Study methods make the drug pipeline, and the methods are lacking

Much epidemiological work on cancer focuses on identifying risk factors for various types of cancer, which kill more than 600,000 Americans every year and render seriously ill millions more. Prevention is, after all, the best medicine.


But with such a wide-ranging and serious disease, there will always be many who fall sick. Cancer clinical trial design could perhaps benefit from more epidemiological support, according to Vinay Prasad, MD, MPH, whose recent research evaluates the success or failure of efforts by pharmaceutical companies and the Food and Drug Administration (FDA) to speed drug approvals and improve patient outcomes.


Prasad finds that drug approvals have indeed been sped, if one considers the number of drugs approved per year, but many of the new drugs have not substantively improved patient outcomes.


The FDA, under pressure from industry and patient advocacy groups, has allowed trials to measure “surrogate endpoints.” They measure the drugs’ effects on things like tumor size that would seem like they should correlate with overall survival, but often don’t. Proposed new treatment regimens also sometimes rely on so-called “real-world” data from electronic health records (EHRs) to compare a treated group to an imagined control group in order to lower the cost of drug development and expand access to experimental drugs.


The methods may sound reasonable, but they are often problematic. (Indeed, many include what Prasad says any trained epidemiologist would recognize as “rookie mistakes.”) Two-thirds of cancer drugs approved from 1992 to 2019 used surrogate endpoints in their clinical trials; in one third, the surrogate endpoint had no previous data to support its use. Just a third of them have since been shown to improve survival or quality of life – the measures that matter. The rest could be no more effective than chicken soup – and they’re an awful lot more expensive.


“If you’re going to have a regulator, it should make sure that the drug works,” Prasad said. “Surrogate endpoints have a role in rare cases, but right now, the exceptions are the rule.”


Part of the problem is that the FDA often stipulates that a drug should receive more study once on the market, but it rarely enforces those demands. The idea is that if a drug appears safe and has effectiveness against a surrogate endpoint, it probably works and should be given to sick patients STAT. Again, this sounds good, but facts don’t fully bear it out. Surrogate-endpoint trials speed the drug approval process by just 11 months over an average of 7 and a half years of development, according to Prasad’s research, which comes out to 12 percent.


So-called real-world data may actually lead to more errors, according to a commentary Prasad and UCSF hematology-oncology fellow Rahul Banerjee, MD, just published in JAMA Network. When real-world studies are compared to randomized clinical trials, the findings align less than half of the time. The real-world studies are more likely to lead to errors that favor recommending aggressive interventions that don’t work. (The patients whom doctors opted to treat in this way without randomization were likely healthier.)


“For people who think real-world data is going to be magic, it looks like it’s going to be very messy,” Prasad concludes.


The criticism often lodged against Prasad’s line of argument is that serious reforms would slow drug development down to a trickle, depriving patients of life-saving interventions. But the trickle might instead, Prasad suggests, accurately reflect the flow of truly transformational drugs, absent those that extend life by 10 days or even just improve a CT scan.


Indeed, the market seems to have found the sweet spot for pharmaceutical profits. Near-total deregulation like that championed by Balaji Srinivasan – once considered as a nominee for FDA commissioner – would lower the cost of drugs by removing the FDA’s seal of approval. A stronger FDA like the one Prasad wants to see would limit the number of drugs companies could market to those that have more meaningful benefits.


“If you held as a standard proving the drugs actually do what they say they do, the industry would retool entire R&D pipeline. You’d see fewer trials, but you would have better trials. We wouldn’t just have options, we’d have good options,” Prasad said.