Contributed Commentary By Ellen Leinfuss 

April 4, 2017 | While a lot has been written about late-phase clinical trial failure, it often seems like a gambling weekend in Las Vegas has better odds than a sponsor has of getting a new drug approved. Even the enormous advances in technology and biology during the past few years, coupled with the industry’s vast operational experience at designing clinical trials, have not altered the success curve in any meaningful way: The phase 2 to phase 3 success rate is only 30.7%, while the phase 3 to New Drug Application/Biologic License Application (NDA/BLA) success rate is just 58.1% (and another 15% still fail to gain approval following the NDA/BLA filing).

After years of significant slope change, culminating in a 19-year high of 45 US Food and Drug Administration (FDA) NDA/BLA approvals in 2015, the bottom looked like it dropped out in 2016. With only 22 approvals, the class of 2016 was below both the five- and ten-year average. And while the agency claims that five drugs were pushed forward and reviewed in 2015 instead of 2016, there were also 14 drugs rejected during 2016. 

Failure rates are hard to properly bracket as they vary widely by both therapeutic area (higher for oncology for example) and type of molecule (higher for small molecules, lower for monoclonal antibodies). But it’s clear that challenges in phase 2 are often carried forward into phase 3, and for phase 3, impacting the likelihood of success in getting a drug approved.

FDA Weighs in on Phase 2 to 3 Divergence

In January 2017, FDA published a paper entitled, “22 Case Studies Where Phase 2 and Phase 3 Trials Had Divergent Results.” The paper’s objective was to demonstrate that even when phase 2 results are promising, phase 3 trials are necessary to prevent patients from being exposed to unsafe or ineffective treatments. This paper appears to be a response to recent criticism about the agency’s speed in approving new drugs and questions regarding the benefits of conducting expensive and time-consuming phase 3 trials. 

“In recent years, there has been growing interest in exploring alternatives to requiring phase 3 testing before product approval, such as relying on different types of data and unvalidated surrogate endpoints,” the FDA wrote. “To better understand the nature of the evidence obtained from many phase 2 trials and the contributions of phase 3 trials, we identified, based on publicly available information, 22 case studies of drugs, vaccines, and medical devices since 1999 in which promising phase 2 clinical trial results were not confirmed in phase 3 clinical testing.”

The study concludes that phase 3 testing did not confirm phase 2 findings of effectiveness in 15 of the 22 case studies; effectiveness and safety in seven cases; and safety alone in one. “These unexpected results could occur even when the phase 2 study was relatively large and even when the phase 2 trials assessed clinical outcomes,” the FDA reported. Further, the agency said two of the experimental products actually worsened the problems they were intended to treat.

The Role of Modeling and Simulation

What’s the solution? In a November 2016 paper in Clinical & Pharmacology Therapeutics, written by representatives from FDA, industry, and academia, modeling and simulation (M&S) is named as a pivotal tool for reducing late-stage attrition. “Many of these limitations in drug development are the result of scientific challenges to predicting efficacy and safety, or characterizing sources of response variability for a drug compound at early, less expensive stages of discovery. The tools, methods, and frameworks (e.g. mechanistic or quantitative) of clinical pharmacology span many distinct sub-specialties and can [have] a significant impact at the interface of these nonclinical and clinical phases. They can greatly reduce uncertainty related to therapeutic targets, dosing, and patient populations in which the novel compound may have the most efficacy.”

M&S combines two transformative and constantly-evolving technologies: computer-aided mathematical simulation and biological sciences. It can be applied at every step along the drug development pathway from discovery through pre-clinical and clinical studies to patient care. It advises new drug candidate selection, first-in-human dosing, and labeling language regarding possible drug-drug interactions (DDIs) and the product’s use in specific populations. It also allows data to be fully leveraged from one phase to the next (both backwards and forwards), from one indication to the next, and from drug development to clinical care. Its use is actively encouraged by global regulators.

Of course M&S technologies come with their own language:

• Pharmacometrics Modeling – Population PK, exposure-response and disease-state modeling are used to predict clinical outcomes; plan clinical trials, including sample size, study duration, and dosing); provide support for dose recommendations, justification and modification; assess trends for safety and efficacy across exposure ranges; and inform “go/no-go” decisions.
• Physiologically-based Pharmacokinetic (PBPK) Technology – This mechanistic modeling informs key R&D decisions related to clinical trial design, informs first-in-human dosing, formulation design, dosing in special populations, predicts the likelihood of DDIs and drug-food interactions, and can be used for bridging studies.
• Clinical Pharmacology – Accounting for more than 50% of a drug label, clinical pharmacology approaches can help reduce late-stage attrition and increase pharma R&D productivity. Expertise in this discipline allows drug developers to reduce uncertainty related to therapeutic targets, dosing, and the patient populations in which the novel compound may have the most efficacy.
• Quantitative Systems Pharmacology (QSP) – This emerging mechanistic modeling approach is focused on target exposure, binding and expression. It is employed to identify biological pathways and disease determinants. QSP enables continuous integration of preclinical and clinical mechanistic knowledge into downstream stages of drug development.
• Quantitative Systems Toxicology (QST) – QST modeling combines toxicity and ‘omics’ data to focus on modes of action and adverse outcome pathways.
• Trial Simulation Software – This simulates the outcome of different clinical trial designs, dosing regimens, patient demographics, and inclusion/exclusion criteria to optimize the probability of a successful trial.
• Model-based Meta-analysis (MBMA) – Proprietary curated databases of publicly-available trial information are used to develop models that compare a drug’s effectiveness against competitor products, optimize clinical trials, scale from biomarker to endpoint, and inform marketing decisions.

To quote Janet Woodcock, director of FDA’s Center for Drug Evaluation and Research, “Modeling and simulation (M&S) tools for drug exposure and its response have been useful in both pre- and post-market settings when questions related to safety and efficacy of therapeutic products arise. Some recent examples where M&S has served as a useful predictive tool include dose selection for pivotal trials, dosing in select populations such as pediatrics, optimization of dose and dosing regimen in a subset patient population, prediction of efficacy and dosing in an unstudied patient population in clinical trials, characterizing exposure and dose-related QT interval prolongation, and using physiologically based pharmacokinetic modeling in predicting drug–drug interactions.”

M&S is used to guide clinical decision-making, support trials, reduce trial size, and even eliminate the need for certain trials. It has been proven to reduce costs and shorten timelines, while optimizing key decision-making. It can address the key reasons for failure: choosing the wrong endpoint, dose, patient cohort or target. However, there is a wide disparity in understanding among members of the drug development community around best practices for leveraging M&S technology, the types of technologies available, and how to communicate the results. The regulators get it—let’s help educate the entire industry!

Ellen Leinfuss  is a senior vice president at Certara, the leading provider of decision support technology and consulting services for optimizing drug development and improving health outcomes. She can be reached at Ellen.Leinfuss@certara.com.