Read the ABT-450 EASL abstracts carefully. One abstract reports a study on IL28B CC allele patients (Incivek reports 100% cure for this group of patients). The contents of the other study are not internally consistent if they had chosen the naive patient group randomly as they should.
Any randomly selected naive patient group contains 55% of potentially SOC-(IFN/RBV) failing patients. Out of the 55% two third will be non-responders and one third will be relapsers. So, 55% x 2 /3 = 37% of naive patients should be non-responders. Now, only 47% of these 37% can be cured with Abbott's quad, and 53% of 37% would fail:
.53 x .37 = 0.196 = 20% = 100% - 80%
This means that even if all SOC responsive patients and relapsers are cured by the quad, the best SVR rate one can hope for is 80%, not 93-95%.
You see why there is this serious inconsistency in their data. One possibility is that their naive patients are not randomly selected and they don't represent the true naive population. Another possibility is their non-responder data were wrong.
But think about this: it is easy to cheat random selection than non-response selection. The latter is almost certainly the most difficult to treat by any protocol while a naive patient group can contain many diverse SOC-responders and there is a greater degree of freedom in selection particularly for a small (N<20) studies reported by Abbott.
Only about 45% of representative GT1-infected patient sample will be cured when they are treated with SOC, and 55% will fail because any naive patient group contains a large number of non-responders(this is inherent to the patients' physiology, not to the history of treatment).
I think you were referring to the below data from ABT abstrct:
"95 percent (18 of 19) of treatment-naive patients infected with HCV GT1 (17 GT 1a, 2 GT 1b) achieved SVR12 with ABT-450/r 250/100 mg dosed once daily (QD) + ABT-333 400 mg dosed twice daily (BID) + ribavirin (Arm 1)."
So are you saying that since the above treatment naive group (GT 1a) did not specify the number of each subtypes CC, CT, TT, it could be that most were CC types and thus skewed to show a better result than if it included equal number of each subtype?
Based on your analysis I agree that if most were of CC type the data does not seem to be as good as it looks.