Neurochem's Risky Bet

The patient numbers are small; only 24 patients remained in the study at 20 months. This makes it hard to draw any good conclusions on Alzhemed's efficacy. The use of historical control data -- especially 12-year-old data -- also is dubious because Alzheimer's care has certainly improved since then.

But the biggest red flag, in my mind, is that these data are potentially biased because it excludes patients who were performing poorly on Alzhemed. Of the 42 patients who started the extension study, 18 -- or 43% -- had dropped out by 20 months.

But the dropped patients weren't included in the efficacy analysis. At least six of these dropped patients stopped taking Alzhemed because of "perceived lack of efficacy," according to the published report. Excluding these patients from the analysis likely makes Alzhemed appear more efficacious than it really is.

If it's hard to extract any confidence in Alzhemed from the phase II study, what about the ongoing phase III studies? Is there preliminary data there that might be used as a proxy to predict the study's final outcome?

Neurochem says such data do exist. The company has compiled blinded efficacy data combining patients on Alzhemed and placebo that it says point to an "encouraging trend" in favor of the drug,

The chart below is the data that Neurochem's Bellini showed investors at the JPMorgan conference. Like the chart above, smaller changes in the mean ADAS-cog score represent a slower loss of cognition and a better outcome for patients.

Blinded results from Phase III Alzhemed study 520 patients followed for 18 months Mean change in ADAS-cog score from baseline
	Blinded mild patients		Blinded moderate patients
	n=404		n=116
improved/stable	41%	-1.6	17%	-1.3
slow progression	23%	4.5	16%	4.4
progression	36%	13.8	66%	15.8
Source: Neurochem presentation Jan. 2007

The takeaway point, from Neurochem CEO Bellini's perspective, was that a majority of the blinded patients with mild Alzheimer's (41%) continue to show improved or stabilized cognition after 18 months of treatment with Alzhemed. Since this data is blinded, Neurochem doesn't know which patients are on Alzhemed and which are taking a placebo, but the overall favorable trend suggests that the drug is having some positive effect, he says.

Again, I believe this analysis falls short. From the above chart, it's possible to calculate the weighted average of the blinded results for the mild and moderate groups.

Mild patients	5.3 point progression
Moderate patients	10.91 point progression
Weighted average for total	6.6 point progression

Examined this way, it's hard to be as confident with Alzhemed's performance, because the drug's blinded performance (a 5.3-point worsening of cognition on the ADAS-cog scale for mild patients) is comparable to placebo patient data in other recently conducted Alzheimer's drug trials.

An analysis of these studies, done by Sprott Securities analyst David Dean, showed that Alzheimer's patients given a placebo show about a 6-point progression on their ADAS-cog score over 18 months. (Dean has a reduce rating on Neurochem and his firm has no banking ties to the company.)

In others words, based on the data available, it's not clear what level of effect, if any, Alzhemed is having on patients with Alzheimer's. This makes the pending release of results from the Alzhemed phase III study a very high-risk event.