‘Fit for Purpose’: FDA recognises MCP-Mod’s utility to improve dose finding


Methodology incorporated into ICON’s ADDPLAN® DF platform for dose finding

MCP-Mod, a powerful statistical tool for reliably predicting optimal dose ranges of new drugs for future confirmatory trials, has been deemed “fit for purpose” by the U.S. FDA. The tool could reduce costly Phase III failures and post-approval dose adjustments.

Suboptimal dose finding remains a major drain on pipeline productivity. In fact, it was the most frequent reason for the FDA to reject first-time marketing applications for new molecular entities (NMEs), according to an analysis of rejections between 2000 and 2012. Rejection may require additional trials, which can erode the remaining time on patent. Of the NMEs that were subsequently resubmitted after a rejection for any reason, the median approval delay was 14.5 months, ranging up to 6.5 years.

Regulators including the FDA and EMA are accordingly encouraging industry to develop better dose-finding methods for early clinical trials.

In that drive, significant attention has been directed to the MCP-Mod approach, which is short for Multiple Comparison Procedure – Modelling. In 2014, the EMA issued a qualification opinion that found the method was more “efficient” with available data and will “promote better trial designs incorporating a wider dose range and increased number of dose levels.”

The FDA recently gave MCP-Mod its “fit-for-purpose” designation for guiding dose selection for Phase III testing. Only one other tool had previously received this designation.

Significantly, the FDA convened a multidisciplinary team, from both the Office of Biostatistics and Office of Clinical Pharmacology at CDER, to review the MCP-Mod methodology based on a submission from Janssen and Novartis. The FDA noted in its determination letter that “the methodology is scientifically sound” and “advantageous in that it considers model uncertainty and is efficient in the use of the available data compared to traditional pairwise comparisons.”

With this in mind, MCP-Mod has been incorporated into ICON’s ADDPLAN® DF, which supports the decision process in the design and analysis of dose finding trials using innovative methodologies. The addition of MCP-Mod was tested with the benefit of the input and expertise of the ADDPLAN DF Consortium, made up of statisticians from nine major pharmaceutical companies. 

Hazards of Current Dose Selection Approaches

Dose-related problems in late stage trials often stem from the limitations of traditional dose finding methods. Those methods generally fall into two categories: pairwise or multiple comparison procedures and model-based approaches. Each has strengths and weaknesses.

Multiple comparison procedures examine several dose levels, quite often within a narrow dose range focused on the upper end of the presumed dose response relationship. This is robust in assessing response differences among the tested results. No assumption, however, is made about the underlying dose-response relationship, so one cannot reliably predict responses to doses beyond those tested.

Model-based approaches examine several dose levels and then fit a pre-selected dose-response curve to the response data. This enables predictions of responses to doses not tested, depending on where they fall on the presumed dose-response curve. But if the selected model is wrong, so are its predictions.

The hazards of restricting Phase III dosing to a narrow range of doses along an unknown response curve are clearly illustrated by a set of trials conducted by a large manufacturer in the early 1990s, as reported in Nature Reviews Drug Discovery.

In this example, a Phase III trial tested three doses ranging from 80 to 160 mg and detected no efficacy difference. A second three-dose trial ranging from 40 to 120 mg also found no difference. It took a third Phase III trial examining three doses from 2.5 to 40 mg against a placebo to find the true dose response curve needed for approval. This miscalculation, based on pairwise comparisons of early trial data, cost the sponsor three years and tens of millions of dollars.

Advantages of MCP-Mod

MCP-Mod addresses these limitations using a hybrid approach. Like multiple comparison procedures, MCP-Mod collects and compares data from several dose points. It also incorporates multiple preselected dose-response curve hypotheses, giving it great predictive power while reducing the risk of choosing a completely wrong model. MCP-Mod works best when the dose levels are widely spaced, by an order of magnitude or more, and a placebo is included.

The first step of the procedure (MCP-step) assesses the presence of a dose-response signal using a trend test deduced from the pre-specified candidate models. The second step (Mod-step) relies on parametric modelling to choose the best fit among the preselected models or model averaging to estimate an optimal dose range for confirmatory trials.

In its detailed statistical and pharmacometric reviews of MCD-Mod, the FDA found the methodology facilitates more informative Phase II trial design by encouraging:

•      testing three or more active and well-separated doses

•      investigating doses on the ascending part of the dose-response curve

•      interpolation within data points and extrapolation beyond data points to select dose(s) for pivotal trial(s)

The agency also encouraged developing MCP-Mod beyond its current applications in univariate response studies to include multivariate responses for safety and efficacy, and exposure response.

Practical Implementation
One potential challenge of MCP-Mod is that it requires at least three active doses plus placebo, demanding a larger Phase II trial than is typical today. Currently, most trials test only one or two doses.

However, combining the approach with an adaptive dose finding design can substantially reduce sample size requirements and trim costs. (For a strategic guide to adaptive dose finding trials, read this white paper.) Furthermore, Phase II MCP-Mod data may be used in some cases as a comparator in Phase III, reducing the need for concurrent confirmatory Phase III trials, which may dramatically reduce costs.

ICON updated its ADDPLAN DF platform with functionality to combine MCP-Mod and adaptive design, giving study designers an efficient, precise tool to answer questions on the most informative doses for the study, while adding insight into the underlying dose-response relationship. As more companies switch on to the multiple benefits of adaptive trials and innovative dose finding methodologies, a better understanding of the dose-response relationship should ultimately play a key role in reducing late phase attrition and curtailing the duration and expense of clinical development.

More information
More Information about improving dose-finding designs for your trials: