Lifecycle modeling and simulation: poised to transform drug development

January 19, 2012 – 9:48 am by Chris

(editor’s note: this is a special guest post from Quintiles)

By Michael O’Kelly, Senior Biostatistics Director, Center for Statistics in Drug Development, Innovation, Quintiles

To build their 1903 flyer, the Wright Brothers modeled and tested wing shapes in a wind tunnel to reduce costs. A century later, 800,000 simulation hours on the Cray Supercomputer fast-forwarded the design of Boeing’s 787 Dreamliner. Computer-based modeling and simulation reduced the number of prototype wings tested, from 77 for the earlier Boeing 767, to only 11 for the Dreamliner, the most advanced aircraft of its kind.

Computer-based modeling and simulation (M&S) has transformed industries from meteorology to finance. IBM’s Deep Thunder simulation predicts potential hurricane damage to city and business infrastructure in “personalized forecasts.� Financial models and simulations show investors tradeoffs between risks and returns and enable instant portfolio updates on sites such as morningstar.com and finance.yahoo.com.

With the cost of drug development at $1.3 billion per approval and success rates languishing near 8% no industry needs the power to test assumptions and predict outcomes more than biopharma. M&S-assisted drug development has been a dream for decades, but realization has been limited by three difficulties: the enormous complexity of biological systems; the lack of population-level data on real-world health outcomes; and uncertainty about regulatory acceptance of M&S applications in drug evaluation. New technology, data and experience are beginning to overcome these challenges.

M&S is poised to transform drug development by generating predictions that advance planning and decision making across the innovation lifecycle, from discovery to marketing to safe, effective drug use in medical practice. Here’s a glimpse of that future…

• A multiple sclerosis genetics study recently published in Nature used M&S to identify 29 new genetic variants linked to MS, providing a potential guide for drug target identification.

• During the 2008 influenza epidemic, FDA used M&S to identify and approve a safe pediatric dose of the experimental drug peramivir, which had never been studied in children. Results from later pediatric trials confirmed FDA’s simulated outcomes

• The American Diabetes Association worked with Archimedes Inc. to simulate a 30-year clinical trial in a modeled patient population to compare the effectiveness of current diabetes management approaches.

M&S is still in its infancy in the health sciences. But applications like these suggest its power to expand knowledge and improve efficiencies on two great frontiers of 21^st century medicine—leveraging deeper knowledge to understand interactions at the molecular level, and leveraging broader knowledge to understand health outcomes at the population level.

Computational power is growing at an astonishing rate, enabling models that integrate huge volumes of molecular and genetic information with large-population clinical datasets. Systems biologists are constructing models so complex that they are being built by consortiums for open use in “pre-competitive� research. One is the $6.7 million partnership between Sage Bionetworks and the National Cancer Institute to build models to predict breast, colon, liver and pancreatic cancer. Mountains of population-level health data are accruing in databases such as Kaiser Permanente’s EHR database of more than 28 million members spanning 40 years. These data give rise to the possibility of more and more accurate and reliable models of health outcomes.

The most mature clinical research application, PK/PD dose modeling, has been streamlining trials for a decade. M&S uses preclinical and Phase 1 PK/PD data to predict optimal doses for clinical testing, gain insight into potential impacts of dose on efficacy endpoints, and test assumptions about biomarker impacts on clinical outcomes. Population PK data are used to predict dose adjustments for special populations including children and elders.

At FDA, the number of new drug applications incorporating M&S has increased six-fold in 10 years. The 2009 guidance on end of Phase IIa meetings encourages M&S in dose selection and trial design, and FDA uses M&S to extend clinical findings and guide additional testing. Agency initiatives include DILI-sim, an agency-industry collaboration to build open-use models to predict species differences in liver toxicity.

At Quintiles, sponsor requests for M&S are increasing. Current projects include development of numerous population PK/PD models; simulating risks and benefits of various design choices to evaluate treatments for neurodegenerative disease; and modeling disease progression in rheumatoid arthritis to inform design of future programs. M&S is also being used to improve portfolio management by addressing greater levels of real-world uncertainty.

M&S is expanding rapidly as applications begin to break down silos and bring multiple disciplines together to apply data and test assumptions. Given predictions that full M&S realization could reduce development costs by 50%, it will soon be a “must have� tool across the development lifecycle.