Maximal Tolerated Dose (MTD) to Recommended Phase 2 Dose (RP2D) - a shift in early oncology trial designs

 As the field of oncology moves from systemic cytotoxic chemotherapies to targeted agents and immunotherapies, the paradigm for dose selection is undergoing a historic shift. For decades, the Maximum Tolerated Dose (MTD) was the "gold standard" for early-phase trials, but today’s clinical trialists and statisticians are increasingly prioritizing the Recommended Phase 2 Dose (RP2D) as a more robust and patient-centric metric.

This evolution is spearheaded by the FDA’s Project Optimus, which emphasizes "dose optimization" rather than simply finding the highest dose a patient can survive.

From "More is Better" to "The Optimal Balance"

The traditional MTD-centric approach was built on the assumption that a drug's efficacy increases linearly with its toxicity—a rule that often held true for classical chemotherapy. However, for modern targeted therapies, the Optimal Biologic Dose (OBD)—the dose that achieves maximum target saturation—often occurs well below the MTD.

Feature	Maximum Tolerated Dose (MTD)	Recommended Phase 2 Dose (RP2D)
Focus	Toxicity-driven; finding the safety ceiling.	Value-driven; finding the therapeutic "sweet spot".
Observation	Short-term (Cycle 1) Dose-Limiting Toxicities (DLTs).	Long-term tolerability, PK/PD, and cumulative safety.
Assumption	Efficacy increases with dose ("More is Better").	Efficacy may plateau while toxicity continues to rise.
Clinical Utility	A safety guardrail to prevent overdosing.	A strategic decision for registrational success.

Why RP2D is Preferred over MTD

For the modern statistician, the RP2D represents a "totality of evidence" that the MTD simply cannot provide:

• Sustainability vs. Intensity: MTD focuses on what a patient can tolerate for 21 days. In contrast, RP2D considers the long-term tolerability necessary for chronic treatment, preventing premature discontinuations that can derail a trial's efficacy results.
• The Sotorasib Lesson: FDA reviews, such as those for sotorasib, have highlighted the "dosing conundrum" where initial MTD-based doses led to excessive toxicity, eventually requiring post-market studies to find a more optimal, lower dose.
• Target Saturation: Modern agents often reach a Pharmacokinetic (PK) plateau where increasing the dose adds no therapeutic benefit but significantly increases the rate of low-grade, chronic toxicities.
• Dose-Response Nuance: As discussed in previous explorations of Determining the Dose in Clinical Trials, while the MTD is a safety limit identified through escalation, the RP2D is a comprehensive recommendation for further evaluation that aims to expose as few patients as possible to intolerable doses.

The Statistical Shift: Beyond 3+3

To find a true RP2D, statisticians are moving away from the rigid "3+3" rule-based designs to more flexible, model-informed approaches. These include:

• Bayesian Optimal Interval (BOIN) designs that allow for a more nuanced exploration of the therapeutic window.
• Randomized Dose-Ranging Studies: Encouraged by Project Optimus, these trials evaluate multiple doses early to compare safety and efficacy side-by-side.
• Dose Expansion Cohorts: Used to refine the RP2D by gathering deeper data on preliminary efficacy and late-onset toxicities in specific patient subgroups.

Conclusion

The shift from MTD to RP2D is more than a regulatory requirement; it is a clinical necessity. By identifying an optimized RP2D early, sponsors can avoid the "safety pitfalls" of MTD, improve patient quality of life, and build a stronger evidence chain for final approval. In the era of precision medicine, finding the right dose for the right patient is just as important as finding the right drug.

Excessive number of clinical trial protocol amendments due to complex trial design

In a previous blog post "Protocol amendment in clinical trials", I discussed the impact of protocol amendments on the clinical trial performance and cost and the reasons for driving the protocol amendments. Protocol amendments are unavoidable, but we can try to think about the study design and execution proactively to minimize the number of protocol amendments. Sometimes, the excessive number of protocol amendment are driven by Complex Innovative Trial Design or CID in short (for example, adaptive design, basket/umbrella/platform trial design, expansion cohort design, Bayesian design...).

We noticed an extreme case of a clinical trial with the study protocol amended 50 times. This refers to Study P001 (also known as KEYNOTE-001, NCT01295827) by Merck, which was a large, multi-cohort Phase 1 trial with numerous expansion cohorts that supported the initial accelerated approval of pembrolizumab. in Statistical Review and Evaluation, BLA 125514, FDA Center for Drug Evaluation and Research, August 2014, The FDA reviewer noted this high number of amendments while discussing the complexity of the trial design. KEYNOTE-001 was a massive "seamless" adaptive trial that evolved from a traditional Phase 1 dose-escalation study into a large study with multiple expansion cohorts (Part A, A1, A2, B, C, D, etc.) covering different tumor types (Melanoma, NSCLC) and dosing regimens. The "50 times" figure likely includes all global and country-specific amendments up to the time of the BLA submission in February 2014.

The high number of protocol amendments for KEYNOTE-001 was a direct result of its innovative, "seamless" adaptive study design. Initially launched as a standard Phase 1 dose-escalation trial, the study evolved into a massive, multi-cohort trial that eventually enrolled 1,235 patients.

The 50 amendments occurred primarily due to the following reasons:

• Addition of Expansion Cohorts: As early data showed promising results, the protocol was repeatedly amended to add new expansion cohorts for specific tumor types, most notably melanoma and non-small cell lung cancer (NSCLC).
• Sample Size Increases: Striking patient responses led investigators to increase sample sizes within existing cohorts to better evaluate efficacy endpoints like overall response rate (ORR).
• Adaptive Dosing Changes: The protocol was amended to change dosing regimens based on emerging safety and efficacy data. For example, Amendment 7 changed dosing from every two weeks (Q2W) to every three weeks (Q3W), and Amendment 10 shifted all participants to a fixed dose of 200 mg.
• Biomarker Integration: Amendments were used to add co-primary endpoints related to PD-L1 expression after researchers observed its correlation with drug efficacy. This included the validation of a companion diagnostic assay.
• Regulatory Speed: This "seamless" approach allowed Merck to skip traditional Phase 2 and 3 steps for certain indications, leading to the first-ever FDA approval of an anti-PD-1 therapy.

While efficient, the FDA's statistical reviewers noted that such frequent changes (averaging more than one amendment per month during the most active phases) created significant operational and analytical complexity for the trial. The main challenges in analyzing the KEYNOTE-001 trial data, as noted in the FDA's statistical and medical reviews, stemmed from the extreme complexity of a "seamless" design that was modified more than 50 times. 

The primary analytical hurdles included:

• Statistical Integrity and Type I Error Risk: The frequent addition of new cohorts and subgroups—often based on emerging data—increased the number of statistical comparisons. This raised concerns about "multiplicity," where the probability of finding a significant result by chance (Type I error) increases with every new hypothesis tested.
• Operational and Data Management Complexity: Maintaining data quality was difficult when different sites were often operating under different versions of the protocol simultaneously. The FDA noted that this led to potential adherence issues and made it difficult to isolate single cohorts for clean, standalone submissions.
• Shifting Dosing and Regimens: The trial transitioned from weight-based dosing (2 mg/kg or 10 mg/kg) to a fixed dose (200 mg) and changed the frequency of administration (every 2 weeks to every 3 weeks) mid-study. This required complex "pooled analyses" to prove that efficacy and safety were consistent across these varying schedules.
• Biomarker Selection and Validation: The protocol was amended to include a PD-L1 companion diagnostic while the study was already underway. This created a challenge in defining "training" vs. "validation" sets within the same trial population to establish the diagnostic's cutoff levels without introducing bias.
• Lack of a Control Arm: Because the trial was essentially a massive Phase 1 expansion, it lacked a randomized control arm for several indications. This forced reviewers to rely on cross-trial comparisons and historical data, which are inherently more prone to bias than randomized controlled trials (RCTs).
• Patient Selection Bias: The "adaptive" nature allowed for rapid accrual in specific successful cohorts, which, while beneficial for speed, made it difficult to ensure the final patient population was representative of the broader real-world population.

Although the excessive number of protocol amendments, the results from the KEYNOTE-001 resulted in the FDA approval of pembrolizumab in the treatment of multiple tumor types. KEYNOTE-001 study was also the basis for the NEJM article "Seamless Oncology-Drug Development" by Prowell, Theoret, and Pazdur.

One-way versus two-way tipping point analysis for robustness assessment of the missing data

Tipping point analysis (TPA) is a key sensitivity analysis mandated by regulatory agencies like the FDA to assess the robustness of clinical trial results to untestable assumptions about missing data. Specifically, it explores how much the assumption about the missing not at random (MNAR) mechanism would have to change to overturn the study's primary conclusion (e.g., a statistically significant treatment effect becoming non-significant). See a previous blog post "Tipping point analysis - multiple imputation for stress test under missing not at random (MNAR)"

One-Way Tipping Point Analysis for Robustness Assessment

A one-way tipping point analysis is a sensitivity method used to evaluate the robustness of a study’s primary findings by systematically altering the missing data assumption for only one treatment group at a time—most commonly the active treatment arm. While the missing outcomes in the control group are typically handled under a standard Missing at Random (MAR) or Jump to Reference assumption, the missing outcomes in the active arm are subjected to a varying "shift parameter" (δ). This parameter progressively penalizes the imputed values (e.g., making them increasingly worse) until the statistically significant treatment effect disappears, or "tips." By identifying this specific value, researchers can present a clear, one-dimensional threshold to clinical experts and regulators, who then judge whether such a drastic deviation from the observed data is clinically plausible or an unlikely extreme.

Two-Way Tipping Point Analysis for Robustness Assessment

A two-way TPA is an advanced method to assess robustness by independently varying the missing data assumptions for both treatment groups (e.g., the active treatment arm and the control/reference arm).

Missing Data Assumptions (MAR vs. MNAR)

The two-way TPA is used to assess the robustness of the primary analysis, which is typically conducted under the assumption of Missing at Random (MAR).

• Missing at Random (MAR): Assumes that the probability of data being missing depends only on the observed data (e.g., a patient with a worse baseline condition is more likely to drop out, and we have observed the baseline data).

• Missing Not at Random (MNAR): Assumes that the probability of data being missing depends on the unobserved missing outcome data itself (e.g., a patient drops out because their unobserved outcome has worsened more than what is predicted by their observed data).

Robustness Assessment

The two-way TPA evaluates robustness to plausible MNAR scenarios. This is done by imputing the missing outcomes (often starting with an MAR method like Multiple Imputation) and then applying a systematic, independent "shift parameter" (or δ) to the imputed values in each arm.

• Process: The shift parameters (δActive and δControl) are varied systematically across a two-dimensional grid, typically in a direction that reduces the observed treatment effect.

• Tipping Point: The δActive and δControl values at which the primary conclusion (e.g., statistical significance) is "tipped" or overturned define the tipping point.

• Robustness: The larger and/or more clinically implausible the combination of shift parameters required to overturn the conclusion, the more robust the original result is considered to be under different MNAR assumptions.

Two-Way Tipping Point Result Tables

The results of a two-way TPA are typically presented as a grid or heat map table where:

• One axis represents the shift parameter applied to the missing outcomes in the Active Treatment arm (δActive).

• The other axis represents the shift parameter applied to the missing outcomes in the Control/Reference arm (δControl).

• The cells of the table contain the resulting p-value or estimated treatment difference for that specific combination of assumptions.

The goal is to find the boundary of the grid where the result crosses the significance threshold (e.g., p >= 0.05 or the lower bound of the confidence interval crosses the null value).

---

Comparison: One-Way vs. Two-Way Tipping Point Analysis

The choice between one-way and two-way TPA is a trade-off between simplicity and comprehensiveness.

Feature	One-Way Tipping Point Analysis	Two-Way Tipping Point Analysis
Missingness Assumption	The shift parameter (δ) is only applied to one arm, usually the active treatment group, while the missing data in the control arm are imputed based on the MAR assumption (e.g., Jump to Reference).	Independent shift parameters (δActive and δControl) are applied to both arms simultaneously.
Sensitivity Explored	Explores MNAR scenarios where dropouts in one arm have systematically worse/better outcomes than assumed by MAR, relative to the other arm's MAR assumption.	Explores a two-dimensional space of MNAR scenarios, allowing dropouts in both arms to vary independently.
Complexity	Simpler to calculate and interpret (one dimension).	More computationally intensive and complex to interpret (two-dimensional grid).
Plausibility	Often viewed as less comprehensive, as it does not model the possibility of simultaneous, independent MNAR mechanisms in both arms.	Considered more comprehensive as it allows for a wider range of clinically plausible and implausible MNAR scenarios.
Result Presentation	A line plot or simple table with a single 'tipping point' value.	A grid/matrix table or heat map showing the boundary of non-significance.

In essence, the two-way TPA is generally preferred by regulatory agencies for its superior ability to assess robustness because it explores a more realistic and exhaustive range of asymmetric MNAR mechanisms.