New draft guidelines from the FDA endorse the use of multiple expansion cohorts in first-in-human clinical trials as a means of speeding lifesaving medications to market, but the document warns that the difficulties and dangers of using such complex studies so early in the testing process necessitate unusually careful design and oversight.
New draft guidelines from the FDA endorse the use of multiple expansion cohorts in first-in-human (FIH) clinical trials as a means of speeding lifesaving medications to market, but the document warns that the difficulties and dangers of using such complex studies so early in the testing process necessitate unusually careful design and oversight.
“Sponsors and investigators have found in recent years that they could save significant amounts of time and money by increasing the size and scope of initial studies rather than formally closing each study and going through the whole process of getting another few studies approved. However, there was a tendency to expand in a relatively unplanned way that didn’t always lead to the best data,” said Joseph Paul Eder, MD, director of the Early Drug Development Program at Yale University. “This seems like the FDA’s way of acknowledging the value of multicohort expansions and trying to help sponsors and investigators do them right.”
Typical phase I trials use 20 to 80 patients to determine the pharmacokinetic and pharmacologic actions of an investigational drug as well as the adverse effects associated with increasing doses. Their main goal is to uncover the information needed to design safe, well-controlled, and scientifically valid phase II studies of safety and efficacy. Any preliminary efficacy data uncovered by a phase I study are a bonus (FIGURE).
FIH trials that employ multiple, concurrently accruing patient cohortseach designed to assess different aspects of the safety, pharmacokinetics, and antitumor activity of a particular drug—seek to cut years from drug development time by seamlessly moving from the initial determination of pharmacokinetics, pharmacologics, and dosing to the estimation of antitumor activity and safety.
However, the FDA believes that rapid enrollment of hundreds (or thousands) of patients in very early trials exposes more people to relatively unknown toxicities, while the necessity of doing such trials at multiple sites, often spread around the world, makes it hard to communicate toxicity data immediately and adjust dosing accordingly. Moreover, potential on-the-fly adjustments to treatment protocols, data collection, and other factors can compromise the validity of data from poorly planned trials.
The FDA’s new guidance applies to trials with at least 1 traditional FIH dose-escalation cohort and at least 3 other cohorts, with cohort-specific objectives such as assessing antitumor activity in a disease-specific setting, establishing safety and efficacy in special populations (eg, pediatric cohorts), testing alternate dosing schemes, validating biomarkers, and combining the experimental medication with some other drug.
The dose-escalation cohort usually begins first, but the additional cohort studies almost always get under way before the dose-escalation phase has produced conclusive data, so patients in all cohorts are exposed to what is essentially an untested drug. The FDA believes that the obvious risk of such trials necessitates that they be undertaken only in patients with serious, incurable diseases. Agency experts also believe that some types of experimental drugs (eg, immediate-release oral products, nonliposomal injections) are probably more appropriate than others, such as those with large inter- and intrapatient variability.
Protocols for Assessing Efficacy
No matter the patient population or the drug type, sponsors or investigators hoping to win approval for FIH trials with multiple expansion cohorts should provide a robust rationale for such a trial design, demonstrate that their experimental drug has the potential to meet the criteria for breakthrough therapy, and provide a scientific basis for including each of their proposed cohorts.
Trial plans should quantify the acceptable risks for each cohort’s population based on that cohort’s prognosis, treatment alternatives, and comorbidities. They should also include a plan for statistical analysis that justifies each protocol’s size, as well as its data-collection protocol, and contains rules for stopping treatment because of lack of observed activity.
The draft guidance warns that different types of cohorts will require different types of precautions to ensure patient safety and good data. For example, cohorts that assess the effects of food consumption, organ dysfunction, or concomitant treatments on drug performance will typically require at least preliminary results on pharmacokinetics and safety from the dose-escalation study before they can get under way. Cohorts designed to optimize dosage should randomize patients among 2 or more dosages and justify group sizes, while cohorts designed to validate biomarkers should use biomarkers that are logically related to the experimental drug’s action and are analytically validated by prior studies.
The FDA strongly recommends the inclusion of pediatric cohorts whenever significant numbers of children suffer from untreatable cancers and could benefit from the experimental drug because prospective inclusion of such cohorts (rather than separate dose escalation and activity estimating phase I trials) can shorten development timelines and increase options for an underserved patient population. However, the agency also warns that pediatric cohorts require extra planning and precaution. Except under special circumstances, pediatric cohorts should launch only after other cohorts have established safe adult dosages, enroll older children rather than younger children, include extra monitoring to evaluate and adjust dose size, and include only children with incurable cases of relapsed or refractory disease.
Launching multiple cohorts without complete phase I trial data make it far harder to select appropriate patient populations, dose amount, and dosage timing, so all but the best designed trials can possibly expose patients to needless risks, miss potential benefits, or produce misleading results. Designers must state each cohort’s goals in advance and show that their plans for cohort size, patient selection, data collection, and statistical analysis will meet those goals. Plans for cohorts that test antitumor activity should prespecify the magnitude of antitumor activity that would warrant further evaluation of the drug. In a nonrandomized cohort, the FDA recommends the use of a Simon 2-stage design1 to evaluate activity and limit exposure to ineffective drugs. In cohorts in which patients are randomized between groups, patient populations must be large enough to demonstrate preselected signs of activity, with statistical significance, in reasonable time frames.
The draft guidelines mention many potential pitfalls of multicohort trial design, and they warn investigators repeatedly against the temptation to make direct comparisons between or among different cohorts. Prior trials have attempted to make such comparisons to reach conclusions about the experimental medication that could not be made from any single cohort’s results. However, except when such analyses are part of a protocol design that is reviewed and approved by expert bodies, the FDA believes that valid comparisons between or among different cohorts will be nearly impossible to achieve. Each cohort, therefore, will generally produce results that are entirely independent of any other.
Protocols for Assessing Safety
There is, of course, a single major exception to this injunction: using information about medication toxicity in 1 cohort to protect patients in the other cohorts. The draft guidelines actually urge trial designers to create protocols for sharing safety information across cohorts in something approaching real time. Indeed, the FDA believes that the potential hazards of exposing many patients to untested drugs necessitate that FIH multicohort expansion trials all incorporate systematic procedures that ensure rapid communication of all potential safety issues to all investigators as well as regulators and institutional review boards.
Before trials even begin, trial designers should agree with regulators and institutional review boards about which adverse events should lead to the immediate termination of any or all the cohorts, which should lead to certain preselected modifications in treatment regimens, and which would be considered tolerable enough to continue with unmodified protocols. Trial designers can submit their plans to the institutional review board at each participating treatment center, but to facilitate multicenter trial, the draft guidelines recommend a single, central institutional review board for each trial.
Trial designers should then recruit medical monitors who have both training and experience in cancer treatment, as well as clinical trials, and they should instruct all clinicians who sign on as coinvestigators in each trial’s chosen system for reporting adverse events.
“Early-stage trials used to be so small that they could be handled entirely by specialists,” Eder said. “These larger trials require participation from far more oncologists, and although the early-stage specialists will still handle the intensive safety monitoring you see in dose-escalation trials, clinical oncologists who choose to participate in these expansion trials may have to do more safety monitoring than is typically required in the sort of late-stage trials that most clinical oncologists are more familiar with.”
The FDA believes that reports of adverse events should go to either an independent safety assessment committee or an independent data monitoring committee.
These independent committees should conduct real-time review of all serious adverse events and periodically produce both summaries of all adverse events and suggestions for protocol modifications that would reduce risks for all trial patients. Possible modifications could include changing trial eligibility criteria if certain patient populations suffer higher-than-expected rates of adverse events, adjusting the dose or timing of the experimental medication or amending information given to patients about the possible risks of trial participation. Indeed, informed consent documents should be updated regularly as new safety information becomes available, and higher-than-expected rates of serious adverse events should trigger the submission of those amended consent forms to patients who are already participating in the trial.
Adverse events that meet prespecified criteria, moreover, should be reported immediately to regulators and all participating investigators. Even in instances in which adverse events do not rise to a level that necessitates immediate notification to the FDA, the draft guidelines support protocols for routinely submitting cumulative safety summaries on some preselected timeline that is more frequent than annual.
“There is certainly a greater burden for clinical practitioners conducting these early-phase trials than there is in doing the late-phase registration trials. Necessarily closer monitoring is required, which is something to be aware of, but the standards discussed in these guidelines are certainly manageable for physicians who want to take part in what is very definitely valuable work,” said Howard A. Burris III, MD, founder of the Sarah Cannon Research Institute in Nashville, Tennessee, and president-elect of the American Society of Clinical Oncology.
“Overall, these seem like reasonable guidelines that balance the desire for speed with the need for safety and rigor,” said Burris.
Simon R. Optimal two-stage designs for phase II clinical trials.Control Clin Trials.1989;10(1):1-10.