Those who have followed my writings over time will know that I haven’t been inclined to adopt a reflexive strategy of ordering molecular testing without good evidence that having this information will improve outcomes. Testing tumors for EGFR mutations is advocated by a vocal minority of lung cancer experts in Boston and New York City, but this hasn’t been advocated by the broader lung cancer community yet, or adopted as routine clinical practice. Although I’ve felt that we’ll have enough evidence to support broad use of molecular variables to individualize treatment for lung cancer patients in the next few years, I haven’t felt that they have a role yet, but I’ve now seen some results that I believe are enough for me to change that conclusion.

The trial results that have changed my impression are based on the IPASS trial that I previously described (see prior post), in which 1217 previously untreated East Asian advanced NSCLC patients with a history of never-smoking or having quit 15 or more years previously and having smoked less than 10 pack years (packs per day x years smoked) were enrolled and randomized to receive either chemotherapy with carboplatin/taxol (paclitaxel) or the oral EGFR inhibitor iressa (gefitinib). This was a remarkably different population from the patient population we typically see in North America. This group was far younger (median age 57, vs. a median age of about 70 for those newly diagnosed with NSCLC, and about 63 for typical North American NSCLC trials), nearly 80% women, and 94% never-smokers (not surprising if you need to have smoked little or nothing to enroll), and nearly all with adenocarcinomas. If there was ever a group of patients who would do well with an EGFR inhibitor, it’s this population. But they didn’t require patients to have a mutation, assuming instead that these clinical variables would provide a close approximation.

The primary endpoint that they evaluated was progression-free survival (PFS), and the curve highlights that there were two different populations in the trial with a major crossover right AT the median PFS, which was essentially identical between the two assigned groups. But the median doesn’t tell the story properly: some patients clearly did far better with chemo (the left side of the curve, while others did far better with iressa (the right side of the curve).

The defining factor appears to be EGFR mutation status. Importantly, one KEY limiting factor is the availability of tissue: among the enrolled patients, 85% gave consent for tissue analaysis, and 56% both gave consent and had tissue available, of whom many didn’t have enough for proper testing. At the end of all of this, a total of only 36% of the trial population were able to have EGFR mutation results reported. This attrition highlights a significant problem with having tissue analyses potentially guide or treatment decisions. The mutation rates in different populations showed that the mutation rates were in the 60% range for most of the clinical variables, except for being more common in women than in men, and in older than younger patients. Although the sex-based difference has been reported frequently, I’m not sure if the association with age is real or just a random finding.

Looking at how the patients with EGFR mutations did, compared with those without EGFR mutations, we see completely different results in PFS. Those with EGFR mutations did significantly better with iressa, while those without EGFR mutations did significantly better with chemotherapy.

The response rate (RR) differences were also striking when reviewed by EGFR mutation status, looking largely the same way. Those with EGFR mutations had a remarkably better RR with iressa, while those without them had a hugely better RR with chemotherapy:

As I mentioned in the prior post on this study, it’s also interesting that those patients with EGFR mutations and assigned to chemo had twice the RR of those given first line chemo but without EGFR mutations. Nevertheless, when RR results are pooled for the overall trial, there aren’t striking differences, even though the clinical population of Asian mostly never-smokers definitely enriches for those patients with mutations. The major difference in RR based on EGFR mutation washes away, and we might have missed an important conclusion had the investigators not dug deeper and checked the molecular variable as well.

So my conclusion is that molecular markers trump clinical factors. I’ve been relying largely on factors like smoking status, race, and tumor histology to guide my recommendations, but the IPASS trial shows that even enriching for the patients most likely to benefit doesn’t seem to correlate nearly as well as using a molecular definition, if you can get it -- which is a real factor when we don't routinely obtain much tissue for patients with advanced NSCLC. I’m particularly struck by the very disappointing results in clinically selected patients who don’t have the EGFR mutation – the very disappointing PFS curve and RR (1%, mind you) of the patients assigned to iressa but without EGFR mutations. Because of this, I will be more inclined to consider checking for EGFR mutation status before recommending that a patient start with first line oral EGFR inhibitor therapy, regardless of their smoking history, race, and other factors. I’m reassured that the overall survival (OS) curves don’t come out different thus far on the IPASS trial, but it’s possible that a mutation-based difference exists but hasn’t been seen yet (we haven’t seen a breakdown of OS by mutation status yet).

I’m leading a couple of clinical trials that give tarceva (erlotinib) and avastin (bevacizumab) together to patients with bronchioloalveolar carcinoma (BAC) in one trial (S0635) and in never-smoker lung cancer in the other (S0636). Between these trials using tarceva instead of iressa, adding avastin, and treating a North American population, the results could be very different from those of the IPASS trial, but the strength of the evidence is in the process of shifting my practice patterns outside of a clinical trial. I should also emphasize that these findings don’t suggest at all that we need to check EGFR mutation status before offering oral EGFR inhibitors to patients who have already received 1-2 lines of therapy. These patients aren’t potentially substituting an EGFR inhibitor for a more common standard alternative, and the evidence does NOT suggest that you need to have an EGFR mutation to benefit at all, particularly against a placebo.

I’ll be extremely interested in new trial results of molecular markers as they emerge. At least for NSCLC, I think we’re on the cusp of re-writing the treatment algorithms for advanced disease, but we may find that availability of tissue is a major hurdle to overcome over the next few years.