Replies to post #176040 on Biotech Values

[AlpineBV_Miller]

EXEL confirmed long time ago the total alpha to spend in COMET-1 is 0.05.

They have never said how they allocated it, though I would have to say I never detected any sense that they would be surprised if the trial went to the full analysis.

FWIW, here is what I've worked up over the years (at BSR and ABV) on COMET-1.

COMET-1 enrollment pattern

May “initiated trial” 280 international sites targeted.
July to December 2012 = 60 patients
February 2013: “50% of trial sites are activated”
May 2013: “Most COMET trial sites are activated”
January to September 2013 = 900 patients
“960 patients enrolled in 14 months”

Statistical assumptions

1*=OS
N=960
Randomized 2:1
Study arm 9.0 month OS assumption (cabozantinib monotherapy)
Control arm 7.0 month OS assumption (prednisone monotherapy)
90% power at HR=0.75, p=0.05
One interim analysis, originally disclosed as 387 patients (later as “2/3rds of total events”)
Final analysis confirmed at 578 events

Other

The 7-month baseline assumption was derived from the COU-AA-301 trial of abiraterone (Zytiga) and represents the duration of treatment on the Zytiga arm between the end of treatment (median time on drug) and death (median overall survival).

There is no SPA.

The trial was initiated May 30, 2013, but management said they enrolled the 960 patients in 14 months.

I expect the trial was enrolled mostly ex-US, particularly in Europe.

Hope everyone is doing well, been a long time since I dropped by.

David

[p3analyze]

Just curious how do you know alpha is 0.05? Mgt never disclosed it. Did you back calculate from the total event goal, delta, beta and randomization ratio, the info fraction and confirm that's the case?

[Summer2762]

EXEL.

In the same exact indication, at similar (but not exact) # of events (387 of 578 vs 305/365 of 511), cabazitaxel pivotal trial which met its study objective by showing a hazard ratio of 0.70 p < 0.0001 but also failed its interim.

The second interim analysis
was added after a protocol amendment to be
performed at the time of the 307 deaths (the 60%
of the 511 deaths in the final analysis of the
protocol) to assess the primary efficacy endpoint of
OS based on the O’Brien-Fleming type I error
spending function. The actual number of deaths at
the second interim analysis was 365 instead of 307.
Therefore, the type I error of the second and final
analyses were adjusted according to the O’Brien-
Fleming type I error spending function. The
corresponding statistical significance levels for the
interim analysis and the final analysis of OS were
0.0160 and 0.0452, respectively.

http://www.accessdata.fda.gov/drugsatfda_docs/nda/2010/201023Orig1s000StatR.pdf

I just read the related PR.

http://www.exelixis.com/investors-media/press-releases?cpurl=http%3A%2F%2Fir.exelixis.com/phoenix.zhtml?c=120923%26p=irol-newsArticle%26ID=1912367%26highlight=

The PR does not state the reason why DMC recommended continuation.

Similarly we do not know whether cabazitaxel hit its required significance level at interim or DMC was worried about AEs and recommended continuation.