 6.3 points on the
>> NIHS scale, in contrast to control patients whose average
>> improvement was 4.7 points. The degree of improvement
>> compared to placebo for the highest dose group was statistically
>> significant at p < 0.04.
Once again the low dose group was only 0.5 scale points above placebo
and the high dose group only 1.6 points above placebo. This constitutes effect sizes of only 1% and and 4% improvement respectively. It is no wonder that only the highest dose was marginally significant. In addition, it should be noted that only the difference scores between entry scores and 30 or 90 days were reported. This was probably required as the result of a ceiling effect for the placebo group and effectively covers up a major defect in the analysis (see below).
> -- At Day 90, of the patients who had entered the study with
> more severe impairments, 32 to 40% of those who received the
> highest dose of CERESTAT(R) had no, or minor disabilities, in
> contrast to only 19% of such patients who received placebo.
> These responses were measured on the Rankin and Barthel
> functional scales respectively.
This is really a stretch. When you have to subdivide your groups
like this to find a difference your not being honest. You can subdivide groups in an infinite number of ways and keep doing it till you find a difference. You can do this with random numbers. I seriously doubt that this result is statistically significant. If it was they would have said so and the n here was probably quite small.
After all there were 132 subjects divided into 4 groups, which
suggests that the groups only contained 33 subjects each. Then you
select out only those with the greatest impairment. Assuming this
means the top third, then there are only 11 subjects per group
left. Thus the 32 to 40% means 3 to 5 people showing improvement
compared to 2 for the placebo.
It's truly a sad commentary on the strength of the drugs
effect that they have to stretch the data like this.
In addition, why are the data for the Rankin and Barthel
scales not reported for the entire sample? If they were even vaguely
in the right direction, I'm sure they would have reported them like
they did for the 30 day NIH Stroke Scale data.
> Dr. Edwards commented that even before the treatment groups
> were identified in the double-blind trial, some patients appeared
> to have a superior course of recovery. "My impression, before
> the blind was broken, was that some patients had a course of
> recovery that was much better than I would have expected
> without treatment, and that impression has been confirmed by
> subsequent analysis."
This is a typical non-scientific dodge. They're using a testimonial
composed of some anecdotal impressions to bolster their case. This
is only necessary when a research study is short on hard evidence !
> Dr. Edwards added that the placebo group had more mild strokes
> than the treated groups, so that despite the trend towards greater
> improvement on the NIHS Scale at day 30 in the two higher dose
> groups, the effect was not statistically significant.
This is a real problem for the study. The lower recovery rate for
the placebo group could be due to a classic ceiling effect, wherein
the placebo group cannot improve more because they aren't as impaired
to start with. This was not considered in any of the analyses that
they report.
This is a clear case of playing with statistics to make very
weak results appear to be meaningful.
John de C |
