Loading ...

Last June, Patricia Hart — now 59 years — heard words no one wants to hear: triple-negative invasive ductal carcinoma.

Her care team was frank about the danger: This form of breast cancer is aggressive, and the tumor, they warned, was growing rapidly. Hart agreed to move quickly onto a regimen drawn from the Keynote 522 trial that combines standard chemotherapy with the checkpoint inhibitor pembrolizumab (Keytruda) before undergoing surgery.

For the first few months, the plan seemed to be going well. Apart from fatigue and mild weakness, she tolerated the drugs well enough to keep going to the gym.

But weeks before her final chemotherapy and pembrolizumab treatment, things took a sharp turn.

“I started feeling very fatigued, weak, and I couldn’t eat,” Hart explained.

An emergency room visit revealed Hart was in adrenal crisis. If not treated quickly, she could go into shock, develop organ damage, even die.

Physicians traced the core problem to secondary adrenal insufficiency, which had caused a severe cortisol deficit. Hart would now require lifelong steroid replacement therapy and face the constant threat of another crisis.

While her breast cancer remains in remission, the experience left Hart questioning the cancer treatment.

“If they said that this could cause a lifetime injury to your pituitary and adrenal glands, I would have said, ‘Let’s wait and see. Let’s see how the chemo does, and if I need more, I’ll do more,’” Hart said.

Stories like Hart’s highlight an uncomfortable reality: The benefit/risk equation reported in the controlled setting of an oncology clinical trial may appear favorable but the reality in everyday practice can look quite different.

Trial protocols often enroll healthier patients with few additional illnesses — people more likely to tolerate high drug doses and report fewer complications. Trials may exaggerate the benefits of a regimen and downplay the toxicities by characterizing adverse events with softened language such as “manageable” or “tolerable’ or not adequately discussing them. The Keynote 522 paper, for instance, only noted the risk for adrenal insufficiency with pembrolizumab — 2.6% overall and 1% more severe cases — toward the bottom of a table, and a Merck spokesperson pointed us to page 14 in a 173-page document on prescribing information for the drug.

But once treatments reach a broader population, doctors may find that patients experience more serious hazards than anticipated and that even “tolerable” side effects are not so tolerable.

Immunotherapies, in particular, straddle a fine line. By training the immune system to attack tumor cells, these drugs can also unleash it against healthy organs — thyroid, colon, lungs, and, for Hart, the pituitary.

Kathy Miller, MD, a breast oncologist, believes oncology needs more sophisticated tools to predict both benefit and harm before therapy begins. “It would also be really helpful if we had a better way of knowing who’s at greater risk of long-term, life-threatening, and permanent toxicities,” noted Miller, of Indiana University Melvin and Bren Simon Comprehensive Cancer Center, Indianapolis.

Until those tools arrive, patients and physicians must navigate the gray zone between urgency and uncertainty.

How Clinical Trials Downplay Toxicities

Physicians often find that the toxicities they observe at the bedside exceed what’s reported in clinical trials. That gap makes it hard for oncologists to give patients a clear picture of how they might tolerate a new therapy.

“If you just look at the trial and tell patients that, ‘oh, don’t worry. You know, based on the trial, the side effects are pretty easy. It’s very well tolerated.’ And then after a couple of cycles, the patient ends up in the ICU, then it feels like the patient did not get correct information,” said Bishal Gyawali, MD, PhD, a medical oncologist at Queen’s University in Kingston, Ontario, Canada.

One reason for the disconnect is the mild language used in many trial reports. In a 2018 BMJ analysis of 122 phase 2 and phase 3 oncology studies, Gyawali and colleagues looked at how these trials reported toxicities and found that about half of the papers relied on phrasing that did not match the actual toxicity profile.

“They were using what we call ‘downplaying terms,’” Gyawali said. Even in the presence of severe, serious, or fatal adverse events, authors would write that “toxicities were acceptable,” “the treatment was safe,” or the drug “has a manageable and mostly reversible safety profile.”

For example, a trial of liposomal irinotecan for pancreatic cancer states that it “has a manageable and mostly reversible safety profile,” yet many required dose reductions and five people receiving the treatment died from drug toxicities.

“If a patient is dying from toxicities, by definition, it’s not manageable,” Gyawali said.

Industry sponsorship amplifies the pressure to soft-pedal harms. Roughly 90% of oncology trials are funded by drug companies, said Gyawali. “When billions of dollars are at stake, then there is incentive for them to massage the messaging in a way that makes it more palatable and more acceptable to the oncology community,” he explained.

Miller routinely sees that tendency in manuscripts. “Industry really doesn’t like people to focus on the toxicity of their drug, but they don’t even like you to call it toxicity. They would prefer you call it ‘treatment emergent adverse events’ because toxicity sounds bad,” explained Miller.

The clinical significance of adverse events can also be hard to gauge. Trials grade toxicities from 1 (mild) to 5 (death), yet they rarely specify how long those events last.

Miller points to grade 2 diarrhea — five to seven extra bowel movements a day — as an example. “Grade 2 diarrhea, 1 or 2 days out of the month is probably not such a big deal,” she noted, but “grade 2 diarrhea, most of the time, that’s a very different experience.” Chronic grade 2 diarrhea means your whole life revolves around where the toilet is, making it difficult to travel or even run errands.

Statistical sleights of hand can further blur the picture. Miller has seen papers that reported all grade 2 or higher adverse events together as a single percentage, which can make toxicities in treatment and control arms look more similar than they are. “If you’re lumping anything grade 2 or greater together, and one arm has predominantly grade 2 toxicity and another arm has a greater portion of grade 3 and 4, then the severity has shifted,” she explained.

Finally, life-threatening reactions may hide under the heading of treatment discontinuations. Side effects are sometimes cited simply as a reason for stopping therapy without specifying which toxicities occurred. Other times, withdrawals are labeled “patient preference” or “physician discretion.” 

“Those should make you wonder,” Miller cautioned. “There can certainly be patient specific reasons that someone needs to discontinue, but I am positive that a decent number of those are because of the toxicities.”

But even when trial authors describe toxicities accurately, there are still reasons why side effects in a general population might not be accurately captured in trials.

Cherry Picking Clinical Trial Participants

When drug makers or academic teams design a clinical study, they start with an extensive list of exclusions — medical conditions, lab values, or prior treatments that automatically disqualify would-be volunteers. The intent makes sense: Protect participants and strip away confounding variables so researchers can see if a treatment is effective.

“When you test out a new drug, you may not even know if it works, and you want to test it on the patient who’s most likely to see a response,” explained Gary Lyman, MD, codirector of the Hutchinson Institute for Cancer Outcomes Research in Seattle.

The strict eligibility criteria, however, produces a study population destined to do better than nontrial patients, one that bears little resemblance to the people doctors treat every day.

The consequence is trials “do not always reveal all of the side effects that might be experienced in the real-world setting,” Lyman added.

When treating liver cancer in community clinics, for instance, patients may have cirrhosis, diabetes, or other liver-related issues. In other words, “you expect that the liver has been damaged,” Gyawali explained. But in trials, eligibility criteria are often limited to patients with pristine liver function, which “means these are not our normal patients,” he said.

In Gyawali’s experience, researchers might report a 30% rate of severe toxicities in a study environment, while in the clinic, oncologists may see that rate swell by twofold. That’s why he treats the published toxicity rate as the minimum those patients will experience.

Support systems inside a trial widen the gap even further. Enrolled patients see physicians, nurses, and research coordinators far more often than standard care allows, catching complications early and smoothing adherence. “There’s plenty of evidence that patients who enroll in clinical trials do better just by virtue of being a trial participant compared to patients who don’t get enrolled in a trial,” said Gyawali.

The result is a paradox: The very safeguards that keep early testing ethical also make it harder for clinicians and patients to predict how a therapy will behave once it leaves the controlled environment of a study and enters the messy reality of routine practice.

More Is Not Always Better

Cancer therapy once followed a stark principle: Deliver as much drug as the patient could physically withstand in order to wipe out every last malignant cell. Not surprisingly, that “maximum tolerated dose” strategy brought a barrage of toxicities.

“The premise was,” Lyman explained, “that for traditional cytotoxic therapies, the effect on the cancer was very much dose dependent. The higher the dose, the more tumor cells killed.” 

With immunotherapies and other targeted agents, however, the classic dose-response curve often flattens. Piling on more drug no longer guarantees better tumor control, and it often means more harm.

“Drugs are approved at excessive dosages because there’s been a fallacious belief that more is better. For some drugs, more is better, but that requires proof,” said Mark Ratain, MD, a hematologist and oncologist at University of Chicago Medicine and founder of the Optimal Cancer Care Alliance. “There doesn’t need to be a single dose that is best for all patients.”

Ratain points to Amgen’s sotorasib, a targeted therapy for non-small cell lung cancers harboring the KRAS G12C mutation. The drug won accelerated approval at 960 mg daily, but based on Amgen’s data, the FDA also required a study comparing a 240 mg dose to the 960 mg dose — a fourfold dose reduction.

The results of the comparison trial, published in 2024, showed the higher dose extended overall survival by roughly 1 month, on average, but led to 89% more treatment-related grade 3 or higher side effects. Gastrointestinal problems dominated with 42% of patients developing any diarrhea, Ratain noted.

Because of a gap in the FDA’s authority, it may be too late for the agency to adjust the approved dosing and Amgen has an interest in protecting its revenue, Ratain explained.

Amgen defended the drug dosing, telling Medscape Medical News that “the current data suggests a different perspective.” Based on recent findings, including the 2024 trial comparing the two doses, “we are confident that patients with lung cancer are benefitting from the FDA-approved 960 mg dose,” said Annik Allen, associate director of global media relations at Amgen. Allen highlighted the “numerically higher” overall response rates and overall survival at the higher dose and noted “no meaningful difference” in adverse events leading to dose interruptions (45.2% vs 39.4%) or discontinuation (16.3% vs 12.5%) between the two doses.

Another facet to the ‘more is better’ mantra is the idea that combining more drugs together will improve patient outcomes. But that’s not always the case.

In Hart’s scenario, she received standard chemotherapy plus pembrolizumab to treat her breast cancer, but do patients in her situation really need both drugs? 

According to Miller, probably not.

“The number of patients who are cured because I added an immuno-oncology agent is quite small,” Miller said. “Many breast cancer patients would be cured if I just gave them chemotherapy.” 

Leaving out pembrolizumab would have also saved her from a life-changing toxicity.

The antiquated “more is better” default underscores a wider reckoning in oncology: Treatment should be guided by evidence of optimal benefit — especially when some newer agents will ramp up toxicities without much benefit to survival.

Can the Situation Improve?

Ultimately, the FDA decides which therapies reach the market and sets the rules for how data are gathered and interpreted. Many oncologists argue the agency could do far more to curb avoidable harms.

“We’re not doing the right trials, and the FDA is not asking the right questions,” said Ratain. “I see this as a regulatory problem.”

One major gap lies in the earliest stages of drug development, where the so-called “maximum tolerated dose” still dominates trial design. A modern approach, experts say, would build phase 1 and 2 programs around randomized dose-optimization studies, intended to determine when higher doses add toxicity without improving efficacy.

Broader eligibility criteria would also make safety data more realistic. “We need to relax eligibility criteria so that participants in the trial are more reflective of real-world patients,” Gyawali said. Excluding people with mild organ impairment or common comorbidities may keep early studies tidy, but it hides how a drug behaves in the clinic.

Within ongoing trials, researchers could be extracting more information about tolerability from patients. Real-time patient-reported outcome measures of fatigue, pain, and daily functioning could, for instance, flag emerging toxicities sooner than clinician-graded checklists.

Once trials wrap up, transparency in publishing becomes critical. Gyawali, a founding member of Common Sense Oncology, helped draft a checklist urging journals and sponsors to present adverse events by grade, duration, and reason for treatment discontinuation rather than lumping them into buckets.

Most oncologists also want safety monitoring to continue well beyond FDA approval. “I think that it’s really important that we don’t stop examining the benefit and harm of treatments once they get approved based on randomized trials, that we continue to capture data,” said Lyman. Too often, he added, clinicians decide how severe a toxicity is, when “it’s really the patient’s experience which should be the gold standard.”

Deeper analyses of post-approval data are already uncovering disparities. “We just recently identified and now confirmed that Black women are much more likely to have severe peripheral neuropathy with weekly paclitaxel compared to Caucasian women,” said Miller. Switching these patients to docetaxel can spare them months of pain and numbness.

Scientists, however, still lack reliable markers to flag patients at highest risk for side effects from cancer therapies. One way to track risk levels would be to require companies to bank biospecimens and share de-identified datasets so independent teams can look for predictive signals. Another would be to require greater data transparency to allow outside investigators to re-analyze raw safety data and publish independent dose-optimization research.

“If anyone can make a change, it would be the regulators,” Gyawali said, whether by rewriting eligibility templates, mandating patient-reported outcome collection, or tying approval to open data.

Hart lives with the consequences of today’s system. The irreversible adrenal damage from pembrolizumab “is a lifetime worry,” she said, one that might have been avoided.

She also hopes her story will remind clinicians to slow down long enough to discuss not only the promise of new therapies but also the possible life-altering dangers that could accompany them.