On October 15th there was a press release that, as far as I can tell, went almost entirely unnoticed. News outlets reported that Roche (owner of Genentech, the maker of Avastin (bevacizumab)) reported to OSI Pharmaceuticals (the maker of Tarceva (erlotinib)) the final overall survival results from the ATLAS trial. There have been a number of posts on the ATLAS trial in the past, including one by me this summer after the primary results were presented at ASCO. To review, ATLAS was a phase III trial that randomized patients with advanced NSCLC, all of whom had completed 4 cycles of first-line chemotherapy plus Avastin, to either maintenance Avastin alone (the standard arm) or maintenance Avastin plus Tarceva. ATLAS was one of two maintenance Tarceva trials, the first of which was called SATURN and did not include Avastin. The SATURN trial and ATLAS trials both met their primary endpoints, significantly prolonging progression-free survival (PFS). However, much to the surprise of many people (like me), the SATURN trial was also reported at the 2009 World Conference on Lung Cancer to be positive for an improvement in overall survival (OS). The ATLAS overall survival data was not yet mature at the WCLC, so we were left to see if there would also be an OS benefit from the combination of maintenance Avastin and Tarceva (or what I like to call “Almost the Most Expensive Regimen Ever”). This week, however, without a hint of fanfare, we apparently got our answer. I can’t even find the original press release, but these results were confirmed to me by a contact at OSI.

During a routine data sweep of the ATLAS trial, Roche concluded that ATLAS could not achieve a statistically significant OS benefit, as had been seen in SATURN. Neither study was powered to detect significant differences in overall survival, which is what made the results from SATURN somewhat remarkable. OS was only one of the secondary endpoints, albeit the one most people were interested in. So we don’t know any more details than that, but my guess is that this should be enough to say that the combination of Tarceva and Avastin probably won’t achieve regulatory approval in the maintenance setting. The more important question is WHY did this trial not achieve an OS benefit when maintenance Tarceva alone was able to show such a benefit? For one thing, the OS benefit in the SATURN trial was quite modest, with an improvement in median OS of exactly one month, so it wouldn’t take much statistical fudging to see that disappear. The ATLAS trial was about 15% smaller than SATURN, with only 768 patients compared to 889, so the trial probably had insufficient power to detect the same magnitude of benefit seen in the larger trial. It is also possible that maintenance Avastin alone has some level of OS benefit, which would possibly diminish the magnitude of difference between arms compared to SATURN (which only had to beat the placebo arm). A final possibility is that SATURN was actually the fluke and that maintenance Tarceva has minimal value, although I think that all of the positive maintenance trials over the last couple of years have definitively proven that this strategy does improve PFS and probably OS too, so I don’t think that's it. It will be very interesting to see if maintenance Tarceva gets FDA approval based upon SATURN, which I think it probably will. If it does, docs will probably use Avastin and Tarceva together a’la ATLAS despite this negative result. Which I’m sure won’t increase health care costs at all!