Cancer Clinical Trials Are Excluding the Patients They Are Meant to Help

In her five years as a clinician treating blood cancers, Elizabeth Goodall noticed a critical failure point: Many patients who wanted to enroll in clinical trials or try a new drug were being told they couldn’t.

“I listened to patients not only describe their symptoms, but also heard their confusion and disappointment if they were not eligible for a clinical trial or new therapy that they were desperate to have access to, for often quite arbitrary reasons,” said Goodall, who is now a PhD candidate at the Oliver Newton-John Cancer Research Institute.

To further dig into this issue, Goodall and her colleagues systematically evaluated ongoing clinical trials for diffuse large B-cell lymphoma (DLBCL) and applied the trials’ eligibility criteria to real-world patients. This research showed that based on eligibility criteria, not a single person in a real-world cohort of 180 relapsed DLBCL patients would qualify for all seven recent DLBCL clinical trials that led to landmark drug approvals.¹ With between 31 and 45 different criteria, each of the trials excluded the majority of real-world patients—a finding that shocked Goodall and her colleagues. “More than half of [the patients] were not eligible for any of the seven trials,” Goodall said. “I didn't expect it to be so stark.”

The reason behind this chasm is restrictive trial requirements. For example, measures of organ function and disease burden automatically eliminate many patients from the trial population. Prior lines of treatment also disqualify many patients, even though those patients are typically the most willing to try experimental new therapies. “We know that clinical trial criteria are getting more restrictive over time, and that's been proven in the literature,” Goodall explained. “That [doesn’t] help our patients because we know that we have an aging population with an increasing amount of chronic disease.”

For DLBCL patients, these results help explain why so many new treatments elicit such lackluster responses; given the stringent participation requirements, patients who are treated with experimental new drugs in clinical trials simply do not serve as a good proxy for those who are given the same drugs once they are approved.¹ It means that despite the recent drug approvals and the glut of clinical trials for DLBCL—over 1500 have been officially registered for the condition since 2019—many patients’ hopes will be dashed.

The team’s findings in DLBCL echo what researchers studying other types of cancer—and even other diseases—have been saying for years: Clinical trials for experimental new drugs are getting more complex, with strict and extensive inclusion-exclusion criteria resulting in increasingly homogeneous experimental groups that fail to reflect the reality of broader patient populations.

So, as medicine advances and clinical trials become more restrictive, how can scientists and clinicians ensure that real-world patients aren’t paying the price?

Maximizing the Internal Validity of Trials and Speed to Approval for New Drugs

Unlike the immortalized cell lines and inbred laboratory animals used in preclinical research—essentially genetic clones—human patients are much more complex; they have different genetic ancestry, cancer subtypes, microbial environments, and levels of baseline health and resilience. They may have been treated at different medical centers, had a range of prior lines of treatment, and be differently debilitated by their disease. All these factors muddy the waters when it comes to understanding the real effects of experimental treatments.

“We are talking about human beings who are not only affected by one condition, and the outcomes might be blurred by other conditions existing,” said clinical oncologist Luigi de Petris of the Karolinska Institute. “In some [cases] you have a very high symptom burden related to the cancer, so even if you have a patient with different other comorbidities [such as heart or lung disease], you know that at that moment the patient is mainly affected by the cancer in question. But other diseases are not the same.”

There are valid reasons why sponsors of clinical trials want to test new drugs on as homogeneous a patient population as possible, said Jorge Nieva, a clinical oncologist at the University of Southern California. “The ideal research study would be conducted in a group of 100 identical twins who all have been raised on identical diets and all have the identical cancer, which was implanted in them on the same day,” Nieva said. Unfortunately, he added, “Nothing like that actually happens in the real world.”

A clinical trial of a new drug generally aims to demonstrate that it is safe and effective, thereby establishing a risk-benefit ratio that regulators such as the FDA carefully weigh. “A clinical trial, especially when you are exploring a new drug or new combinations, is to show whether the new one is better than [the current standard-of-care], and to do this you need to have a clear statistical setup,” said de Petris, who has conducted research about the gap between trial populations and real-world results in small-cell lung cancer.²

Trial sponsors, therefore, need to minimize bias and confounding variables as much as possible, said Nieva; patients must be functionally strong, have good end-organ function, and show biomarker values that fall within very specific ranges. “All this is done to increase the internal validity of the trial and increase the likelihood that we'll be able to see a signal from the effect of an intervention or a new drug,” Nieva explained. “Unfortunately, over time and over the course of my career, this balance has shifted towards more internal validity and less external validity, as economic pressure drives pharmaceutical companies to want to have the most homogeneous patient population possible for conducting their research.”

Nieva’s lab has performed multiple analyses of pharmaceutical, academic, and government-sponsored trial studies to explore the gap between trials and real-world populations. “The most restrictive clinical trials are the ones that are pharma sponsored,” he added. According to Goodall, based on restrictive trial criteria, less than five percent of patients with any specific disease actually make the cut and are included in clinical trials for that indication. However, once a drug is approved based on those restrictive trial groups, it can be used to treat the other 95 percent of patients with the same disease, even if they have significant comorbidities, less-than-perfect organ function, or biomarkers outside the required range.

Kerstin Vokinger, a medical doctor and lawyer at the University of Zurich and ETH Zurich who specializes in the regulation of medicines and technology, said that in oncology in particular, drugs can be subject to rapid approval based on safety and efficacy measures from the most restrictive early-phase trials, but these data cannot always be extrapolated to broader patient populations. “We can see [in the research] that there is a strong focus on speed to approval, which is very important,” said Vokinger. “But if we speed and accelerate the process and we have less evidence, I believe it's crucial at the very least to conduct further studies also after approval, where we really try to also include, as much as possible, the real-world setting.”

Arbitrary Inclusion Criteria, Administration Burden, and a Lack of Diversity Plague Cancer Trials

As trials become increasingly complex, Nieva said, so too does the administrative burden. “When I started in oncology 30 years ago, a clinical trial study protocol would typically be 30 to 50 pages long,” said an Nieva. “Now, a typical study protocol is 300 pages long and includes additional handbooks and guides, bringing the total content of material, oftentimes [to] nearly 1,000 pages of documentation and processes.”

Goodall also argued that many of the eligibility criteria are not always meaningful. “It seems incredibly arbitrary in a lot of cases where renal function limits are made and cardiac impairment limits are made,” she remarked. Even if patients could meet the criteria, proving it is another question. “The level of evidence that you need for that patient to even come and be able to sign onto that trial is becoming ridiculous,” Goodall added.

Vokinger’s research has shown that there are other mismatches between trial populations and the real world, including sex and age differences.³ She also emphasized that there are other factors to consider. “It's not just about how sick or not sick [the patients are], or comorbidities, or fitness, or age, or gender, but it's also about low- and middle-income countries that need to be included, it's about race and ethnicity,” she said.

In a recent study, researchers showed that despite being a major contributor to the global burden of disease, African countries are vastly underrepresented in clinical trials.⁴ “As a result of that, you then rely on enrollment of US patients of African ancestry to try to understand diversity within that group, but the enrollment numbers are really quite small for that group,” Nieva remarked. However, this is not because they do not want to participate in research, Nieva said, but because academic medical centers where trials take place are typically located in wealthier areas. “Community health centers, individual doctors’ offices, and clinics that serve urban populations don’t have the ability to conduct clinical trials,” he added.

‘No Carrot or Stick’: The Struggle to Broaden Trial Criteria

How can clinical trials strike the right balance between maximizing internal validity and better reflecting real-world patients? Some time ago, Nieva had an idea that involved enrolling patients in trials based on fewer selection criteria so that there would be both a broader population to assess safety measures and a narrower group in which to analyze drug efficacy. The patients could be stratified by their level of fitness and lab biomarkers. “You could envision a system where we were able to capture large groups of patients and be able to get a sense of efficacy in the real world, while at the same time not penalizing the pharma company who developed the trial, by allowing the analytical population to be a much narrower group to look for effect size,” he explained.

However, Nieva continued, any change to the current status quo must begin with government systems and regulators. “Everything that we see in the system that we have is the system our regulators have built, and the response of pharma has always been appropriate for the regulatory environment that it lives in,” he said. Recent calls to liberalize inclusion-exclusion criteria have been met with mixed responses. “Right now, there really is no carrot or stick provided by regulators to make those recommendations anything more than simply recommendations,” Nieva added.

Pragmatically, Goodall suggested that trial sponsors should focus on more evidence-based exclusion criteria, rather than arbitrary values. “Think about it a little bit more and include patients that are safe to get your drug but might be a more general representation of the population,” she said. Increasing crosstalk between stakeholders could also increase positive outcomes. “All the different stakeholders have different skills, understanding, and perspectives on this issue,” said Vokinger. “It's an issue that we have to sort of solve together, so building these bridges [between stakeholders] would be really beneficial for patients.”

Vokinger hopes that by actively sharing the results of these studies with government bodies, the pharma industry, and the media, she and other researchers can contribute to positive change for patients—a sentiment echoed by Nieva. “The best thing that the scientists can do is put out publications like [Goodall’s] in the British Journal of Hematology that recognize the problem,” Nieva said. “Then it's really up to government officials to decide if they want to act on this information or not.”

According to Goodall, it’s not all doom and gloom; some real-world studies show that the effects of new drugs in real-world populations are as good as, or even better, than seen in trial data, largely because adverse events such as toxicity are not only expected but better recognized and treated. “We should all, as an entire group, work towards recognizing [the gap between trial populations and real-world] and just have a bit more thought about why some of these patients are excluded, particularly as our population ages and gets more complex,” she said. “It's not an absolute disaster.”