Factors that Influence Attrition Rates

A number of factors influence the attrition rates described in the previous section.  Kola and Landis found that attrition rates vary by stage of development and therapy area. 1 And within each stage of development there is variability amongst therapy areas.  A recent study found a higher overall success rate for New Biological Entities than for New Chemical Entities 2 Kola and Landis also cite a CMR study that show in-licensed compounds tend to do better in the clinic and that small companies tend to show better success rates than larger companies (Ref  1).  These same factors were also shown to be important by DiMasi et al. in 2010. 3.

Assay/Model Predictivity.  Of course, at specific stages of Discovery and Development tests are performed that measure a) the relevance of the biological target to the potential disease indication, b) the efficacy of the candidate drug in a variety of in vitro and in vivo disease models, c) the ability of the candidate drug to reach the site of action in the body in sufficient concentration to elicit an effect, d) the potential for safety issues with the candidate drug, e) which patients are likely to respond to the drug, and f) early signals of efficacy of the drug in patients.  If all such tests were completely predictive of outcome in human patients one would expect that the probability of success in Phase II would be considerably higher than 40%, and in Phase III higher than 60%. 

Bjornsson 4 notes that total predictivity, PT is the product of the individual key predictivities, of efficacy, safety and compound properties (and we would add biomarker predictivity), or

PT = PE X PS X PC X PB

Total predictivity PT is related to total attrition AT as

PT = (1 – AT )

Each of these individual predictivities, Pi is related to the proportion of that individual attrition ai, relative to total attrition, or

Pi = [1 – (ai X AT)] 

Which is to say that every step in the R&D process has an impact on the total predictivity, and an important activity is to determine the contribution of the individual attrition ai to total attrition, presumably some activities will have a greater influence on attrition than others.

Predictivity of Preclinical Efficacy Models. Bjornsson has provided a considered study of the predictivity of preclinical efficacy models.  Preclinical efficacy models provide the following data: a) the intended indication, b) the biological targets and pathways influenced by the drug, c) potency and selectivity, d) efficacy in animal pharmacological or disease models in a relevant species, e) the onset (immediate/delayed) and time course (short/long) of the effect, f) the dose/concentration/exposure required to produce the effect, f) the biomarkers that demonstrate efficacy and g) the imaging methodologies required to observe the effect.  The efficacy models have a wide range of predictivity from reasonably high (blood pressure) to very low (Alzheimer’s) varying by target and disease.

Predictivity of Safety Models. Preclinical safety models provide the following data: a) in silico predictions of toxicity issues early on, b) demonstrations of target organ toxicity, c) demonstrations of toxicity outside the target organ, d) indications whether these effects occur upon acute or chronic exposure to the drug, e) determination of the minimal dose to manifest toxicities – the NOEL, no observable effect limit, and the NOAEL, no observable adverse effect limit, and f) the potential for reproductive toxicity and genotoxicity as well as carcinogenicity. These are a variety of in vitro and in vivo models mandated by regulatory authorities. 5 

Predictivity of Compound Properties. There are a variety of properties related to the delivery of the drug to its site of action and the interactions of the organs of the body with the drug that may reduce its efficacy. A number of in vitro and in vivo models have been developed to characterize drug disposition and exposure.  These models provide information on a) pharmacokinetics (PK) in animals, b) drug degradation (drug metabolism, DM) and the resulting metabolites, c) the impact of other drugs on the DM/PK of the drug (drug-drug interactions), d) absorption and bioavailability (the percent of drug administered versus the amount found in the blood stream), and e) physicochemical properties, such as salt form, solubility, pKa, lipophilicity.  In recent years considerable advances have been made with physiologically based pharmacokinetic predictions with software that show a predictivity averaging about 0.6 (Ref 4).  Medicinal chemists have developed a short set of desired properties in a drug candidate, the Lipinski Rule of Five – not more than 5 hydrogen bond donors (nitrogen or oxygen atoms with one or more hydrogen atoms), not more than 10 hydrogen bond acceptors (nitrogen or oxygen atoms), a molecular mass less than 500 daltons, and an octanol-water partition coefficient log P not greater than 5. 6  The list has been expanded to involve the concept of “druggability”.

Predictivity of Biomarkers.  The lynch pin of personalized medicine and translation medicine, the ability of a biomarker to predict which patients will respond to a drug is an extremely important predictivity. 7 The success of a clinical trial will be enhanced by excluding that portion of the patient population that is unlikely to respond to the drug.  Identified in early Target Discovery activities, the biomarkers first help pick relevant disease targets, then help improve the relevance of preclinical efficacy models in identifying the best clinical candidates, finally determining which patients should receive the drug and possibly providing an early indication of efficacy in a clinical trial.  A simple biomarker would be blood pressure or heart rate, which may be sufficient for some cardiovascular indications.  Many more diseases require considerably more sophisticated biomarkers, usually the up- or down-regulation of specific disease markers.  Best if the biomarkers are detectable from blood or urine samples.

Where to Start.  Attrition is influenced by so many factors perhaps the biggest challenge is to determine where to start.

“If we knew what we were doing it wouldn’t be called Research” – Albert Einstein.

A logical approach to attrition would be to determine which factors are likely to have the most impact on attrition and start with them.  In 2007 the PHRMA determined the spend on R&D by stage, Table 1. 8  In their analysis, “preclinical stages” includes the Discovery Stages and the Preclinical Development stage, (the numbers may be under-quoted since only resource is projectized in the Discovery stages).  What may be a surprise to some is that the spend on large Phase 3 clinical trials is almost as much as the spend on all “preclinical stages”, Table 1. 

 

Table 1, Success Rates {(a) described in Figure 1 of Attrition)} and Total Spend per Stage {(b)Ref 8)} per Stage ($ millions, for all R&D by PHRMA members in 2007).

If we assume that the spend is roughly similar in each Preclinical stage, then we get $3272 per Preclinical stage, comparable to Phase 1. The facts that Phase 2 has the lowest success rate, 44%, and that Phase 3 has the highest cost, $13.665 million suggest that attrition in these stages needs to be addressed first.

This discussion continues in the section Attrition – Reasons for Failure

  1. “Can the pharmaceutical industry reduce attrition rates?” I. Khola and J. Landis, Nature Reviews / Drug Discovery 3, 711, 2004
  2. BIO CEO & Investor Conference February 15th, 2011BIO
  3. DiMasi, J.A., Feldman, L., Seckler, A., and Wilson, A., “Trends in Risks Associated With New Drug Development – Success Rates for Investigational Drugs”, Clin. Pharm. Ther., 87, 272-277 (2010)
  4. “Does Pharmaceutical Predictivity Translate to Productivity in Drug Development? If So, How?”, Thorir Bjornsson, May 2010, presentation to NGP Summit in Los Angeles in 2009, and at FDA in Silver Spring, Maryland, and Critical Path Institute in Tucson, Arizona, in 2010, http://dl.dropbox.com/u/6886618/Pharmaceutical%20Predictivity%20May-2010.ppt.pdf
  5. Olson H, Betton G, Robinson D, Thomas K, Monro A, Kolaja G, Lilly P, Sanders J, Sipes G, Bracken W, Dorato M, Van Deun K, Smith P, Berger B, Heller, “Concordance of the toxicity of pharmaceuticals in humans and in animals”, Regul. Toxicol. Pharmacol. 32, p56-67 (2000)
  6. Leeson, P., Springthorpe, B., “The influence of drug-like concepts on decision-making in medicinal chemistry, Nat. Rev. Drug Disc. 6, p881-890 (2007)
  7. Please refer to the section Translational Research, Translational Medicine, Biomarkers and Personalized Medicine.
  8. Table 5, PHRMA, “Pharmaceutical Industry Profile 2009