The Economics of Drug Discovery and Development are Not What You Think

Introduction

Simply stated, without a dramatic increase in R&D productivity, today’s pharmaceutical industry cannot sustain sufficient innovation to replace the loss of revenue due to patent expirations for successful products (Nature Reviews, March 2010). In addressing this crisis, next-generation approaches to clinical development have received disproportionate attention within industry, academia and regulatory agencies (1). At Evince, we are committed to substantially reducing the cost by increasing productivity in drug discovery via an artificial intelligence based in silico platform. A careful analysis reveals that the capitalized cost of discovery and preclinical development is 46% of total R&D costs, compared with 54% for clinical development (Nature Reviews, March 2010). It is counterintuitive that the cost of discovery and preclinical development should be such a large percentage of the total cost. The reality is that the high rate of project failure and the substantial time between project initiation and drug approval amplify tremendously the cost of early-stage activities. Improving productivity at the discovery stage can therefore make a huge difference in the overall cost of bringing a drug to market. This is sufficiently important that we have chosen to use two blog entries to discuss this topic:

  • Results of modeling drug discovery and development costs for Big Pharma / Big Biotech
  • A new model developed at Evince specifically for companies prosecuting small, targeted portfolios, such as those of small pharma / small biotech or managed within an individual business unit in a larger organization.

Modeling drug discovery and developments costs for Big Pharma / Biotech

Although the out-of-pocket cost of discovery for any single project relative to clinical cost is not all that much (estimated at slightly over 7% based on the model below), the total number of projects needed at the discovery stage to produce a new drug is what drives the cost associated with these early stages. Think of looking for oil in the 1920’s. There are many candidate sites for new wells and maybe 19 dry holes drilled for every producing well — lots of early-stage costs to achieve success. Here the analogy fails. When you drill a well you know immediately whether you’ve got a gusher or not. From its beginning as a discovery project, a new drug takes roughly 14 years to reach the market, and failure can happen at any point along the way.

For many years, analysts at the well-respected Tufts CSDD (Center for the Study of Drug Development) have been conducting and updating quantitative analyses of the interplay between drug discovery and development costs, failure rates by stage of the process, and the troubling paucity of new drugs coming to market. There is also a fine and highly relatable article from employees at Eli Lilly that analyzes cost and efficiencies: How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Paul et al. (2010) Nature Reviews Drug Discovery 9:203. The following table summarizes the key elements of the model that they use for their analysis by stage of the drug discovery and development process.

Note: cost of capital is assumed to be 11% annually.

We have used this model as the basis of our analysis and demonstration that focusing on drug discovery — in particular, decreasing the cost and increasing efficiency of lead optimization — really does substantially impact the bottom line. The total capitalized cost of Discovery and Development to bring one new drug to market is $1.8 billion. Discovery accounts for 46% of this cost (the Tufts CSDD analysis estimates 43%) and Lead Optimization alone accounts for 23% ($414 million) of the total (2).

We believe that in silico discovery approaches that use artificial intelligence to process input data will increasingly improve the efficiency of drug discovery. Small molecule Lead Optimization is a highly iterative process involving many cycles of design, synthesis, and testing in biological systems. A smart in silico approach that applies artificial intelligence techniques evaluates the proposed chemical changes to a molecule to prioritize the best prospects. The prospects that score highly are synthesized. The result is that fewer cycles are necessary to conduct lead optimization, which dramatically decreases the cost and time required to successfully complete this stage. The same principles are entirely applicable to the Hit-to-Lead stage. Decreasing the duration of these two phases (Hit-to-lead and Lead optimization) by 50% and the out of pocket cost of each of the two stage by 60% produces a net overall savings of nearly $400M ($1398M vs $1778M). The effect of cost by stage for the standard model and the test case are shown in the following table and figure:

Note: Cells with changed entries are highlighted in green; changes in output values are shown in brown text.

In addition to the direct impact of these increases in efficiency on Hit to Lead and Lead Optimization, the decreased overall time to discover a new product results in indirect savings in capitalized cost for the first three discovery stages.

We have included the above model on the Evince Website so the user can manipulate the model to determine how changes affect the overall cost.

Conclusions

The economic challenges facing the pharmaceutical industry are manifold. One of the more important is the overall cost of bringing new medicines to market. The ongoing biological revolution, including sequencing the human genome, has not yet led to fundamental changes in the way we create new medicines or in the probability of success at any stage in this difficult, drawn-out process. There has been a wave of new companies that use machine learning / artificial intelligence to drive decisions in discovery and development. Using these approaches to generate even moderate improvements in efficiencies during the discovery stages has a large impact on overall cost.

(1) There are many worth-while efforts to improve efficiency of clinical development and provide participants with greater benefit. For example, biomarker or genomics-driven patient selection and subsequent evaluation of linkage to treatment response is de rigueur for current cancer therapy trials. In this author’s opinion, perhaps the most interesting and encouraging innovations in clinical development are the use of adaptive trials and bucket trials — as well as buy-in by regulatory agencies like the FDA that these innovations should be embraced.

(2) It should be noted that this model understates costs in three ways without necessarily affecting key findings like proportion of spend associated with Discovery. First, the model was published in 2010. In 2016, the CSDD reported that costs had increased substantially (out-of-pocket costs, $1,395M and capitalized cost, $2,558M) per new approved drug. Second, this is a generalized model that includes combined data for both small molecules and biologics. Efficiencies are even worse and consequently costs higher for small molecule only portfolios. Third, significant costs are not included — for example, costs of (i) target identification and validation and (ii) support activities not directly associated with discovery and development per se but that are otherwise necessary to support the R&D organization.

Contributed by James (Jim) Appleman, Ph.D., Sr. Vice-President and CSO, Evince Biosciences. Dr. Appleman has nearly 30 years of experience in building successful drug discovery and development organizations. He also has expertise in the implementation and application of novel information technologies as applied to drug discovery.