In releasing a draft guidance earlier this year on dose optimization for cancer therapies, the US Food and Drug Administration is steadily advancing its decades-long campaign to move the industry away from the default use of a maximum-tolerated dosing paradigm in clinical trials and toward a more patient-centered approach.
On the front lines of that campaign are cancer patients who often endure unnecessary suffering due to excess doses of targeted therapies, which, in the era of molecularly informed precision oncology, they tend to receive and be exposed to for months or years. When Chris O’Sullivan of New York experienced her second recurrence of HER2-negative, estrogen receptor-positive metastatic breast cancer, she had already suffered devastating side effects from her previous therapies. The hormone drug letrozole caused pain and stiffness, and Amgen’s Xgeva (denosumab), meant to reduce bone fractures, limited blood flow to her jaw, causing her to have part of it and her sinus surgically removed. Mindful of those past experiences, she asked her oncologist at NewYork-Presbyterian Columbia University Irving Medical Center if she could start her new course of treatment, Eli Lilly’s CDK4/6 inhibitor Verzenio (abemaciclib), at a low dose and increase it if she was tolerating it well.
However, her oncologist advised that she start Verzenio at the recommended dose, 150 mg, which she took while on AstraZeneca’s hormone therapy Arimidex (anastrozole). But when O’Sullivan started experiencing diarrhea, stomach cramps, and nausea, her oncologist had to lower her Verzenio dose to 100 mg and then again to 50 mg. “I know somebody who’s 180 pounds and I’m 110 [pounds], and we were on the same [Verzenio] dose,” she said. “That makes no sense to me.”
Patients push for change
O’Sullivan’s experience isn’t unique, according to Anne Loeser, a metastatic breast cancer patient and founder of the Patient-Centered Dosing Initiative. In 2020, the organization surveyed around 1,200 metastatic breast cancer patients about the prevalence and severity of their treatment-related side effects, and an overwhelming majority, 86 percent, said they experienced at least one “bad” episode, while one in five respondents had to visit the emergency room or hospital because of a side effect. More than two out of five (43 percent) had to skip at least one treatment due to side effects.
Loeser’s organization presented those results at the 2021 American Society of Clinical Oncology’s annual meeting, where it caught the attention of the US Food and Drug Administration, which launched Project Optimus in early 2021. Within this initiative, the agency’s Oncology Center of Excellence works with stakeholders across industry to optimize drug dosages for efficacy, safety, and tolerability before approval.
When Loeser was starting the Patient-Centered Dosing Initiative and setting goals, she and her group wondered whether they could ask the FDA to guide sponsors toward a more patient-centric approach than the standard maximum-tolerated dosing (MTD) paradigm in cancer drug development. Loeser didn’t think it was going to fly. “Who are we? We’re just a bunch of patients,” she recalled thinking. But the FDA seemed very receptive to the survey data, and from there, Loeser said momentum built behind the movement to replace the default MTD paradigm with dose optimization.
Shortly after Loeser’s ASCO presentation, Richard Pazdur, director of the FDA’s Oncology Center of Excellence, and coauthors Mirat Shah, Atiqur Rahman, and Marc Theoret penned an opinion piece in the New England Journal of Medicine calling for sponsors to more carefully evaluate drug response, efficacy, and safety from early therapeutic trials for the purpose of selecting the best dose, rather than defaulting to the MTD.
They recognized in the editorial that under the pressure of expedited timelines in oncology drug development, sponsors tended to select higher doses in registration trials to maximize their odds of proving efficacy. However, they noted that initial clinical trials of two PD-1 inhibitors, Merck’s Keytruda (pembrolizumab) and Bristol Myers Squibb’s Opdivo (nivolumab), included an analysis of wide dose ranges, leading sponsors to choose doses lower than the MTD while still preserving efficacy.
“Sponsors should carefully evaluate exposure-response, efficacy, and safety data from early trials to inform dose selection, rather than automatically selecting the maximum-tolerated dose,” Pazdur and colleagues concluded.
The MTD approach was developed on chemotherapies, based on the principle that the therapeutic effect increased linearly in relationship to the dose. The most effective dose, then, would be the largest dose the patient could possibly tolerate, even to the point of enduring significant toxicity. In an MTD-based trial, the drug dose is increased gradually until dose-limiting toxicities are reached, and a dose close to that limit, or just below it, becomes the default dose for subsequent clinical trials.
The traditional design for determining MTD in a Phase I trial is known as “3+3.” That means that three patients are enrolled at an initial dose level, and, if no dose-limiting toxicities are observed, investigators enroll three more patients at the next dose level. If a dose-limiting toxicity is encountered, three more patients are enrolled at the same dose level. If dose-limiting toxicities are observed in two or more out of six patients at a specific dose level, the MTD is considered to have been exceeded and further dose escalation is not pursued. Six patients are then enrolled at the next dose down and so forth, until there is a cohort in which at most one patient out of six experiences a dose-limiting toxicity.
In 1994, the FDA implemented guidelines issued by the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) and advised sponsors to collect dose-response data in clinical trials to identify “an appropriate starting dose” and a dose beyond which further benefit is unlikely or side effects are unacceptable. However, until recently, there’s been little movement in industry to identify “an appropriate dose” in cancer drug trials, and the focus has remained on establishing the MTD.
“The oncology regulators at FDA never followed through on that,” said Mark Ratain, director of the Center for Personalized Therapeutics at the University of Chicago Medicine. “They basically gave oncology a pass because there was concern that we needed these drugs quickly, and therefore, optimizing the dose wasn’t a priority, and [there was] a mistaken belief that more [drug] was better.”
The FDA is now working hard to change that belief. In its January draft guidance, “Optimizing the Dosage of Human Prescription Drugs and Biological Products for the Treatment of Oncologic Diseases,” the agency offers some specific recommendations for early trial design. A key piece of advice is to compare multiple doses based on preliminary dose-response data and to use a randomized, parallel dose-response trial design.
“Project Optimus can be thought of as a response to the changing oncology drug development landscape,” Brian Heiss, a clinical reviewer in oncology at the FDA’s drug division and a liaison to Project Optimus, said over email. “Most oncology drugs in development right now are targeted therapies, for example, kinase inhibitors, antibody-drug conjugates, and monoclonal and bispecific antibodies.” Those newer drugs differ from cytotoxic chemotherapy, Heiss pointed out, and they require a different approach for dose evaluation.
“Unfortunately, most oncology drug development is stuck in the ‘more is better’ paradigm of cytotoxic drugs, where the maximum-tolerated dose is sought to achieve what is believed to be the greatest anti-tumor effect,” Heiss said.
He noted that an FDA-approved drug has demonstrated a favorable benefit-risk profile in its intended patient population, but that doesn’t always mean that the dosage has been thoroughly optimized. Doses below the MTD may not be well characterized, and there could be better options.
There are additional considerations in the case of biomarker-driven therapies. “Many targeted therapies are different from cytotoxic chemotherapy in that the duration of therapy is much longer, and long-term tolerability is incredibly important,” said Heiss. “MTD-centered drug development is inadequate to assess these longer-term toxicities as it focuses only on short-term severe toxicity assessments, generally only in a limited number of patients.”
As in O’Sullivan’s experience with the CDK4/6 inhibitor Verzenio, Heiss said patients on such therapies may require frequent dose modifications, limiting their overall exposure to an otherwise efficacious drug. This, in turn, can reduce the benefit the patient sees from the drug. That can affect patients on an individual level, but on a larger scale, dose reductions can obscure the benefit of a drug in a clinical trial. Low-grade symptomatic toxicities can also impact patients’ quality of life, and in some cases, their treatment compliance. Optimizing the dose before the drug gets to market can reduce or avoid those issues, Heiss said.
While Heiss wasn’t able to give details about how many molecularly targeted drug applications are receiving guidance under Project Optimus, he said the number of drugmakers seeking such advice is increasing, particularly in the oncology space, where randomized dosage trials have been relatively uncommon compared to other therapeutic areas. “FDA is receiving questions and providing advice on dosage optimization as early as the pre-investigational new drug meeting, before the drug has been given to humans,” he said. “Our goal is that the drug dosage has been optimized for benefit-risk for patients at the time of drug approval.”
Advantages of optimizing the dose prior to registration include avoiding the need for additional trials later and allowing more patients to stay on the drug as long as they are receiving benefit, Heiss added.
Although the agency has recommended randomized dose-finding trials for oncology drugs, “they are just one piece of a reimagined dosage selection paradigm for cancer drugs,” he continued. “In a traditional dosage selection paradigm, patients are evaluated for short periods of time for serious or life-threatening toxicity to determine the MTD, and there is not much consideration of other key data.” Other data that can contribute to identifying a potential range of therapeutic doses includes pharmacokinetics, pharmacodynamics, safety, and preliminary activity.
Case studies in dose optimization
The details of how Merck and Amgen arrived at the marketed doses of their respective cancer drugs, the checkpoint inhibitor Keytruda and the KRAS inhibitor Lumakras (sotorasib), have become case studies in precision oncology for experts tracking what drugmakers are getting right and how they can better align with the spirit of the FDA’s dose-optimization guidelines.
In the case of Keytruda, the dose-escalation strategy Merck used in early clinical trials to identify the recommended Phase II dose of 200 mg every three weeks, later became the approved, marketed dose. Merck has touted this as an example of a dose-finding strategy that bucked the MTD norm. The company used a pharmacokinetic model to estimate the dose required for target saturation, which served as a surrogate for maximum pharmacological effect for antagonist monoclonal antibodies.
Gideon Blumenthal, Merck’s VP of global regulatory affairs and oncology, said that early on in Keytruda’s development, Merck studied a 2 mg/kg and 10 mg/kg dose, and randomized this in some early pivotal Phase III studies to identify 2 mg/kg as the maximally effective lowest dose.
Blumenthal, who was the former deputy director of the FDA’s Oncology Center of Excellence, explained that because the drug has a wide therapeutic window, there was no need to increase the dose higher than 2 mg/kg. He said that dose then “got bridged” to a flat dose of 200 milligrams every four weeks, or 400 milligrams every six weeks. When it comes to novel therapies like antibody-drug conjugates, biomarker-driven therapies, and immune-oncology agents, Blumenthal said Merck is “very committed to using all [available] methodologies to come to the right dose prior to marketing authorization.”
Blumenthal, who actively participated in workgroups hosted by the advocacy organization Friends of Cancer Research in support of Project Optimus, said Merck has had “great interaction and advice” from the FDA across its oncology programs, including early feedback and interactions, even before beginning Phase I trials. “Certainly, at the end of Phase II, we’re engaging the agency to make sure we’re optimizing the dose before we move into our pivotal studies,” he said.
Optimizing the dose is not only important because it benefits patients, but also for commercial reasons, according to Blumenthal. “From a competitive standpoint, we’re seeing now in oncology that the days of having just one player for a given target are over,” he said. For example, there are numerous companies working on KRAS inhibitors. In a crowded market, “if you don’t get the dosing right,” Blumenthal said, “you’re at a competitive disadvantage.”
Still, opinions differ on what it means to get the dosing right. In a letter published in Clinical Pharmacology and Therapeutics in May 2022, Ratain and his coauthors critiqued the dose-optimization strategy Merck used to develop Keytruda. Even though Merck randomized 2 mg/kg and 10 mg/kg doses of Keytruda, Ratain and colleagues pointed out that, “Unlike with nivolumab, there has never been a randomized dose-ranging study that included any dose less than 2 mg/kg every three weeks, and there has never been any demonstration of a dose-response relationship for any PD-1 inhibitor at doses administered to patients.” They concluded based on this rationale that it is not possible that the marketed Keytruda dose of 200 mg every three weeks — bridged from the 2 mg/kg dose — is optimal, “only that it is the dose that pembrolizumab’s manufacturer has chosen to market — even if that dose is excessive.”
Ratain also pulls no punches when it comes to another oncology dose-optimization case study, Lumakras.
In its original Phase I trial, the drugmaker conducted a dose-escalation study, aiming to elicit a 20 percent to 33 percent incidence of dose-limiting toxicity. The doses used were 180 mg, 360 mg, 720 mg, and 960 mg. Investigators deemed the highest dose of 960 mg to be well tolerated, even though 58 of the 129 patients had serious complications and nine discontinued due to toxicity. In other words, it was a standard MTD trial.
“That drug, in my opinion, was completely misdeveloped,” Ratain said. “Now, they’ve done a Phase III trial that, in our opinion, confirms that that dose is not optimal.”
Ratain is referring to the CodeBreak 200 Phase III trial, comparing the 960 mg Lumakras dose to the chemotherapy docetaxel in previously treated patients with KRAS G12C-mutated non-small cell lung cancer. At the European Society of Medical Oncology’s annual meeting in September, Amgen reported a modest improvement in median progression-free survival among Lumakras-treated patients (5.6 months) compared to those on docetaxel (4.5 months). There was no overall survival advantage. However, about one-third of patients on Lumakras had a grade 3 or higher treatment-related adverse event leading to numerous dose interruptions, reductions, and discontinuations.
“The intolerance to sotorasib at 960 mg reflects inappropriate dosing, particularly regarding diarrhea, which is attributable to the low bioavailability of this poorly absorbed drug,” Ratain and his coauthors wrote in The Cancer Letter in September 2023.
In one of the first known examples of the agency giving teeth to its recommendations on dose optimization, the FDA’s accelerated approval of the KRAS-targeted drug in May 2021 came with a post-marketing requirement for Amgen to carry out a trial comparing how patients fare on the approved 960 mg dose and a 240 mg dose.
“This is a beautifully designed, beautifully targeted drug that was expected to be effective at a very low dose and was effective at a very low dose,” said Ratain, speculating that Amgen could reasonably expect full efficacy for Lumakras at a dose as low as 120 mg. Amgen declined an interview for this story.
One size doesn’t fit all
In oncology, because a cancer diagnosis usually engenders a sense of urgency, sponsors often don’t follow the rules and guidelines that are commonplace in other disease settings, such as depression or hypertension, observed Lesley Seymour, deputy director of the Canadian Cancer Trials Group. “Cancer patients don’t have time to wait around,” Seymour said. “If there’s a new drug that works, you need to get it to them ASAP, otherwise they might succumb to the disease before they can access the new drug.”
That means that many procedures that are now considered standard in other diseases haven’t been implemented in cancer trials in the interest of running faster results. “For the last 10 to 20 years, there have been discussions about how we could make all of this more precise by doing things like randomizing [as early as] Phase I, but certainly in Phase II studies,” Seymour said. “But to be frank, there has been a fair amount of pushback [from patients and physicians] … fearing that any delay in developing a drug would make that drug get to market, or become available, later.”
Of course, there are also commercial advantages to expediting a drug to market. “The quicker you can develop a drug, the more time it’s on patent while you’re still selling it, and the more money you make,” she said.
When drugmakers do take the time to design dose-optimization trials, Seymour said there are different strategies and no one-size-fits-all solution. “We’re talking about a big range of different drugs with different structures, with different mechanisms of action, some of which are given intravenously, some are given orally,” she said. “There isn’t a magic protocol that could be applied to every drug that could be tested.”
Broadly, sponsors would need to test different doses of a drug in the trial while monitoring not just for toxic effects, but also looking at variables such as blood levels of the drug, effects on the tumor, pharmacokinetics and pharmacodynamics, and biomarkers where appropriate. And ideally, at the end of the dose-escalation portion of the clinical trial, three or four doses are well tolerated by patients, Seymour explained. But then, instead of choosing by default the highest of that set of well-tolerated doses, she said the next step is to go into the second part of the trial and randomize patients between those three or four doses to find the lowest dose that shows optimal efficacy.
One dilemma, which shows up in the Merck and Amgen examples, according to Seymour, is choosing the dose range to use and how many doses. Using too small a range, or choosing too few doses, can result in a total miss of the lowest effective dose. At the same time, she noted, “you don’t want to be too granular and have too many dose levels, but you also want to have enough dose levels where you could realistically see a difference.”
One way to get around the need to guess is to follow blood levels of the drug. “If you look at the blood levels of the drug, and dose level one looks exactly the same as dose level two and dose level three, there’s not much point in trying to differentiate between those three,” she said. Instead, she said investigators should choose doses that show some differences in terms of blood levels, pharmacokinetics, or other parameters while still showing signs of activity.
Seymour further cautioned that clinical trial patients differ from patients in the general population who will get the drug. Study participants are typically healthier and committed to participating in the trial, often to the point that they won’t admit to problematic side effects because they want to stay on the trial.
That can result in more dose reductions in patients receiving treatment outside of a trial, which sponsors may then try to compensate for in the real world with higher pricing, Seymour said. “If a company is developing a drug and there’s a risk people will use the drug at a half dose, they’re either going to lose profit or they’re going to pitch the price of the drug much higher just in case it’s used at a lower dose, which increases societal costs of access to effective therapies,” she said. “And if you end up actually using it at the approved dose, then you’ve got a drug that’s double the price it should be.”
Challenges for smaller companies
While large pharmaceutical companies are embracing dose-optimization strategies, if imperfectly, the added complexity and cost of those trials remain challenging for small biotech companies with more limited resources.
“I don’t think big pharma is going to resist, because big pharma knows how to develop drugs,” said Ratain. “It’s the little biotech companies that have timelines, and valuations based on those timelines, [and] they’re going to say, ‘We don’t have the resources.'”
Those tensions have put smaller companies like Oxford, UK-based Exscientia in the vanguard of clinical trial design innovation. Exscientia uses artificial intelligence to streamline and accelerate development of its pipeline of targeted therapies. In November, the Belgium Federal Agency for Medicinal and Health Products gave the company the green light to begin its IGNITE-AI Phase I/II trial of an A2A receptor antagonist EXS-21546 with anti-PD-1 therapy in patients with relapsed or refractory renal cell carcinoma and NSCLC. While evaluating safety, efficacy, pharmacokinetics, and pharmacodynamics in up to 110 patients, Exscientia will be testing an A2A biomarker signature for patient enrichment purposes.
In order to optimize the trial for cost efficiency and, more importantly, for patient benefit, Exscientia went with an adaptive dose-optimization design over a traditional 3+3 design or a randomized fixed-parallel dose response study. In the latter trial design, an equal number of patients would be randomized to each dose level. In an adaptive trial design, modifications to the trial can be made after it begins based on data as it is collected or at an interim analysis.
Exscientia is adding a layer to its adaptive dose-finding design by using modeling to guide the assignment of patients into dose cohorts. Rather than assigning an equal number of patients to each dose cohort, investigators are using a continuous reassessment model to allocate patients across those doses. The algorithm will use accruing data in the trial to establish an MTD of EXS-21546, as well as the best dose to use in the expansion phase, taking into consideration data regarding drug safety, patient response, pharmacokinetics, and pharmacodynamics.
Model-informed adaptive trial design is “nothing new,” said Mike Krams, Exscientia’s chief quantitative medicine officer, pointing out that MD Anderson Cancer Center is using this type of design in its early development programs.
“Identifying the maximum [tolerated dose] is an appropriate research question to ask,” said Krams. “The more important question to ask after that is which dose and treatment regimen is ideally suited to create benefit for patients … integrating across a number of different endpoints, not just safety.”
While Krams acknowledged that Exscientia’s model-informed adaptive design saves money, he emphasized that its purpose is efficiency. “[Money] is an irrelevant point,” said Krams. “We must look at the value provided to patients within the clinical trial and the future. If I need to find some knowledge and I can either achieve this by working with 10 patients or 10,000 patients … to work with more than 10, that’s basically just an expression of inefficiency.”
The FDA’s new draft guidance on dose optimization covers most clinical trial situations in oncology, but not all. The guidelines do not address dose optimization for radiopharmaceuticals, cellular and gene therapy products, or cancer vaccines, for example.
Targeted radiopharmaceutical therapies lend themselves to dosing strategies that are unique for several reasons, according to Ratio Therapeutics CEO Jack Hoppin. Ratio is developing a prostate-specific membrane antigen-targeted radiopharmaceutical agent for metastatic prostate cancer in partnership with Bayer and developing an imaging agent with Lantheus to identify fibroblast activation protein (FAP)-expressing cancer cells on PET scans.
With radiopharmaceuticals, toxicity is driven by radiation exposure, not by the pharmacology of the nonradioactive targeting vector carrying the isotope. There is an established dose of radiation that normal tissues can tolerate before there are observable negative effects based on external beam radiation. Additionally, imaging can be used to nail down exactly how much radiation exposure a given tissue is receiving based on how long the radioisotope is present. The need to manage the radiation dose, rather than a direct toxic effect, changes the requirements of dosing trials.
“As we become better at designing targeted radiotherapies with greater therapeutic indices, we can extend the number of cycles given to a patient, extending the usefulness of the therapy for a given disease,” said Hoppin. “We do not need to push the concept of MTD in a similar fashion to conventional chemotherapies.” Still, he noted that the burgeoning radiopharmaceutical field needs to establish the normal tissue tolerance of such therapies since “the quality of radiation differs from external beam radiotherapy and the amount of radiation that is given over time is also very different.”
Also outside the purview of the FDA’s guidance are considerations for optimizing doses of already-approved therapies, which concerns Loeser. While companies developing new drugs can now expect to be challenged by the FDA if they don’t collect sufficient data on optimal dosing, she suspects the draft guidance is unlikely to spur drugmakers to optimize dosing for marketed therapies that they’ve developed using an MTD approach and that patients are already receiving.
“If I’m a pharmaceutical company and my drug is already on the market and being prescribed, what incentive do I have to go back and do dose-finding studies on it to determine the best dose based on safety and efficacy?” said Loeser. “We’re up against a huge challenge.”
Loeser said doctors and researchers may be able to make some calculations using data from past trials or real-world data on patients who had doses reduced due to toxicities, and estimate the optimal dose of marketed drugs. “Maybe we could get some learnings on that, but that’s kind of a long shot,” she said.