Study suggests NCCN off-label recommendations valid based on available evidence
Click Here to Manage Email Alerts
Although National Comprehensive Cancer Network guidelines frequently recommend use of drugs in settings beyond their FDA-approved indications, the strength of evidence for such recommendations is robust, according to results of an analysis published in Annals of Oncology.
Researchers found a significant subset of these drugs later receive FDA approval or support of the off-label use by randomized controlled trials.
“Off-label drug use is a frequent component of care for patients with cancer in the United States,” Razelle Kurzrock, MD, director of the Center for Personalized Cancer Therapy and Rare Tumor Clinic at University of California San Diego Moores Cancer Center, and colleagues wrote. “Recommendations without randomized controlled data are often for mechanism-based drugs with high response rates in rare cancers or subsets without effective therapies.”
As HemOnc Today previously reported, a study published last year in British Medical Journal by Wagner and colleagues examined the differences between recommendations by NCCN guidelines and FDA approvals between 2011 and 2015 for 47 drugs intended for adult hematologic or solid cancers.
Investigators, including Vinay K. Prasad, MD, MPH, hematologist-oncologist and assistant professor of medicine at Oregon Health and Science University and a HemOnc Today Next Gen Innovator, found that FDA approved these therapies for 69 indications, whereas NCCN recommended those same therapies for 113 indications. The researchers concluded that the evidence for the additional indications recommended by NCCN was weak.
For the current study, Kurzrock and colleagues sought to determine the strength of the evidence by performing an in-depth reanalysis of the 44 off-label recommendations listed in the NCCN Clinical Practice Guidelines in Oncology. Their results showed that, of the 44 off-label recommendations, 14 were later approved by the FDA and/or supported by randomized controlled trial data.
An additional 13 recommendations were either minor extrapolations from the FDA label (n=8) or were on label (n=5). Of the 17 remaining extrapolations, eight were for mechanism-based agents applied in rare cancers or subsets with few available treatment options, with a median response rate of 43%. Additionally, seven recommendations were based on nonrandomized data with response rates greater than 50%, according to the researchers.
Ultimately, Kurzrock and colleagues concluded that the strength of evidence supporting NCCN recommendations is “robust,” directly contradicting the conclusion of Wagner, Prasad and colleagues.
In response, Prasad told HemOnc Today that “the paper by Kurzrock [and colleagues] contains a number of methodologic weaknesses and ultimately reaches an incorrect conclusion. Our conclusion remains that the majority of NCCN recommendations are based on weak evidence, per the long-established hierarchy of medical evidence developed by David Sackett, OS FRSC, and colleagues.” To read a full response from Prasad to the paper by Kurzrock and colleagues, click here.
HemOnc Today spoke with Kurzrock about the study, the potential implications of the findings and the directions she believes subsequent research should take.
Question: What prompted this research ?
Answer: Our analysis stemmed from a prior article by Wagner and colleagues that stated that the NCCN guidelines made recommendations with low levels of evidence. We wanted to carefully go through the levels of evidence and look at what had been critiqued to see if the critique was valid.
Q: How did you conduct the study?
A: We examined the prior paper by Wagner and colleagues published in BMJ. We looked at all the tables and supplements, and went through all the indications in the NCCN guidelines. We wanted to determine how the recommendations were made by NCCN and to match up what we saw in the guidelines with what had been reported previously.
Q: What did you find?
A: Based upon our findings, we did not agree with what had been previously reported by Wagner and colleagues. First, many of the reported extrapolations were either minor or fully on-label. In addition, 32% of the extrapolations were later FDA approved and/or validated by randomized trials, suggesting that NCCN guidelines were getting some of these drugs with unquestionable activity to patients with lethal cancers faster.
For 15 of the 44 extrapolations, however, our disagreement with Wagner and colleagues is based on principle. Wagner and colleagues imply that drugs should be available to patients only when randomized controlled trials prove their value. However, for these 15 drugs, response rates were high — often higher than 50% — in fatal cancers with no effective therapies. In addition, many of these cancers were rare, and so performing a randomized trial would take years. Patients with fatal cancers cannot wait years. Or, the randomized trials might not be doable since the rarity of the condition would make them exorbitantly expensive and patients and physicians might question the ethics of the control arm.
Q: What should clinicians do with this information?
A: NCCN guidelines are widely used, and our data support a high degree of trust in these guidelines. Panels of physicians who dedicate their lives to specific diseases, and hence are experts in the field, compile the guidelines. The guidelines are well-respected, and our data suggest that this respect is well-deserved because the data are robust.
Again, a lot of recommendations were later FDA approved or confirmed by randomized controlled trials. Where “the rubber hits the road” is in a small proportion of the guidelines where one has high responses in rare diseases.
An example of this involves hairy cell leukemia — a rare condition. The hallmark of the disease is the BRAF mutation, which is found in 100% of patients. The guidelines prescribe the BRAF inhibitor vemurafenib (Zelboraf, Genentech) for this condition. However, this guideline was one of those criticized. When looking at the data, this guideline was based on two multicenter studies with 50 patients between them; of the 50 patients, only one patient did not respond, so the overall response rate was 98%. The responses make sense, because we know that the BRAF mutation is the fundamental biological characteristic of the disease. The question for us as clinicians is: Do we need a randomized trial if the response rate is almost 100%? What if the response rate is 50% in a disease that has no treatment and is fatal — should that therapy be available to patients? The experts at NCCN felt, and we agree, that it is important to make those drugs available to patients because they often have no adequate therapy. When a drug has low toxicity and it is unlikely that anybody is ever going to conduct a randomized trial, the agents should be available to those patients.
Q: What is next ?
A: NCCN will continue to monitor its recommendations to make sure there is strong evidence for all the guidelines.
The other issue that we need to debate is about the randomized trial as the gold and only standard for approving therapies. This standard was developed decades ago, and it is a great standard when we have two therapies that are indistinguishable. However, when we have a therapy with high response rates in a lethal disease, do we really need a randomized controlled trial? Moreover, is a randomized controlled trial even ethical? Decades ago, we needed randomized controlled trials because we had drugs with marginal efficacy and considerable toxicity, and we did not want to expose patients to toxic drugs. However, we now have drugs that have high response rates in lethal diseases. We need to rethink the standards from 40 years ago for the new era, where we really understand cancer and have highly effective drugs that are well-tolerated.
Q: Is there anything else that you would like to mention?
A: One of the general arguments that Wagner and colleagues make is that if we do not do a randomized controlled trial, we are at risk for significant errors. They discuss medical reversals, with one of the best examples being transplants for breast cancer. We did autologous transplants for breast cancer for years. Then, a randomized trial found that these harsh transplants were not efficacious for breast cancer. So, Wagner and colleagues argue that we need to stop these medical reversals, and the only way to stop them is by having randomized controlled trials. I understand this argument. But, it does not give the whole picture.
First, it does not recognize that randomized controlled trials are not always correct. The more important argument is that there is no perfect system. This is not a question of whether we will ever make a mistake by allowing drugs to be approved or placed in guidelines without a randomized controlled trial. Undoubtedly, a mistake could be made. The real questions are, what is the rate of mistakes and how many people are harmed by the mistakes that result from not having a randomized controlled trial vs. the benefit from giving access to patients who have lethal disease to these drugs earlier?
What is the harm done by withholding the drugs?
My argument would be that the number of patients who would be harmed by not having access to these drugs because we are waiting years for a randomized controlled trial — which may not be doable in rare diseases — is far greater than the risk for medical reversals in cancer. When drugs have high response rates in lethal cancers, greater harm is done by withholding those drugs while people with no options die waiting for a randomized controlled trial that may never happen. – by Jennifer Southall
References:
Kurzrock R, et al. Ann Oncol. 2019; doi:10.1093/annonc/mdz232.
Wagner J, et al. BMJ. 2018;doi:10.1136/bmj.k668.
For more information:
Razelle Kurzrock, MD, can be reached at the University of California San Diego, Moores Cancer Center, 3855 Health Sciences Drive, La Jolla, CA 92037; email: rkurzrock@ucsd.edu.
Vinay K. Prasad, MD, MPH, can be reached at Oregon Health and Science University, 3181 S.W. Sam Jackson Park Road, Portland, OR 97239; email: prasad@ohsu.edu.
Disclosures: Kurzrock reports research funding from Foundation Medicine, Genentech, Grifols, Guardant Health, Incyte, Konica Minolta, Merck Serono, Pfizer and Sequenom; consultant fees from Actuate Therapeutics, Genentech, Loxo, NeoMed, Roche and X-Biotech; speakers fees from Roche; and equity interest in Curematch Inc. and IDbyDNA. Prasad reports no relevant financial disclosures.