There was a very sobering piece in NEJM by the FDA last month in which the authors try to explore what went wrong with the Keynote-183, Keynote-185 and checkmate 602 trials testing PD-1 inhibitors combinations with pomalidomide or lenalidomide and dexamethasone in multiple myeloma. Interim analysis of Keynote 183 and 185 revealed detrimental effects on overall survival (OS) with hazard ratios of 1.61 and 2.06, not explained by differences in toxicities alone. The checkmate 602 trial was also halted in light of these findings and also showed higher mortality in the nivolumab combination arm.
In the thoughtful NEJM piece, the authors make at least three important points. First, they question why these PD-1 inhibitors were tested in combination despite their having limited single-agent activity. In fact, a couple of years ago, Vinay Prasad and I asked the same question: why are novel cancer drugs being tested in combination despite having limited activity as a single agent? We found that these drugs, even when ultimately approved, provide relatively low value and recommended that drugs with poor single agent activity not be tested in combinations unless there are specific reasons to expect synergy.
The second important point in the article is that many cancer drug approvals are lately based on durable response rates in single arm trials without a control group, a situation in which it is difficult to evaluate the safety and efficacy of drug combinations. Indeed, without an RCT, the oncology community would never have known these signals of detrimental effect. If the FDA had approved these PD-1 inhibitors in multiple myeloma on the basis of non-randomized trials, which it often does in other oncology contexts, who knows how long it would have taken to recognize the increased mortality in patients—and at what cost. This is another reason why we need RCTs more now than ever. Finally, the authors point out that these PD-1 inhibitors in multiple myeloma were directly advanced to phase 3 trials after phase 1 trials were completed, without phase 2 information. Indeed, in a recent paper, Alfredo Addeo and I showed that a substantial percentage of drugs that fail in phase 3 trials do not have supporting phase 2 data.
PALOMA-3: For the birds?
How often do you see an absolute difference in median survival times reported in a trial publication? For example, atezolizumab plus abraxane improved progression-free survival versus abraxane alone in the IMPassion130 trial but the difference was only 1.7 months. This difference was not reported which is not a big deal, because we all know how to add and subtract. But in the report of PALOMA-3, an absolute difference in overall survival is reported. Absolute difference in median overall survival with palbociclib plus fulvestrant versus fulvestrant alone was 6.9 months. Why was this difference reported? One thought: in the Kaplan-Meier graph, we usually see the stratified HR reported, but in this trial report, unstratified HR is also provided. Because the difference in overall survival wasn’t significant but looked numerically impressive, we had the rare privilege of seeing an absolute difference reported, as well as the unstratified hazard ratio as well (although I don’t know what to do with that information, except to recognize that the p-value dropped from 0.09 to 0.05).
Old is GOLD
Although I rarely comment on surgical oncology, I couldn’t let this opportunity pass to make a point about the harms of adopting medical practices without solid evidence, something Vinay Prasad and Adam Cifu covered in their book Ending Medical Reversal. The first sentence of the abstract of this study reads: Minimally invasive surgery was adopted as an alternative to laparotomy (open surgery) for radical hysterectomy in patients with early-stage cervical cancer before high-quality evidence regarding its effect on survival was available. Why? Why do we often widely adopt interventions in medicine without evidence? Anyway, better late than never, 2 studies—one RCT and one cohort study—have been done and unfortunately both show harms with modern minimally invasive surgery compared with the classic open surgery. Sometimes excitement over a “new procedure” can be iron pyrite.
Newer is Scarier
I don’t remember ever seeing a hazard ratio of 5 for survival outcomes in cancer, which means that patients in the experimental arm have a five times higher hazard of dying than patients in the control arm. Unfortunately, that’s the hazard ratio in this RCT of cetuximab-radiotherapy versus cisplatin-radiotherapy in patients with low risk oropharyngeal cancer that were HPV positive. This finding was consistent with another similar non-inferiority trial in which the hazard ratio remained over 1. The toxicity profile wasn’t much different between the two regimens, although that was the main reason for doing the trials—namely, the assumption that cetuximab had better side effect profile than cisplatin. Kudos to the investigators for conducting these trials that now clearly show that HPV-positive patients with oropharyngeal cancer are harmed by substituting cisplatin with cetuximab. I have always had reservations about the role of cetuximab in head and neck cancers. Even for locally advanced disease, cetuximab was tested only in one trial while cisplatin has been tested in multiple trials. The costs of cetuximab are very high compared with cisplatin. Another similar anti-EGFR antibody panitumumab has failed in two different RCTs in this setting. That’s why I included cetuximab-RT for locally advanced head and neck cancer as an example of low-value practices in oncology in my 2017 paper. These recent two trials in HPV positive patients with oropharyngeal cancer also now show that the argument of reduced toxicity with cetuximab is probably not very valid. So, why exactly should oncologists use cetuximab in any head and neck cancer? Or at least, why not test cetuximab against cisplatin for all locally advanced and metastatic head and neck cancers because I am not aware of any such head-to-head (or neck-to-neck!) comparisons yet.
Is Vitamin D supplementation VITAL?
The central theme of the entries this month seems to be “When RCT was finally done, the benefits of intervention X we assumed were gone.” After several observational studies, we finally have a placebo controlled RCT of Vit D supplementation. Results from this VITAL trial show that vitamin D supplementation has no preventive effect on cancer, cardiovascular deaths or death from any cause. Two key messages here: first, one RCT can come to a different conclusion than years of observational studies and second, publicly-funded RCTs help answer important clinical questions.
Pay Heed to STAMPEDE
It’s hard to imagine any trial cooperative group contributing as much information to treating any cancer than the STAMPEDE trial group informing prostate cancer treatment. The STAMPEDE trial group answers important questions with smart trial designs, robust statistical methodology, and laudable reporting. The latest results to come from the STAMPEDE trial team describe the effect of local treatment of primary tumor with radiotherapy for newly diagnosed metastatic prostate cancer patients. The researchers found that radiotherapy to the prostate did not improve overall survival for unselected patients but had an 8% increase in 3 year survival rates for patients with low metastatic burden. Again, the STAMPEDE team models how subgroup analyses should be done and reported in clinical trials. They use OS as the primary endpoint, prespecify the subgroups, appropriately power the trial for subgroup analyses, perform limited number of prespecified subgroup analyses in which the direction of effect has a rationale, perform consistency analyses to increase the confidence in subgroup results and yet, refrain from highlighting this subgroup effect in the abstract. I read a lot of poorly planned, poorly conducted, and poorly reported trials in oncology that make me feel as if “I am glad I am not a part of this trial”; but time and again STAMPEDE reports make me feel “I wish I were a part of this trial team”.
Let me take a selfie
In fact, please allow me to take a couple of selfies. First, our research—which we had also previously presented at ASCO this year—on quality of life information in cancer drug trials has been published. We show that nearly half of cancer drug trials don’t include quality of life as an endpoint and of those that do, nearly a quarter don’t report on prespecified quality of life endpoint. We also showed that the correlation between progression free survival and quality of life was poor. In another paper I published in the BMJ with my colleagues from Japan, we examine how frequently cancer drug trial reports use terms that downplay toxicities of the drugs (such as “the toxicities were acceptable”) and discuss why the use of such language is problematic.
Dr Gyawali is a research fellow at Program On Regulation, Therapeutics And Law (PORTAL) at Brigham and Women’s Hospital/Harvard Medical School. The opinions expressed herein are his own. You can read his previous blogs here.