In the high consequence environment of pharmaceutical development, any assumption made earlier in the process can prove extremely costly if uncorrected once more information becomes available.

From a business perspective, it is essential to create a safe avenue for communication of concerns regarding the drug candidate’s efficacy, safety, toxicity or pharmacological function immediately as the researchers become aware of them.

Compliance, IND submissions, and business incentives
The Food and Drug Administration (FDA) requires researchers to provide specific data about new drug candidates before the drug can to proceed to human clinical trials. The IND submission package includes preclinical data on animal pharmacology and toxicology as a proof that the drug candidate is reasonably safe for human trials; manufacturing information to ensure consistent quality of the product; and clinical protocols and investigator information. The application also includes information on clinical professionals and a commitment to obtain informed consent and a review by the institutional review board (Food and Drug Administration, 2017).

The pharmacology/toxicology package contains pharmacology studies, acute, subchronic and chronic toxicity studies, special toxicity, carcinogenicity, reproduction toxicity, mutagenicity and absorption, distribution, metabolism and excretion (ADME) studies (Food and Drug Administration, 2017). This information generally comes from in vitro tests, computer modeling, and from experiments on laboratory animals. All experiments on animals in the U.S. have to conform to Part 58 of Title 21 of the Code of Federal Regulations Good Clinical Practice for Nonclinical Laboratory Studies (eCFR — Code of Federal Regulations, 2017).

From a business risk perspective, compliance and due diligence are the easy parts. The hard part is the predictive power of this information whether or not the drug candidate has the potential to show safety and efficacy in clinical trials and obtain approval. There is a universal pressure to conduct these tests in the timeliest manner practicable, in order to pass the IND submission landmark. This step typically indicates the company’s willingness to undertake a huge investment in clinical trials, without any guarantees of success. The impact of IND submission on stock price is typically neutral (Picardo, 2017). The dropout rate is especially high for new molecular entities with the potential of becoming first-in-class. The dropout rate in phase III is especially worrying.

Publication bias in pre-clinical research
The majority of preclinical research never reaches the publication stage. For commercial research, the reasons mainly include the need to protect intellectual property. Groundbreaking research in the biotech field helps to make a business case for investors. On the other hand, academic research often has no such ambitions, the publication of the results is often part of the contract or grant award, and the main obstacles to publication are negative or inconclusive results and the lack of statistical validity. Verification of claims made in pre-clinical research based on published research is difficult, if not impossible. Whilst all human clinical trials available in databases and registries (World Health Organization, 2017) and (US. National Institute of Health, 2017), no such registry exists for animal studies. According to Matosin et al. (2014), the main motivation for publication in academic circles is to have own research cited, hence the reluctance to invest time into attempts to publish negative or inconclusive results (Matosin et al., 2014). Ter Riet et al (2012) explored what are the main factors whether preclinical research is published or not, what is the extent of non-publication, and what are the consequences of the lack of publication. Academic research is more likely to be published (50%) than commercial research (10%). Main causes of non-publication were the lack of statistical significance and the inability to pass peer review (ter Riet et al., 2012).

Reproducibility of pre-clinical research
Low reproducibility of pre-clinical research is an important source of assumptions made in the early stages of preclinical research. Freedman et al. (2015) argue that 50% of all pre-clinical research is irreproducible. Research that produces no useful results exceeds US$28 billion in the United States only. The reasons include study design, methodology, materials, reference standards and reagents used, laboratory protocols, and analysis and reporting (Freedman, Cockburn and Simcoe, 2015).

Predictive power of animal research data
The predictive power of data from preclinical research is apparent from the attrition rate in clinical trials. The Tufts Center for the Study of Drug Development estimates the cost of a newly approved drug to be at about US$2.6 billion. On average, about 60% drop out in Phase I of clinical research and only 8% percent of drug candidates from the discovery phase reach the market. The attrition rate in the preclinical stage is approximately 70%. Unsurprisingly, the highest attrition rate is among first-in-class drugs (Booth, 2017).

Early failures
The lack of reliability and predictive power of data from preclinical research costs human lives as well. Whilst these events are extremely rare, their consequences are potentially profound. In 1993, a hepatitis B drug fialuridine had to be dropped from a Phase II trial because of severe hepatotoxicity, in seven patients, of whom five died and another two only survived due to a liver transplant. The accident prompted the development of more sensitive tests for the detection of liver toxicity. The cause of the deaths was a rare type of long-term toxicity undetected in preclinical trials (Manning and Swartz, 1995). The TGN-1412 First-In-Man failure became one of the most studied clinical research disasters due to the unanticipated toxicity of the drug. Not only the accident led to overhaul of the rules for Phase I trials, it also caused serious concerns about the predictive power of animal studies. The life-threatening symptoms of a cytokine storm were the result of assumptions made about the biology of the drug in preclinical studies. Research conducted on cynomolgus monkeys did not translate into the human biology of immune response (Attarwala, 2010). The most recent instance of unanticipated toxicity occurred in 2016 when serious neurotoxicity developed in healthy volunteers who received a new experimental drug BIA 10-2474, a reversible inhibitor of fatty acid amine hydrolase (FAAH) that increases the levels of endocannabinoids. Cumulative toxicity that was to blame was not detected in preclinical studies (Kerbrat et al., 2016). In response to the disaster, the European Medicines Agency (EMA) published a concept paper with suggestions how to improve the safety of First-in-Man trials (European Medicines Agency, 2017).

Phase III failures
The reasons why drugs fail in phase III boil down to a flawed basic science, inappropriate animal disease models, the use of and incomplete understanding of the disease biology and targets, or uncorrected assumptions made earlier in the process of drug discovery and development. Clinical study design is an important factor because of the use of surrogate endpoints, changes in exclusion and inclusion criteria and patient population, inappropriate dose selection in the transition from Phase II to Phase III, operational execution and overly optimistic presentation of findings, leading to “go ahead” decisions for projects that should have been stopped (Shanley, 2017).

Less discussed causes of phase III failures are of organizational origin. It is human nature to protect own work and own projects to ensure continuing participation in projects that receive the organization’s approval and funding. Halting a project is an event every project manager prefers to avoid. The reasons are obvious and understandable – from a career perspective, it is always better to participate in a single long project than in multiple projects that all failed in a short time. Bringing up concerns regarding the viability of a project and the results that are coming in will always represent a career risk to the individual who dares to present the unwelcome news, especially if they come later in the process and the news are likely to result in a major controversy. In the high consequence environment of pharmaceutical development, any assumption made earlier in the process can prove extremely costly if uncorrected once more information becomes available. From a business perspective, it is essential to create a safe avenue for communication of concerns regarding the drug candidate’s efficacy, safety, toxicity or pharmacological function immediately as the researchers become aware of them.

Any plan must proceed on assumptions. An information collection plan needs to exist to facilitate timely replacement of assumptions with facts as the situation develops. Indicators of risks and potential adverse consequences, specifically potential safety concerns and lack of efficacy, have to be recorded in the original research plan and tracked throughout the drug development process. These Priority Information Requirements (PIR) and Critical Information Requirements (CIR) need to correlate with decision-making to be meaningful as a protective mechanism against the risk of an extremely costly late-stage failure (Sheckler, 2017).

In the drug development industry, a pro-active feedback loop between pre-clinical and clinical stages of development needs to exist to facilitate continuous verification of assumptions and consequent adjustment of plans.

Emerging technologies
New technologies such as human organs-on-a-chip (Wyss Institute, 2017) have the ability to transform preclinical research fundamentally. The technology involves a dynamic 3D model of human tissue on a computerized model that allows a detailed understanding of the disease biology and the effect of drug candidates on disease targets. The number of repetitions is near endless, giving the results the necessary statistical power. It is safe to assume that the technology will become more affordable as it matures. Metabolism of xenobiotics, including pharmaceuticals, is a complicated affair that is difficult to observe directly in living creatures, animals and humans alike. Our understanding of absorption, distribution, metabolism, and excretion, and lack thereof, directly depends on the tools we have. At a theoretical level, we all know that the metabolism of xenobiotics depends on the genetic makeup of the cell, qualitative characteristics and capacity of cytochromes CYP450 to process the compound, one way or another. Yet in clinical trials, we still heavily rely on chance. New technologies allow in-vitro and in-silico modeling of scenarios that will inevitably occur when the drug reaches real patients such as hypoxic state, raised levels of inflammatory markers, and decreased liver capacity due to NAFLD / NASH or perfusion changes. Abaci and Shuler (2015) argue that the technology can be used for pharmacokinetic and pharmacodynamics modeling in drug development, and explored methods how to achieve scaling up of the organ models in order to replicate organ-organ interactions. μOrgans-on-a-chip (μOOC) can be used in preclinical research to mimic a physiological system using human cells to predict behavior or validate assumptions made in earlier stages of research. Once the PK/PD is known, μHuman-on-a-chip (μHOC) can be used to model inter-organ interactions and to model and predict drug partitioning, metabolism rate, permeability rate and so on (Abaci and Shuler, 2015). Oleaga et al. (2016) successfully utilized the system for modeling organ toxicity of new drugs under development. The model was tested on drugs with known toxicity on functional models of cardiac, muscle, liver, and neuronal tissue. The results showed promising results in regards to the predictive value of the models for in-vitro toxicity screening (Oleaga et al., 2016).

Better times ahead?
The published materials seem to be a tip of the iceberg of extensive research that explores the potential of the technology in drug discovery and development. Reduction of attrition rate in clinical trials and especially in the advanced stages requires a combination of technological solutions and organizational changes. Information without appropriate action is meaningless. To facilitate appropriate organizational response to new coming information that is likely to halt a high-value project, the perceptions, incentives, and motivators within R&D functions have to change to accommodate timely termination of a potentially costly project as a good outcome that is good for business rather than a personal or team failure.

eCFR — Code of Federal Regulations (2017). eCFR — Code of Federal Regulations: Title 21, Chapter I, Subchapter A, Part 58: Good Laboratory Practice for non-clinical laboratory studies. [online] Available at: [Accessed 1 Jun. 2017].

Food and Drug Administration (2017). Investigational New Drug (IND) Application. [online] Available at: [Accessed 2 Jun. 2017].

Food and Drug Administration (2017). Drugs: Guidance, Compliance & Regulatory Information – Guidances (Drugs) – Pharmacology / Toxicology. [online] Food and Drug Administration. Available at: [Accessed 2 Jun. 2017].

Shanley, A. (2017). Preventing Phase III Failures. [online] Available at: [Accessed 2 Jun. 2017].

World Health Organization (2017). ICTRP Search Portal. [online] Available at: [Accessed 1 Jun. 2017].

US. National Institute of Health (2017). Home – [online] Available at: [Accessed 1 Jun. 2017].

Sheckler, R. (2017). All organizational dysfunctions track directly to leadership failings. (Part 2). [online] Arete-Zoe. Available at: [Accessed 2 Jun. 2017].

Wyss Institute (2017). Human Organs-on-Chips. [online] Wyss Institute. Available at: [Accessed 2 Jun. 2017].

ter Riet, G., Korevaar, D., Leenaars, M., Sterk, P., Van Noorden, C., Bouter, L., Lutter, R., Elferink, R. and Hooft, L. (2012). Publication Bias in Laboratory Animal Research: A Survey on Magnitude, Drivers, Consequences and Potential Solutions. PLoS ONE, 7(9), p.e43404.

Matosin, N., Frank, E., Engel, M., Lum, J. and Newell, K. (2014). Negativity towards negative results: a discussion of the disconnect between scientific worth and scientific culture. Disease Models & Mechanisms, 7(2), pp.171-173.

Freedman, L., Cockburn, I. and Simcoe, T. (2015). The Economics of Reproducibility in Preclinical Research. PLOS Biology, 13(6), p.e1002165.

Booth, B. (2017). A Billion Here, A Billion There: The Cost Of Making A Drug Revisited | LifeSciVC. [online] LifeSciVC. Available at: [Accessed 1 Jun. 2017].

Manning, F. and Swartz, M. (1995). Review of the Fialuridine (FIAU) clinical trials. 1st ed. Washington, DC: National Academy Press.

Attarwala, H. (2010). TGN1412: From Discovery to Disaster. Journal of Young Pharmacists, 2(3), pp.332-336.

Kerbrat, A., Ferré, J., Fillatre, P., Ronzière, T., Vannier, S., Carsin-Nicol, B., Lavoué, S., Vérin, M., Gauvrit, J., Le Tulzo, Y. and Edan, G. (2016). Acute Neurologic Disorder from an Inhibitor of Fatty Acid Amide Hydrolase. New England Journal of Medicine, 375(18), pp.1717-1725.

European Medicines Agency (2017). European Medicines Agency – News and Events – Proposals to revise guidance on first-in-human clinical trials. [online] Available at: [Accessed 1 Jun. 2017].

Abaci, H. and Shuler, M. (2015). Human-on-a-chip design strategies and principles for physiologically based pharmacokinetics/pharmacodynamics modeling. Integr. Biol., 7(4), pp.383-391.

Oleaga, C., Bernabini, C., Smith, A., Srinivasan, B., Jackson, M., McLamb, W., Platt, V., Bridges, R., Cai, Y., Santhanam, N., Berry, B., Najjar, S., Akanda, N., Guo, X., Martin, C., Ekman, G., Esch, M., Langer, J., Ouedraogo, G., Cotovio, J., Breton, L., Shuler, M. and
Hickman, J. (2016). Multi-Organ toxicity demonstration in a functional human in vitro system composed of four organs. Scientific Reports, 6(1).