comparing vx-809 400 mg bid dosing and 600 qd with kalydeco 250 mg bid, looking for some thoughts on why the mean fev1 was not statistically significant for day 0-56. There seems to be an upward trend..is the duration not long enough to see significance?
Here is the data: day 0-28 day 28-56 day 0-56
400 bid/250 bid -4.3%(p=.04) +6.6%(.01) +1.9(.57)
600 qd/250 bid -2.9%(.07) +6.1(
H.S., you pointed out an intriguing puzzle in a very important data set. Both Cohort 3 and Cohort 2 show that the p-values for Day 0-56 data are almost two orders of magnitude greater than the Day 28-56 data. In addition, the sums of data points for Day 0-28 and Day 28-56 do not give the data for Day 0-56. This can be understood by observing that data for Day 0-56 depends only on the FEV1 data at Day 0 and Day 56, but not on the data at Day 28. Moreover, it is conceivable that the noise level at the beginning is high because of individual variation such as age, progression of lung damage, and other factors, but the effects of the drugs are more or less uniform (for example, a person with a low FEV1 gets almost as much benefit as another person with a high FEV1).
Because of the uniform Kalydeco effect on individuals, the data for Day 28-56 have the low p-values and the high statistical significance. This indicates that the Kalydeco effect has a large signal-to-noise ratio. And VX-809 also give rise to the fair effect for Cohort 3, but not as uniform as the effect of Kalydeco, and the p-value for Day 0-28 is not as low. The Cohort 2 datum is more significant than the datum for Cohort 3 purely because there were more participants in Cohort 2.
Then, how can we understand that the data for Day 0-56 have such a low statistical significance? First, the number of participants is small particularly for Cohort 3, and secondly, the variabilities of the two drug effects add to give the noise in the overall drug effect (a person may respond well to both drugs, while another responds poorly to both drugs, which will give a large standard deviation of the mean).
I think that the conundrum is solved if not completely or quantitatively.
Vertex had repeatedly said the 400 bid was not powered for FEV but only safety and it was only like 10 participants so too small to be statistically significant. I’ve been bouncing for joy though after the presentation because the 400 bid resulted in much tighter cmax/cmin and better exposure levels. Vertex showed a slide demonstrating 70% of those in the 400 bid dosing were in the top quartile of drug exposure and that the top quartile of drug exposure had the best FEV improvement. Significantly, then, during Q&A someone asked what it was for 600 mg and it was only 40% in top quartile of exposure. Extrapolate that improvement from 40 - 70% with the the responder analysis for the 400 mg dosing and I think you’ll come to the same conclusion I did. . . the 400 bid dosing is going to greatly improve FEV. Also, if you looked at the graphs showing the 0-28 and 28-56 days FEV you’ll see amazingly parallel responses, but with 400 bid dropping lower in monotherapy arm. This shows it is a dose response (as Vertex said), but also gives me hope of even better FEV improvement when there is no monotherapy arm in Phase 3. Vertex said it does not expect an initial decline in FEV when dosed together and it would know since Phase 2 661 is not doing a monotherapy arm and since Phase 2 is dose-escalating, Vertex must know early results since is in the 4th and final dosing for Phase 2. In short, I think approval of the combo is assured by February of 2015 AND more importantly, that the improvement will be much closer to Kalydeco than for double deltas.