BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Should Roche's Failed Trial Give Hope To Alzheimer's Patients?

This article is more than 9 years old.

Roche’s Genentech unit just released a mid-stage study of a drug that seeks to slow the progression of Alzheimer’s by blocking the accumulation of a protein called amyloid beta. Like every other medicine that has attempted this approach – which remains the main one being pursued by drug companies to battle Alzheimer’s – the drug failed to yield a statistically significant benefit.

Unlike those previous failures, however, there is a glimmer of hope in the data on Roche’s drug, crenezumab, which was presented today in Copenhagen at the annual meeting of the Alzheimer’s Association. Why? The data seem to tell a consistent story that the drug may have an effect.

  1. Crenezumab didn’t result in a statistically significant benefit compared to placebo on performance in tests designed to assess how well patients were doing, but the group that got the drug did do better. This difference was consistent across different subgroups.
  2. The patients with the mildest disease seemed to get the biggest benefit from crenezumab. This is very much what Eli Lilly saw with its amyloid beta drug, solezumab. Again, the consistency of the data is a reason for hope.

There are reasons to believe that crenezumab has a better chance than previous beta amyloid blockers. Bapineuzumab, developed by Pfizer, Johnson & Johnson, and Elan, caused dangerous brain swelling. Crenezumab, invented by Swiss biotech AC Immune and licensed by Roche, doesn’t seem to cause this side effect, and that has allowed Roche to give it at doses 50 times higher than those used with bapineuzumab. The dose in this study was 50% higher than in previous ones, and Roche is still working to figure out if it can push it still higher.

That is not to say that the odds of success are high. Generally, in drug development, companies now follow an approach that they don’t want to develop drugs unless they show clear benefit in early studies. What is being done in Alzheimer’s is much riskier: companies are swinging wildly for the fences in the hope that they will hit an out-of-the-park home run. As Timothy Anderson at Bernstein Research wrote in a note to investors:

“As a general rule, drug companies normally won’t advance compounds into phase 3 development unless the probability of success is estimated to be in the range of 60-70%. With the Alzheimer’s disease drugs described in this report, the probabilities are likely much lower (even optimistic experts won’t claim a figure higher than ~25%). What drives this? The tremendous size of commercial opportunity. As we have shown previously (Exhibit 4), a drug that successfully modifies AD progression could make Lipitor (sales peaked at ~$12B) look like a mid-sized product. “

We’re in the land of long shots here. But Roche is taking an innovation it learned in cancer – to use the mid-stages of the clinical trial process to conduct rigorous tests of an experimental medicine’s potential – here. And if you squint the right way, the crenezumab data do indicate that it might be worth moving into a larger study aimed at getting the drug approved.

A low dose of the drug was studied in 184 patients, 62 of whom got placebo; this part of the study failed unequivocally. Another 247 patients (84 on placebo and 163 on crenezumab) got the higher dose, and here there’s some hope.

The goal was for patients to do better both on two sets of tests: the 12-item cognitive subscale of the Alzheimer's Disease Assessment Scale (ADAS-cog12), and the Clinical Dementia Rating-Sum of Boxes (CDR-SOB). By strict statistical measures, which require a p-value of less than 0.05, meaning a less than 1 in 20 chance that a result happened randomly, the study failed on both measures.

However, Carole Ho, Roche’s Group Medical Director of Exploratory Clinical Development, says that Genentech powers its phase II trials to deliver a p value of less than 0.2; by this measure, the study succeeded on ADAS-Cog12, but not on the second measure, CDR-SOB.

Here is what that result looks like:

As you can see, this isn’t a big difference. The placebo patients saw their ADAS-Cog12 scores drop 10.56 points, compared to 8.79 points for those on crenezumab. That’s just a 1.78 point difference (and that p value is just 0.19, barely making Roche’s internal hurdle). The amount of overall decline in both groups is a lot bigger than the difference between them; would individual patients even be able to notice this difference?

But things look a lot better if you cut the data into the patients into different subsets based on the Mini-State Mental Examination (MMSE) at the beginning of the study. A higher score on this test means better cognitive function.

In the least sick patients, the difference between the crenezumab and placebo groups is nearly doubled, and the p value would be significant if just those patients had been tested. (Don’t get too excited; statisticians don’t count subgroups like this because the odds are if you look at enough subgroups, *something* will turn out to be statistically significant.)

And hey, using another standardized test, the Dantes Subject Standardized Tests, you see the same trend, with a much bigger benefit on the patients with higher MMSE scores.

And in BLAZE, another clinical trial, the drug again seems to result in a difference in ADAS-Cog12 in those with a high MMSE score, but not in those with a low one.

Great, but the results reveal the same problems that have dogged Alzheimer’s drug development so far. All of these are just measures on standardized tests, and that means the data are noisy. Worse, it’s hard to tell how many patients have Alzheimer’s, compared to other forms of dementia. Roche is dealing with these problems head-on – including with a fascinating trial that uses the drugs before the disease starts in patients who have a gene that makes them destined to get it. The company will need to do more.

Based on this data, a real test of this data would require recruiting patients based on biological evidence they have Alzheimer’s – perhaps a PET scan of levels of amyloid protein in the brain – when their disease is still mild. Then it might be possible to show a benefit in a big clinical trial. This is similar to the strategy that Lilly is following with its solanezumab. It might work. But the odds are still very, very long. It’s hard to know what investors, patients, or scientists should make of the data so far. The data are just good enough that it’s hard not to say Roche should keep going.

Roche allowed me to review these data ahead of their release, on the condition I not write on them until they were made public.