Over the past several years, oncologists have experienced an evolution in the way we think about stable disease, at least in the context of lung cancer.  Historically, oncologists have graded our work by looking at response rates, or the percentage of patients with tumor shrinkage of 50% or more of their lesions as measured in two dimensions.  We considered a drug or combination to be “active” if it had an “objective response rate” (ORR) of 15% or more, and we generally discarded approaches that had a lower ORR than that. 

   But largely as a response to the recognition that some targeted therapies may potentially halt tumor growth but not significantly shrink or kill a cancer, we’ve come to acknowledge that achieving stable disease can translate to a valuable clinical benefit for patients (as in, patients are likely to live signfiicantly longer).  Of course, we all want to see tumors melt away, but at least in the setting of advanced lung cancer, having the cancer shrink just a little or just stay the same for several months at a time is a real improvement over the natural history of what the cancer would do, which is grow steadily over time.  We also often see that patients who have chemo with or without radiation before surgery sometimes show no significant tumor shrinkage, but after surgery we find that the tumor contains mostly or only dead tumor cells.  So again, stable disease underestimated clinical benefit of treatment. 

First, we need to remember that what is called ”stable disease” (SD) in the formal oncology grading system is pretty diverse.  This can range from mild tumor growth to modest tumor shrinkage that doesn’t meet the set criteria for a response.  In the last few years, oncology researchers have started to present “waterfall plots”, which I’ve shown in a few prior posts.  In them, the vertical bars show the change in the total volume of disease measured on CT scans during treatment, going from most growth on the left side to most shrinkage on the right side, and the horizontal line in the middle being no change at all.  Bars extending up above the line (on the left side) mean the cancer grew overall, and bars extending below the line (on the right) are tumor shrinkage.  But you can see that in this trial (with sorafenib/nexavar, which had no official responders) some patients had a lot of tumor shrinkage but didn’t count as a responder, potentially because their response was not confirmed in another CT scans or some other exact criterion of response on the trial that they didn’t meet.  Other patients with SD had modestly increased tumor volume.  Here’s the plot:

 (click to enlarge)

So we can see that our classification system includes a broad range for what falls under SD, from modest growth to good but less than 50% shrinkage.

   We began to recognize that there could be a benefit in SD as we began to study second-line treatment of NSCLC more over the last 6-8 years.  Dr. Frances Shepherd in Toronto led a trial of 104 patients who had previously received one line of chemo for advanced NSCLC to then receive taxotere/docetaxel or supportive care (abstract here).  This was one of the key trials that led to the US FDA approving it for second-line therapy, which was based on the improvement in median survival (7.0 vs. 4.6 months) and one-year survival (29% vs. 19%) for taxotere compared to supportive care alone.  Although the ORR was a mere 6% overall, Dr. Shepherd noted that 43% of the patients had SD, so half of the patients had “disease control”.   While this was not really a commonly reported endpoint in 2000, Dr. Shepherd raised the point that the improved survival with taxotere must be from more than a mere 6% response rate, likely from the nearly 50% of patients who had at least SD, now also known as “disease control” (sometimes also known as “non-progression”).

   The experience with tarceva compared with placebo (abstract here) also highlights that SD must translate to improved survival.  EGFR tyrosine kinase inhibitors like tarceva were known to occasionally induce remarkably and sometimes very long lasting tumor shrinkage in a minority of patients.  When the important BR.21 trial (reviewed in a prior post, and abstract is here) reported a survival benefit for recipients of tarceva compared to a placebo, but there was only a 9% response rate with tarceva, one central question was whether the whole benefit for the tarceva group was propelled by a phenomenal benefit from a very small number of patients.  Some complicated mathematical models were developed that actually showed that is would be impossible for such a small proportion of patients, the 9% that responded significantly, to account for the survival benefit.  In fact, there is a borderline significant improvement in survival with tarceva even if you remove ALL of the responders from the analysis:

And as I discussed in my prior post on the differences with tarceva looking at different patient subgroups, women are more likely to have a significant response than men, and folks with adenocarcinomas are more likely to respond than those with squamous tumors (as reflected by the higher RR in those groups as shown on the following figures), but all of these groups have the same survival benefit with tarceva (as shown by the gaps in the curves and the “hazard ratios” all about the same, where a decimal less than one measures benefit of tarceva vs. placebo):

And to really illustrate the point that you can improve survival without having significant tumor shrinkage, we have the group of men with squamous cancers and a smoking history, who are the ones we’d least expect to show a response.  In fact, the ORR for that group was less than 5%, but they had an improvement in survival with tarceva that was every bit as good as that seen for the trial as a whole:

So I think it’s settled.  People can live significantly longer by achieving SD on treatment.

  From all of this evidence in the last few years, most of the experts have come to accept that the disease control rate (DCR) is a truly valid endpoint, and that we’re potentially discarding useful anticancer therapies if we require a 15% ORR cutoff, especially in settings where it can be so hard to achieve major tumor shrinkage, like previously treated lung cancer. Improved survival is the most important factor (at least if comes with reasonable quality of life), and that can come without a significant response.  

   Tomorrow I’ll cover some new work that shows how an early look at DCR in clinical trials is able to predict how well patients on that trial will do in terms of longer-term survival.