An adaptive design—and mindset—can accelerate early-phase NSCLC development

Lung cancer, primarily non-small cell lung cancer (NSCLC), remains the most common and deadly cancer worldwide.¹ Since 2020, the FDA has approved 43 new drugs to treat lung cancer. Yet, patients with advanced disease still face significant unmet needs.² Major pharmaceutical companies and emerging biotechs are racing to develop the next generation of NSCLC drugs.

However, two recent advances in basic science and regulatory policy impact sponsors. First, scientists have long recognized NSCLC as a group of related but distinct diseases rather than a single histological entity. This understanding has expanded as the remarkable molecular and cellular diversity of NSCLC, highlighted by numerous biomarkers, has become clearer. Developers must therefore decide early in the development process whether to target specific biomarker-driven subgroups or all patients. Second, the FDA’s Project Optimus, launched in 2021, requires comprehensive dose optimization for every anticancer drug.

In the highly competitive NSCLC therapeutic landscape, adaptive trial designs and adaptive decision-making are increasingly essential for accelerating development and ensuring commercial viability.

Slow down to speed up

The FDA expects that thorough dose optimization will benefit patients by ensuring that cancer drugs are more effective and safer. Recent studies have found that treating patients with lower doses of anticancer drugs may reduce toxicities, allowing patients to remain on treatment longer and ultimately leading to improved efficacy.³

Before Project Optimus, the goal of dose escalation was to find the maximum tolerated dose (MTD). The goal now is to determine a range of effective doses for further optimization. Dose optimization requires sponsors to enroll more patients, gather more data, conduct additional analyses, and allocate more time and resources to early-stage trials. The new requirements for dose optimization can be challenging for small biotechs with limited funding and in-house expertise.

At Parexel, we advise sponsors to use an integrated, adaptive first-in-human trial design that mitigates risks by systematically removing the uncertainties at each stage of early drug development: target validation, disease selection, and dose optimization.

Recently, we met with a biotech company looking for a simple dose-escalation trial to show progress and attract funding. They thought that an adaptive Phase 1-2 trial with escalation, optimization, and expansion was too complicated and slow. However, this was shortsighted because the FDA later required a more complex adaptive design with multiple arms and doses. Starting with a seamless, adaptive trial from the beginning saves time overall. Failing to comply with the FDA’s Project Optimus could lead to years-long delays in drug development. Spending a few months designing a data-rich trial reduces risks and speeds up development, which can significantly increase an asset’s value. Informed investors understand the importance of a comprehensive drug development plan that includes dose optimization.

Key elements of adaptive trial designs

Efficient study designs can reduce the time spent on dose escalation, optimization, and expansion by combining proof-of-concept and dose-finding. Successful adaptive trials in early-phase NSCLC incorporate six key elements:

1. Initial broad patient cohorts: Early trials may start with a broad "all comers" approach for solid tumors before narrowing the focus to NSCLC and then to specific biomarker-defined subgroups as data accumulates. This enables researchers to assess safety and tolerability, and explore the drug's effect across various populations, potentially avoiding missteps in defining patient populations and biomarker strategies too early or narrowly. Recently, we worked on such a Phase 1-2 trial that began by enrolling patients with solid tumors and progressively narrowed its focus to NSCLC based on accumulating data. The patient population was ultimately refined to include only those with specific mutations.
2. Dose escalation: Model-assisted designs, such as the Bayesian Optimal Interval (BOIN) or Modified Toxicity Probability Interval (mTPI-2) design (Figure 1), are often preferred for dose escalation over traditional 3+3 designs, which can yield inaccurate and variable MTD estimates.⁴ It is operationally inconvenient to limit each cohort to three patients at a time, and there is a risk of missing the target by random chance or rule-based decision-making. In a BOIN design, there is no assumption about the dose-toxicity curve, and a transparent pre-calculated decision table makes it easy to escalate or de-escalate doses based on dose-limiting toxicities (DLTs). There is flexibility in target toxicity (versus the fixed rate of less than 33 percent in a 3+3 design) and in the number of patients per cohort (not limited to 3 or 6 patients). "Backfilling" cohorts at promising, safe dose levels allows for the collection of more data on safety and preliminary signals for antitumor activities.⁵ However, backfilling strategies require careful planning and cross-functional coordination to obtain preliminary evidence on the target validation, disease selection, and dose levels for the randomization dose-finding part.
3. Dose optimization and expansion: At Parexel, we recommend including backfills during dose escalation to identify an efficacious dose range from which two or more dose levels are selected for the dose optimization part in a homogenous disease setting. The agency now typically requires a randomized comparison of at least two dose levels in dose optimization studies to eliminate any potential bias in dose selection based on safety and efficacy outcomes. However, trials do not need to be powered to demonstrate statistical superiority. At Parexel, we recommend that sponsors consider a formal expansion design for each dose cohort, such as the Bayesian Optimal Phase 2 (BOP2) design or Simon’s two-stage design with stopping rules to potentially select the optimal dose at an interim analysis (Figure 1).

Figure 1. Seamless Phase I/II trial design incorporating dose optimization

4. Real-time data analysis and decision-making: A crucial component of adaptive design is the ability to analyze data in real time to make informed decisions quickly. This might include evaluating intermediate doses based on observed adverse events and efficacy signals. For instance, in one recent trial, we introduced two intermediate doses after observing toxicities at higher planned doses. We also added a pre-medication regimen to manage adverse events better. This extended the trial’s timeline. Although the sponsor had initially expected to complete dose escalation within a year, they adjusted their development plan after considering the benefits of thorough dose optimization. They sailed through their End-of-Phase 2 (EOP2) meeting with the FDA due to the breadth of their optimization data and careful planning.
5. Pre-planned adaptations: The ability to make changes is not arbitrary but is governed by predefined rules within the protocol. This includes criteria for adding or dropping dose cohorts, which is a hallmark of adaptive design. Adaptive designs must be supported by predefined criteria and real-time decision-making.
6. Combination therapies: The potential for combination therapies should be considered early in the development process, as many new drugs may be most effective when used with other treatments. The trial design can be adapted to include cohorts that explore these combinations. The doses to be combined can be adjusted based on emerging monotherapy data.

Key elements of an adaptive mindset

An adaptive mindset incorporates three key elements:

1. Adapt to emerging data: Companies must adjust to incoming data at both the trial and program levels to make timely, crucial decisions. For example, many emerging companies might initially draw inspiration from non-clinical data or competitors. They could then develop a positive bias toward a specific target indication without considering expansion to other potential indications. At Parexel, we adopt an integrated approach to drug life-cycle management from the start by asking: Are you certain about the indication? Do you have the flexibility to test additional tumor types? Phase I/II is the ideal time to gather data from multiple indications and explore differences across tumor types. Clinical pharmacokinetics/pharmacodynamics (PK/PD) data collected during the dose-escalation and backfilling phases can help sponsors extrapolate the optimal dose found in the first indication to other indications, saving valuable time and resources.
2. Adapt to regulatory policy changes: Project Optimus has brought about a seismic shift in regulatory policy for oncology drugs, necessitating a more thorough exploration to identify a dose with maximal efficacy and acceptable safety in the target indication. Sponsors must be proactive, data-driven, and flexible to align with regulators’ evolving expectations. Companies that fail to adapt may face significant delays, including requests for a randomized dose-finding study during development, as well as refuse-to-file (RTF) decisions and complete response (CR) letters during the FDA's review of the marketing application. It is also a mistake to underestimate the operational complexity of adaptive designs. The successful implementation of an adaptive trial hinges on proactive cohort management, timely data reporting, and constant communication with sites to address challenges and maintain momentum. In one recent Phase 1 trial of a new agent for NSCLC, Parexel’s trial team engaged in frequent, in-depth meetings—sometimes for 8 to 10 hours a week—to review data, discuss patient safety, and strategize on enrollment.
3. Adapt to cross-functional, flexible decision-making: A sound corporate decision-making structure must evaluate information from all sources and be open to solutions for dose optimization, target validation, and indication selection. While many companies employ adaptive approaches, success is determined not only by the design of a trial but also by its execution. A key differentiator is the ability to leverage a multidisciplinary team of medical, regulatory, and statistical experts who can interpret data in real-time. This expertise is crucial for making pivotal decisions during a study, such as when to initiate backfilling, how to select patient cohorts, and when to approach regulatory agencies to prevent development delays. This strategic know-how provides a competitive advantage over firms that may rely more heavily on large-scale data alone.

Adaptation is a competitive advantage

Using adaptive trial designs in early-phase NSCLC trials is not just about statistical methods; it is a strategic necessity that requires a comprehensive approach. By combining deep scientific and regulatory expertise with operational excellence, it is possible to navigate the complexities of NSCLC drug development and speed up the delivery of promising new therapies to patients. Sponsors may benefit from external experts to help determine when to make strategic adaptations, when to engage regulators, and how to justify dose selection.

Disclaimer:
Parexel provides the information contained in this document for educational purposes only. The information does not constitute legal or regulatory advice. Readers should not act upon this information without seeking advice from professional advisers.

Contributing Experts