Over the past several months the topic of so-called maintenance therapy in advanced NSCLC has been one of the most timely and controversy questions in lung cancer. We've had posts here covering a trial testing immediate vs. delayed taxotere (docetaxel), the updated results of the maintenance alimta (pemetrexed) trial, the SATURN trial of maintenance tarceva (erlotinib), and the ATLAS trial of maintenance tarceva vs. placebo combined with maintenance avastin (bevacizumab). All of these make an argument for an immediate transition from first to second line therapy, but these studies highlight several other conclusions that we might make. 1) Treatment after first line is effective: Recommending additional treatment after the first line of therapy has been common practice for less than a decade, really becoming common practice only after taxotere was shown to improve survival (very modestly) compared with supportive care alone or alternative, less effective chemotherapy. These studies included a broad range of patients and showed rather modest survival benefits overall, actually without a clear improvement in median survival with chemotherapy. In contrast, the maintenance therapy trials with chemotherapy have shown median survivals in the range of a year, which is much higher than the 8 month range seen in earlier second line therapy trials. Why? This is almost certainly because these trials filter out the patients who show progression, serious side effects, and/or a significant decline in their performance status over the first four cycles of first line therapy. If you remove most of the people who are likely to do poorly, you're left with a population that is likely to do very well. We've never seen survival numbers like these for previously treated patients, but they're a reflection on additional therapy being impressively effective in the patients who have stable disease or a response with first line therapy. 2) Clarification of who really benefits and who doesn't: If we identify a group likely to benefit far more, there must also be a group who benefits much less from treatment after the first line. As noted above, the patients included in the maintenance therapy trials appear to derive a very convincing survival benefit, far more than we typically see from second line therapy in trials that enroll a broader population (including patients progressing in fewer than four cycles). This suggests that the patients with early progression receive a considerably lower magnitude of benefit from further treatment. We may do it, since the trials that included them show a modest benefit overall, but this needs to be weighed against side effects and other costs of treatment that we can't expect to be a blockbuster. 3) Timing as "maintenance" may not matter, but treatment only helps if you get it: The main debate right now about maintenance therapy isn't whether further treatment is valuable, but really just whether it's critical to get it as an immediate transition from first line therapy. I don't believe any of these trials make that argument effectively, because the trials that have shown a survival benefit (significant with alimta, a strong trend with taxotere, and an undefined but reportedly statistically significant improvement with tarceva on the SATURN trial) may well be attributable to an absolute difference in the proportion of patients on each arm who actually receive effective therapy (ever). In the trial of immediate vs. delayed taxotere, the survival was the same when immediate treatment was compared to the subgroup of patients on the delayed taxotere arm who actually got it (which was only two thirds of the patients assigned to the delayed chemo arm). In the maintenance alimta trial, only 19% of patients on the placebo arm ever got alimta, and only two thirds got any treatment after progression, while 100% on the maintenance placebo arm got their treatment. We also know that even fewer than two thirds of the placebo recipients on the tarceva trials ever got tarceva. So these trials aren't really purely testing the timing of treatment as maintenance vs. waiting until progression: they're actually testing everyone getting treatment after first line versus only about two thirds of patients getting treated. If immediate therapy isn't pursued, the same benefit may be achieved if patients are followed diligently enough to initiate the next treatment before they experience a clinical decline. 4) Survival differences may be closely related to how many patients crossed over to the effective tretment: The maintenance therapy trial that showed the strongest survival benefit, with alimta, also had by far the lowest crossover rate to the study treatment (19%) among the people who weren't assigned to it as maintenance therapy. We don't know the magnitude of the overall survival benefit with the SATURN trial, but it may also reflect fewer patients on the placebo arm getting tarceva at any point. Progression-free survival may well be the clearest and only convincing benefit with maintenance therapy, while overall survival may really reflect how many people end up receiving effective treatment after first line therapy at all. And the value of progression-free survival isn't entirely clear if most or all patients can end up living just as long if they get start their treatment after a break. My overall conclusion is that maintenance therapy is generally a good idea, since patients who show stable disease or a response after four cycles of first line chemo receive particularly benefit with later treatment, and whether it's because of the timing of maintenance therapy or the fact that they actually get their intended treatment more reliable, they live longer. I don't think every patient must get maintenance therapy, since I honestly believe that patients will do just as well if they just get their next treatment before they decline so much that they miss the opportunity. For a minority of patients, those with a very good response and/or those with indolent disease (and most experienced oncologists can identify the general pace of a particular patient's cancer after caring for a person and seeing the behavior of their cancer for several months), a break from treatment is likely to be perfectly appropriate. Some patients will go 3-6 months or longer before they progress, and for those patients, an immediate transition to the next line of therapy is effectively overtreatment. So in the end, we come back to the idea that the optimal strategy is going to be an individualized approach, though I'm more likely to favor a treatment that I expect will be well tolerated as maintenance therapy in the majority of patients.