A Paradoxical Trend in Drug Development

Research and development (R&D) for drug development is an integral part of society contributing to increased life expectancy and higher quality of life. Over the last 70 years, trillions of cumulative dollars have been spent by both industry and government funded research leading to revolutionary discoveries that have formed new fields and informed novel modalities of drug development. Given these advancements, it would be expected that the overall efficiency of drug development has improved. Paradoxically, there has been an exponential decrease in efficiency. Today, the development of a new drug is estimated to cost between $1.8 and $3.0 billion, takes over a decade to reach market approval, and has success rates estimated as low as 1 in 10,000 [1]. What are the origins of these effects? How can R&D efficiency be improved? To address these questions, this post aims to formalize the drug development process, introduce the declining R&D efficiency problem, discuss the chilling implications of the current trend, and suggest some future lines of inquiry to solve this problem.

Developing a new drug requires multiple steps and is formally summarized in Figure 1. Although this summary is partially agnostic to serendipitous discovery, this is commonly how the industry and academics approach the problem. A potential target for an ailment such a receptor or enzyme must be identified and validated requiring basic science methods (green boxes, Figure 1). This is then followed by drug development to the target by screening methods and/or rational design coupled with hit identification. Once a drug hit has been identified, this is subjected to lead optimization to obtain an ideal in vitro response which is verified in vivo with animal studies (blue boxes, Figure 1). Only then does a drug reach the latter portions of Figure 1 where clinical trials are conducted (purple boxes, Figure 1) in three phases (I, II, and III) with a regulatory agency such as the Food and Drug Administration (FDA) or the European Medicines Agency (EMA). Only upon successfully navigating the entirety of this landscape does a drug come to market where it can treat someone suffering from the ailment.

Figure 1 — The Drug Development Process. Shown from left to right is the sequence of drug development including Basic Science (green boxes), Preclinical Development (blue boxes), and Clinical Development (purple boxes). Beneath each major stage are the specific steps, the cycle time (in years), the probability of transitioning to the next step, and the cost per launch (fully capitalized in Millions [M] of US Dollars). Note that the Cost per launch (capitalized) includes the total costs per approval which includes the costs of failed drugs. The data are taken from [2] and the figure is also adopted from [3] and [4].

It is important to cast Figure 1 within in the context of scientific development from the last 70 years. There have been revolutionary advancements that have improved the efficiency of each one of these steps. These include the rise of model organisms to study disease [5], the molecular biology revolution allowing routine cloning and expression of gene products [6], the human genome project which has now led to the age of genomics [7] and other –omics, combinatorial chemistry which has facilitated an increase in the number of drugs that can be developed [8], and even recent advancements in gene therapy where we have seen glimpses of the age of personalized medicine [9]. The magnitude of these advancements cannot be overstated. Multiple Nobel Prizes have been awarded and entire new avenues of drug development are accessible today that were not present at the turn of the last century. Taken together, these results paint a hopeful and promising picture of R&D efficiency.

Given the apparent improvement in each of the steps in Figure 1, how can we measure the overall R&D efficiency today and compare it to the past? One simple measure of R&D efficiency is the number of new drugs approved per billion dollars of R&D spend adjusted for inflation. Given the advancements discussed in the previous paragraph, we would expect to see an overall improvement in R&D efficiency. Paradoxically, the opposite conclusion is borne out of the analysis. Not only has R&D efficiency decreased over the last 70 years, it is decreasing at an exponential rate as depicted in Figure 2. This is commonly referred to as erooM’s law as an analog to the more commonly recognized Moore’s law [10]. Specifically, the number of drugs approved has remained approximately constant over the time horizon of Figure 2 while R&D spend has increased to well over $60 billion from industry alone. This not only stands at odds with the scientific advancements of the last century, but questions the effectiveness of the entire drug development process.

Figure 2 — The Exponentially Decreasing Efficiency of R&D Development. Plotted are the number of drugs approved per inflation adjusted billions of US Dollars versus year as adopted from [3]. Primary data were manually added for years 2011 through 2015 utilizing approval numbers from [11] and R&D spend from [12]. Two items are of note: (1) There is a strong linear trend when plotted on a logarithmic scale indicating an exponentially decreasing efficiency of R&D spend, and (2) the last couple years have seen a slight attenuation of this trend given a higher number of biologic drug approvals. Whether the trend has truly attenuated is inconclusive from these data.

The implications of erooM’s law are chilling. Put in financial terms, as the efficiency of R&D decreases the return on invested capital (ROIC) [13] does as well. In financial theory, businesses will only pursue a project if it meets the Net Present Value (NPV) criteria: i.e. the NPV must be positive presenting a value creation opportunity to the company since the benefits (drug approval) are larger than the costs (R&D spend). ROIC is integral to the NPV criteria as this determines the potential benefits in the future versus the costs. Given the declining ROIC in R&D implies that the number of viable drug development projects is decreasing exponentially [14] unless drug prices increase or there is an improvement in R&D efficiency. Pricing is already a hot topic issue and even though it seems that erooM’s law has not significantly influenced pricing [15], it seems unlikely that a solution will come from increasing drug costs. Thus, pending an improvement of R&D efficiency, the high-level implication from Figure 2 is that the global drug pipeline will attenuate and become depleted of new projects — i.e. there will be fewer and fewer new drugs developed despite the well-recognized ailments that plague society.

To alleviate the R&D efficiency problem the obvious questions are: what are the origins of this effect and how can we address them? As previously described [3], it seems that there are four primary factors driving erooM’s law:

  1. “Better than the Beatles” problem — the origin of this name is that any new hit in the music industry must be better than the Beatles. This effect manifests itself that the Blockbuster drug of the previous product cycle is today’s generic drug. For instance, a previous hit drug such as Viagra has now lost exclusivity and the market is flooded with generics. Thus, the hurdle to approve a new drug in a new addressable market has necessarily become more difficult as certain markets have become saturated.
  2. The cautious regulator — This refers to the change in risk tolerance of the drug regulators. As certain ailments have necessarily become obsolete, the risk that regulators are willing to tolerate has decreased changing the cost-benefit analysis required to obtain approval.
  3. “Throw money at it” tendency — This refers to the inflated sizes of R&D departments and other investments (such as advertising and human resources) that have inflated over the last 70 years. This may also be conflated with increased competition in certain markets.
  4. “Basic research-brute force” bias — There seems to always be an overestimation of the advancements promised by basic sciences from each new discovery. This can be appreciated by the advancements described in previous paragraphs which have been part of the overall decline in R&D efficiency.

Although not exhaustive and mutually exclusive, these four phenomena are at the heart of the decline in R&D efficiency.

We can further assess how these four effects quantifiably affect R&D efficiency by using the model from Figure 1. A previously proposed model is summarized in the bottom of Figure 1 showing the approximate time, probability of transition to the next step, and total capitalized cost per drug developed [2]. A few important conclusions can be drawn from this analysis. First, approximately 50% of the capitalized costs and 75% of failures occur before clinical trials. This emphasizes the importance of basic research and diverse targets. Second, Lead optimization is the most expensive step despite its high probability of transition implying a significant number of drugs show excellent in vitro and in vivo effectiveness yet fail in clinical trials. Third, two of the three highest costs per new drug are Phase II and Phase III. This reveals the importance of constructing high quality clinical trials to minimize attrition in the more capital intensive later stages of development. Finally, and most importantly, the average drug costs approximately $1.8 billion dollars and takes almost 13 years to develop [16]. Taken together, these results clearly demonstrate that there is no single bottleneck and that the costs can be reduced at almost every step of the process.

How can we specifically address these identified roadblocks? The full answers are not within the scope of this blog post but clear solutions immediately come to mind and two are proposed. First, any improvements to Phase II and Phase III clinical trial design that decrease attrition would clearly increase overall efficiency by as much as $1 billion per drug [2]. Some previously suggested solutions include the utilization of patient data to create smarter clinical trials at enrollment. For instance, a novel obesity drug may have synergistic effects if the patient has a more active lifestyle which can be assessed by utilizing mobile phone data.

Second, more diverse target pursuance or collaboration on similar targets between companies can increase overall efficiency as well. This inefficiency is manifested by the largest drug companies all pursuing similar addressable markets. For example, almost every large drug company is currently pursuing monoclonal antibodies to address inflammation in Rheumatoid Arthritis, Ulcerative Colitis, or Chron’s Disease. As a result, the overall R&D efficiency decreases as many companies directly compete for the same market and spend duplicative R&D dollars on experiments or information that could be more efficiently spent elsewhere. Unfortunately, how collaboration is facilitated efficiently to mitigate duplicative R&D spend without incurring patent disputes and proper recognition of contribution is an open question.

Regardless of the suggested solutions, one conclusion from our analyses is clear: there is a need for resolutions to address the declining efficiency in drug R&D. The roadblocks discussed here are well recognized and simple to identify. The fundamental issue is proposing specific solutions that not only address the overall industry efficiency, but incentivize adoption from individual research groups or companies. What specific solutions conform to this bicameral incentive structure? Are these solutions that can be adopted by an individual organization or do they require industry penetration? These are simple questions with necessarily complicated answers waiting to be found.

Footnotes and Citations:

[1] These numbers are dependent on which source is specifically cited, however it is widely accepted that each drug costs more than a billion dollars, takes a decade to develop, and including basic science steps has attrition rates as described.

[2] Paul, S.M., et al., How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nature Drug Discovery, 2010. 9: p. 203–214.

[3] Scannell, J.W., et al., Diagnosing the decline in pharmaceutical R&D efficiency. Nature Drug Discovery, 2012. 11: p. 191–200.

[4] Tollman, P., et al., Identifying R&D outliers. Nature Drug Discovery, 2011. 10: p. 653–654.

[5] Model organisms and disease model systems play an integral role in preclinical studies and as genetic screening tools.

[6] In the 1970’s to the 1980’s recombinant DNA technology came to fruition allowing a myriad of other techniques to come to fruition including but not limited to: recombinant protein expression, engineering of bacteria for directed evolution, and construction of genetic material for modifying model organisms.

[7] The human genome project and the advent of high-throughput sequencing have led to staggering databases full of sequenced human genomes allowing identification of genetic markers for disease. Additionally, the explosion of the “omics” era including proteomics has allowed for identification of a significant number of drug targets.

[8] Combinatorial chemistry has greatly facilitated improvements to the lead optimization process by allowing a chemist to expand the universe of potential drug targets that can be constructed. It is estimated that this has improved the chemical search space by approximately two orders of magnitude.

[9] New tools such as CRISPR Cas9 and Chimeric Antigen Receptor T-Cell (CAR-T) technologies are just beginning to become approved by the FDA.

[10] erooM is Moore spelled backwards and was first described in [3].

[11] FDA, Novel Drugs 2015 Summary. 2015, Food and Drug Administration. p. 1–17.

[12] PhRMA, 2016 PhRMA Annual Membership Survey. 2016, PhRMA. p. 1–9.

[13] Return on Invested Capital (ROIC) is commonly used to measure the “bang for your buck” in businesses. It is measured in percent and the larger the percentage, the better the returns. Figure 2 shows that ROIC has decreased in drug R&D where return is the number of drugs approved while invested capital is the dollars spent.

[14] Under the assumption of no pricing increases and a finite number of drug targets, as ROIC decreases, so does the number of NPV positive projects. Thus, the number of financially viable drug research projects decreases suggesting a globally depleted pipeline.

[15] Scannell, J.W., Four Reasons Drugs Are Expensive, Of Which Two Are False. Innogen Working Papers, 2015. 114: p. 1–22.

[16] These values vary depending on source cited, however, our estimates are grounded from the citations provided in this post.