I never though presentation of the HR was important as it was obvious to me that we had A versus (61%)AB in the results that increased the comparator arm, but you can censor the cross over data like they did on the sorafenib trial and as far as Cross over patients, its 156 of the 257 of the sorafenib arm, so there would be 101 patients in the control arm for OS and the US population of 18 patients will a stronger signal (they are 2:1 in median OS HR = .5 , so the average sorafenib OS should decrease even if the other patients removed were randomly distributed which is unlikely based on this. From the AVEO report to the ODAC/FDA: Tivozanib has demonstrated anti-tumor activity after disease progression on sorafenib. In the ongoing study 902, at the time of the final overall survival analysis for study 301, the median PFS of the 156 subjects who received next-line tivozanib was 8.4 months (95% CI: 5.5, 12.4) and the confirmed overall response rate by investigator assessment was 13.5% (95% CI: 8.5, 19.8). A waterfall plot of individual subjects’ tumor measurements indicates that 74.5% of patients with measurable disease had some degree of tumor regression on tivozanib in study 902. Incidentally, I analyzed the 8 trials results, OS versus PFS, and for the drug arm there is an 85% R squared correlation. For the comparator its only 50%, until you remove the placebo results (which are effect at OS due to cross over and subsequent care) and then it rises to 84%. So PFS results are a great indication of a drugs likely success with OS. And if you google Pazdur and The oncologist alphamed press, you will find a surprising article Endpoints for Assessing Drug Activity in Clinical Trials that echoes everything I’ve been claiming about the OS trial results. It surreal to me that Dr Padzur is attacking this trial based on this article.
I’ve been an engineer in the medical device/pharma industry for a few decades. I don’t develop the drug formulation, but currently I work on the medical devices that they use (combo products) and get exposed to the clinical trial data and the formulators. I’m a black belt, and I use statistics frequently to optimize, validate, or troubleshoot processes. I believe early stage DOEs or clinical trials statistics don’t prove anything, but they are a great tool to make an informed choice, along with technical knowledge, and the confirmation or long term results is where the proof is established. If you use predictive statistics without technical knowledge, you can often be led down the wrong path. I feel the statisticians on the ODAC board did not have enough specific technical knowledge to actually understand the uncontrolled interactions that these RCC VEGFR trials have, and focused on the trial set up rather than if the trial results were good enough to make an informed decision. (They also let the others run with statistically non-significant data to make judgements – should have at least warned that the trends were not strong enough to base conclusions on, especially with all the 2nd order interactions and a subset of the population that doesn’t respond to VEGF inhibitors)
The quote from the temporary stat member about not encouraging cross over in trials just floored me, she doesn’t understand that in this application, getting the new drug is the motivation to join the experiment. If she doesn’t get that, how is she going to get that the VEGFR inhibitors have interactions with each other and subsequent therapy? I was very disappointed in the scientific process of the ODAC committee, it seemed like Pazdur had already decided that he didn’t like the trial, and they made excuses to agree with him rather than digging deeply into what the drug and the trial are doing.
biwinnin63, I appreciate your posts a lot and specifically log on to look for them. Just wondering, do you have any opinion about Mannkind's Afrezza and the Dreamboat inhaler device they are completed testing (Phase III)? I have a lot riding on it and there isn't anyone with a medical device background at the MNKD message board.
Investigator assessment inflated PFS compared to independent assessment. Also, only 18 subjects in one group makes it hard to draw conclusions. My memory is that even in this analysis, which was done, the trend was clearly favoring sorafenib.
They announced another presentation of a subset for treatment-naïve patients. On the Feb 2013 meeting, they already presented a pre-specified subgroup analysis of treatment-naïve patients for metastatic disease, the PFS benefit of tivozanib was12.7 months versus 9.1 months with sorafenib (P=0.037). I don’t think they would just present that again. I hope they break down the OS for treatment-naïve patients, and then censor the crossover patents. I think this data will be the most beneficial, and may not have as much of the 15% low HIF population that doesn’t respond to VEGFR therapy. Those 15% are what is really bringing down the superiority of Tivozanib, as they appear to even be helping the numbers of sorafenib, the unresponsive group is actually better off on sorafenib (but still is screwed) while the responsive group is really better off with Tivozanib.
They used the independent assessment in the reported OS data. You are wrong about the trend favoring sorafenib. There are several subsets that favor tivozanib, especially when the cross over is censored. The Feb 2013 OS presentation with 1:1 (no follow up or grouped by people still on the original randomization) favors tivozanib over sorafenib, but its not presented as the study's censored HR on OS like they did on the original sorafenib trial. The 18 subjects is only the US population, but they have 101 that did not have the sorafenib/tivozanib therapy, and its enough to have a high enough power- it may not be statistically significant, but most of the trials for RCC OS are not statistically significant. It boils down to if the FDA wants to let the EMA approve the drug way before they do, as thier risk/benefit arguement is now weak, and will be shown to be wrong.
There is even a newer FDA article that supports that cross over affected the comparator arm. Relationship Between Progression-Free Survival and Overall Survival Benefit -A Simulation Study.