Not yet a member?
Print This Post
view / write comments
Substantial resources are spent on large, expensive trials to demonstrate small survival advantages in lung cancer. For example, 889 patients entrusted their care to the phase III SATURN trial of maintenance erlotinib, following treatment with a platinum doublet. At ASCO, we learned that these patients were rewarded with a one week improvement in PFS, and at World Lung, we learned that they lived a month longer. The FDA’s advisory committee recommended against approving erlotinib for maintenance therapy based on this modest benefit.
I am among many oncologists who take a two-faced view on small advances in care. On the one hand, I use advances with modest benefits in my clinic. I do not consider financial cost/benefit when adding avastin to platinum-based doublet therapy for patients with non-SqCC; I even use cetuximab, with more modest benefits, at times. While I use these small advances regularly in my clinic, I take a very different perspective when talking to other investigators, where I argue that we should be swinging for the fences, not trying to get more base-hits. Along the way, we will make minor advances and should use them, but the real goal should be to either cure cancer, or develop non-toxic drugs capable of turning cancer into a chronic disease.
With this perspective in mind, agents that can potentially overcome resistance in lung cancer are of particular interest to me and the histone de-acetylase inhibitors fit this mold. As I am working on a trial of a histone de-acetylase inhibitor for small cell lung cancer, these agents were already on my mind when I read the results of a new study in JCO.
Proteins are the tools of the cell do to all the things that it needs to do — examples of proteins include receptors (here EGFR is probably most famous) and the signal transduction cascade that result from signaling to a receptor. In order to make a protein, the cell starts with the DNA blueprint that is housed in a structure called the nucleus. The nucleus can be thought of us a little fortress-enclosed island containing stacks of blueprints in a big lake that is the cell. The DNA is transcribed into stuff called RNA, which travels out of the nucleus into the cytoplasm (which could be thought of like the water in the lake). In the lake, the RNA attaches to a structure called the ribosome (think of it as a small, floating protein factory) where it is translated into a protein. Most things done by cells, whether normal or cancer, depend on proteins, including the bad things that cancer does and including the death cascades that we’d like to activate within cancer cells. All of the new biologic therapies, such as cetuximab and tarceva seek to modify proteins in some way. For example, cetuximab is an antibody to the EGFR and tarceva is a small-molecule tyrosine-kinase inhibitor (a lot of big words to say that it stops the signaling from the receptor within the cancer cell). Cytotoxic (cell-harming or killing) chemotherapies are the traditional agents. They are all essentially poisons that damage the DNA of rapidly dividing cells. Since cancer cells divide a lot, they are more susceptible to these poisons than normal body cells that, for the most part, shouldn’t be dividing.
Unfortunately, we have witnessed that cytotoxic chemotherapy does not cure lung cancer. How do cancer cells evade therapy? One potential way is by going to sleep. If a lung cancer cell happens to be taking a snooze while the chemo is passing by, instead of replicating, it could avoid dying from the chemo. Enter histone de-acetylase inhibitors (henceforth HDACi).
DNA exists wrapped around structures called histones. These can be wound tightly, or they can be looser. When looser, the DNA is more active and prone to transcription and when tighter, it is more quiescent. Thus, HDAC inhibitors seek to “wake up” the cancer cells in time to see the chemo. They also influence the action of several players known to be important in cancer, death cascades, and chemotherapy sensitivity–heat shock protein 90, hypoxia-inducing factor alpha, and alpha tubulin. For more information on HDAC inhibitors and the rationale for trying vorinostat in lung cancer, see Dr. West’s 2007 post.
So, it’s a cool science idea. What next? (see slides 14-16 of this link for a review of the progression from idea to improved cancer therapy) Well, it worked well in cell-cultures, so it went to phase I testing, where Dr. West covered the promising results. Since it was safe and had hints of efficacy, it went to phase II testing and the results were just published.
The trial randomized patients with stage IIIB-wet or IV to either carbo/taxol/placebo or carbo/taxol/vorinostat. The primary endpoint was response rate (RR), and the study was positive—RR increased from 12.5% in the placebo arm to 34% with vorinostat (p = .02). This may have been the primary endpoint of the study, but as we’ve talked about many times before on GRACE, this is a surrogate measure—we don’t care about cancer shrinking per se unless it leads to people feeling better and living longer. Did it? Here are the progression free survival and overall survival curves:
As you can see, there was a trend towards improvement of both progression-free and overall survival. While neither measure met statistical significance, the trial was not powered for these measures. Further, the survival curve does maintain separation as time passes, with even a hint of widening. While not a cure for lung cancer, the magnitude of the potential survival advantage is meaningful, and the regimen is thus worthy of further study. But, what kind of further study is indicated? The authors look at the toxicity data and do something impressive and rare for a positive study: they suggest trying to do better before going to a phase III trial. Instead of replicating the two large tables on toxicity, I’ve made my own summarizing only the toxicities which were worse with vorinostat:
Toxicity……………………..Vorinostat …………………………..Placebo
|
|
Grade 2 |
Grade 3 |
Grade 4 |
|||
|
Nausea/vomitting |
13 |
3 |
0 |
4 |
0 |
0 |
|
Fatigue |
20 |
8 |
0 |
7 |
1 |
0 |
|
Hyponatremia |
0 |
11 |
1 |
0 |
3 |
0 |
|
Thrombocytopenia |
— |
9 |
11 |
— |
4 |
1 |
The authors looked at this toxicity and then asked the consequence of it. A look at the table for reasons for treatment discontinuations is revealing:
First, 34% of patients on the placebo arm stopped therapy because it wasn’t working, compared to only 18% on the vorinostat arm—so the therapy does control cancer growth. Second, 24% of patients stopped therapy early on the experimental arm because of toxicity compared to only 9% on the experimental arm. So, not only did vorinostat toxicity cause direct side-effects (likely reducing quality of life), it also caused many patients to stop a regimen that can help delay cancer progression (which in turn improves both duration and quality of life). Instead of slapping high-fives and moving on to phase III testing, the authors did something courageous—they concluded that alternative schedules combining these drugs should be first tested in the phase II setting to try to get a better regimen with less toxicity. Kudos to them.
Where is vorinostat going in NSCLC? A search of clintrials.gov reveals much of the ongoing work that we can look forward to seeing results from:
There is also exciting work going on with vorinostat in other cancers (including SCLC) and other HDAC inhibitors are undergoing testing in NSCLC. For now, vorinostat in NSCLC may be headed back to the minor leagues for a season, but it holds the promise of returning as an MVP.
Related posts:
Related posts brought to you by Yet Another Related Posts Plugin.
Posted in: First-line treatment, Lung Cancer, Non-Small Cell Lung Cancer (NSCLC), Other targeted therapies, Targeted therapies, Treatment email to a friend
Print This Post
Dr. Weiss: Much as I want homeruns, too, I’m very skeptical about whether these results even represent a base on balls. Can the authors (or you in their absence) explain why the placebo arm (with a standard first-line doublet) had only a 12.5% response rate (far below the norm for first-line doublets, right?)? I’m guessing that the vorinostat arm did well only in comparison to a relatively unlucky placebo group. I don’t have the survival numbers in front of me, but as I recall (and the graphs you attach seem to bear that out), overall survival in the placebo group (looks like around 10 minths for the median) was somewhat lower than the usual first-line norm as well. What are your thoughts on this?–Neil
Neil,
I like the way that you’re thinking. Two large studies employing carbo/taxol come to mind to address your question about the placebo group—ECOG1594 and ECOG4599. ECOG1594 compared four different chemo regimens to each other—cis/taxo, cis/gem, cis/docetaxel, and carbo/taxol. Response rate and overall survival were similar on all arms. 290 patients were assigned to the carbo/taxol arm and their response rate was 17% with overall survival of 8.1 months. The other is ECOG 4599, which showed improved response rate and survival with carbo/taxol/avastin compared to carbo/taxol. In this trial, 444 patients were assigned to carbo/taxol with response rate of 15% and overall survival of 10.3 months. Patient characteristics were a little different for all three studies and they took place at different points in time. However, they create a context where a 12% response rate and 9.7 month overall survival are reasonable for carbo/taxol.
We must remember, however, that even though there was a randomization, this was a phase II, not a phase III trial. The point of the trial is not to lead to approval and use of carbo/taxol/vorinostat, but rather to demonstrate that vorinostat is active and worthy of future study. An almost tripling of response rate and a more than three month improvement in survival are clinically meaningful and for me, seem to provide such a signal.
You asked about its comparison to other doublets so let’s look at other recent positive trials. In ECOG4599, the winning arm, carbo/taxol/avastin, lived 12.3 months. In the Scagliotti trial, patients with adenocarcinoma reached a median overall survival of 12.6 months and patients with SqCC 10.8 months. In the phase II Madeit study using ERCC1 and RRM1 to target chemo, median survival in the targeted arm was 13.3 months. So I do think that in this context, too, 13 months is promising. I stand by my opinion that we have a base hit here.
And as for the home-run? One of GRACE’s members who has my respect chastised me privately early in my time here for using the “cure” word too much. My mentor thinks that I’m too picky in demanding that studies shoot for cure. Both of them have reasonable perspectives—I don’t want to give false hope to patients whose lives are at stake and I don’t want to call hard working researchers who have dedicated their lives to improving cancer care failures for not finding the cure. However, I do think that cure is what we need to at least work towards. And I have hope that HDAC inhibition could be one part of ultimately getting there. Surely cancer cells have many tricks up there sleeves, but perhaps we’ve learned a way around one of them.
Dr. Weiss: Thanks for the correction. I thought I remembered that typical response rate for doublets was in the 20-30% range. Given the figures that you give for comparison, the results in the vorinostat arm are more impressive than I thought. Same for the survival differences (I again overestimated the median survival with typical doublets by a bit), though as you note, differences in patient characteristics (especially performance status) would have more of an impact on survival differences than on response rate. However, given the relatively small size of the sample (and I recognize that it wasn’t powered for statistical differences; it would take something huge like 34% vs. 12.5% to reach statistical significance), and even though it was randomized, I would want to see (and you probably did in the actual article) a comparison of patient characteristics between the two arms (for instance, were they stratified by performance status? If not, was there any real difference in performance status between the two arms?).
Still, I’ll agree that I overstated the case, and this is (happily) a solid base hit.–Neil
The question of similarity of the two arms is key–if they were different, especially in such a small sample size, that could account for some of the differences observed. Here’s the data:
Characteristic vorinostat arm(%) placebo arm(%)
Male sex 61 63
Median age 64 66.5
Ethnicity white 85 84
Ethnicity black 8 13
Ethnicity other 7 3
Brain mets 16 16
ECOG PS0 39 41
ECOG PS1 61 59
Stage IIIb 8 3
Stage IV 92 97
Histo NOS 44 41
Adeno 29 34
SqCC 19 19
Large cell 3 3
Other 5 3
For a randomized phase II, I would say that this is well balanced. The only slight difference was in stage, but this was not nearly enough to account for the large difference seen.
That’s incredibly well-balanced. Clearly, that doesn’t explain the big difference in response rate. Thanks!–Neil
Just adding some thoughts on response rate in our trials: many of us were unpleasantly surprised by seeing response rates under 20% for our established standard treatments. However, response rate is not as objective as we’d like to think, and it’s very subject to methodology. The newer, RECIST criteria that we use are more stringent than some older standards, and many trials include a confirmatory scan to declare someone as having a response. The actual numbers vary enough from study to study that I don’t think we can read too much into comparisons between one trial and another.
Even survival, which is a very “hard” endpoint not subject to debate or revised definitions, nor is it only detectable when you look for it, unlike scan-based endpoints like response rate or progression-free survival, is not readily translatable from one trial to another because of sometimes significant differences in the trial populations. For instance, the trials that only include “avastin-eligible” patients, such as ECOG 4599 and others, exclude patients with squamous tumors, brain metastases, and several other medical issues. The median survival with carbo/taxol and NO avastin on the ECOG 4599 trial was 10 months, which was 2 months more than the exact same regimen in a trial done by the exact same group (the ECOG 1594 trial that Dr. Weiss noted above). Why? Probably because patients less likely to do well were excluded due to restrictions on who could receive Avastin.
Meanwhile, many European or “worldwide” trials include not only patients who would be agreed upon as “advanced NSCLC” in North America, but also patients with “dry IIIB” NSCLC who would be generally be treated with curative intent with chemo and radiation in North America but are, in my opinion, undertreated when consigned to systemic therapy alone. But they also tend to have a longer survival, so I think we need to be cautious about presuming that a regimen that gives a median survival of 13 months in one trial is better than a regimen that gives a median survival of 11 or 12 months in another. There could be very real differences in patients and treatment patterns.
All of this is to say that a randomized trial in which patients with the same eligibility requirements are treated by the same investigators is very helpful. Yes, a randomized phase III trial is designed to declare a superior approach of one over another, while the randomized phase II trial isn’t able to do that, but the relative differences between arms is certainly valuable, more than comparing results across clinical trials.
-Dr. West
Dr. Weiss, I like the part you described the cell environment and how DNA, RNA, proteins live their lives in their paradise island. Your words sound well like a Flash movie visually demonstrating the process. I am often bothered by the cell biology terms when reading scientific reports or journal articles. So now Molecular Cell Biology by Lodish et al. is on my bookshelf. Any even simplier and shorter book you can recommend?
Neil, you seems have mastered the cautions on reading an report on cancer. : ) I llike your questions.
I also feel that Dr West’s comments supplement his aritcle, Miracles and deadly killers: putting media reports on cancer in perspective, in section Cancer 101. People who have not read it may wanna check it out.
I’m very pleased with this conversation because it’s brought out some of the key issues not just about carbo/taxol/vorinostat, but about how to evaluate trials and how to best move forward. My mentor once told me that he could make licorice look good in a phase II trial through patient selection. I think that one of the key strengths of the randomized phase II design is that allows you to study whatever population you want, but still control for this bias. The cost, however, is patient # and there are real costs to this. This trial required 94 patients to entrust their care to it. The trial was positive, so presumably most benefited, or at least had as much chance of benefiting as they would have from off-protocol therapy. However, this was not known before the trial started and so it is worth noting that the 53 adenocarcinoma patients and 6 large cell patients might have been avastin eligible off study. The trial was likely expensive–it costs many thousands of dollars per patient to rigorously conduct this kind of study (a very rough estimate might be $5-6,000 per patient). The authors hypothesized an increase in RR from 25% to 50%, with an alpha (chance of falsely finding a difference if not really present) of .1 (10%). A “back of the napkin” (actually, SWOGSTAT) calculation shows that you only need 17 patients in a single-arm phase II to do this because of the large effect size postulated.
But wait, you might say–really there was an increase in RR from 12.5 to 34%, not 25% to 50%–if a single arm trial had found that RR of 34%, you would have called the trial negative and you’d be wrong. This criticism is valid. Dr. West outlined well the issues between comparing outcomes between trials. This heterogeneity and unpredictability is one of the weakness of the single arm trial, and in this case, the randomized phase II design “saved” it.
Since this is a forum, I’m going to pretend that I’m Neil or Dr. West for a moment and argue against myself. Perhaps the authors could have done a single arm trial, but made a more rational hypothesis. In the conclusion to their paper, the authors argue that the RR and OS in the control arm were consistent with the literature; so maybe up front they could have chosen a lower RR for vorinostat to beat and still had a good single arm design. Back to the stats–if the authors postulated an increase in RR from 15% (the midpoint of the two studies I listed in my post below) to 30% (a doubling) with all other parameters being the same, this could have been done with 35 patients.
I’m conducting this discussion with Dr.West and Neil and this schizophrenic argument with myself as much to talk about this study as to talk about trial design. At the end of the day, I do think that the authors designed their study well and interpreted their results honestly. They avoided the major pitfall of the randomized phase II design–trying to pretend that you’ve done a pseudo-phase III and declare premature victory. All that a phase II study, whether randomized or not, can show is whether there is enough signal for efficacy for further study. Here, the investigators did so very well. They were also impressively critical of their own results, demanding less toxicity, and hopefully with it, better efficacy. Hopefully, a slight change in the regimen could improve it, leading to a better phase III study and ultimately a better treatment. I look forward to one day writing about this kind of success on GRACE.
8 Comments
Login (Must Be Logged In To Comment)