COMMENTARY
Advancing Drug Discovery: A Pharmaceutics Perspective ELIZABETH KWONG Kwong Eureka Solutions, Quebec, Canada Received 9 July 2014; revised 25 September 2014; accepted 4 November 2014 Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24294 ABSTRACT: Current industry perspective of how discovery is conducted seems to be fragmented and does not have a unified overall outlook of how discovery challenges are being addressed. Consequently, well-defined processes and drug-likeness criteria are being viewed as “broken” and will not maintain future R&D productivity. In this commentary, an analysis of existing practices for defining successful development candidates resulted in a 5 “must do” list to help advance Drug Discovery as presented from a Pharmaceutics perspective. The 5 “must do” list includes: what an ideal discovery team model should look like, what criteria should be considered for the desired development candidate profile, what the building blocks of the development candidate should look like, and how to assess the development C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci risks of the candidate. Keywords: drug-like properties; operating models in discovery; physicochemical properties; Drug Design; ADMET properties; insilico modeling; in-vitro models;target therapeutic profile (TPP); developability risk assessment; MAD; rule of five
INTRODUCTION Despite today’s challenging environment, pharmaceutical industry players continue to strive to deliver scientific excellence, meet unmet medical needs, and drive state-of-the-art innovation. At the same time, the industry and its stakeholders are cognizant that the current model is not sustainable for long term. To fix this situation, one of the analyses on R&D productivity1 has pointed toward the reduction of costs and cycle time. At the same time, shifting compound attrition to earlier during lead optimization before the first-in-human stage of clinical development will be a part of the cost-saving features of this enhanced productivity. In other words, the “preclinical stage” investment is now becoming the new “Phase IIb proofof-concept study” for key “go/no-go” decisions. By their estimation, this paradigm shift will increase the overall probability of technical success in late-stage phases II and III. This analysis therefore suggests a refocus of resources to discovery research and early translational medicine. Finally, for any successful R&D, there is no substitute for good people and good science. Investment in talent and getting the right people at the bench and in decision-making roles is critical.1 Since the publication from Paul et al.,1 there has been a longrange planning in the pharmaceutical industry to focus on process efficiencies in the preclinical stage. This has been achieved by optimizing decision-making in discovery research, which includes faster “go/no-go” decisions about progressing programs into clinical development. In addition, improvements have been implemented in research capabilities based on the newest scientific insights and the optimal use of resources to balance multiple projects, cost, priorities, and productivity. It will be increasingly important for scientists to not only think from the left or right side of their brain but rather thinking using one brain. As a consequence, a conscious effort to re-organize discovery groups evolved, which included the integration of the Correspondence to: Elizabeth Kwong (Telephone: +514-694-6475; E-mail: ec [email protected]) Journal of Pharmaceutical Sciences C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association
preclinical/development functions such as drug metabolism and pharmacokinetics, pharmaceutics, process chemistry research, and toxicology into drug discovery,2,3 in hopes of improving the efficiency of generating candidates for the development. At the same time, this shift of resources from early development to the discovery space allowed integrated teams such as Developability Assessment Groups4 to help drive the optimization of well-balanced “drug-like” properties of the candidate. Continued improvement on this holistic approach and reorganization of discovery team resulted in identifying liabilities early, and enabled more rationale candidate selection decisions. It was also suggested that drug delivery studies, drug–drug interaction assessment, and safety pharmacology assays can be used to provide preclinical information necessary to select a drug candidate with the best overall pharmaceutical profile. Higher sample throughput screens, better in vitro cell models, and computational models have been employed to manage resources, costs, and time.3 It is also worth noting that in the Discovery space, medicinal chemists are usually the leads and the drivers of the program. This element of discovery programs include many fields of research such as biology, pharmacology, biomarker identification, functional genomics, metabolism, pharmacokinetics, toxicology, pharmaceutics, and so forth and most often chemists in partnership with the biologists are at the heart of the team and will progress the program based on structural–activity relationship (SAR). The success of the discovery program will depend in part on how well the different cross-functional scientific information and historical data are captured to ensure the delivery of an effective and safe drug candidate. For example, medicinal chemists from Hoffmann-La Roche had developed ROCK, which is a wiki-based application to capture, browse, and search information, key discoveries, and property effects related to a chemical structure to aid their drug design.5 These types of tools help Discovery scientists to avoid reinvention, but rather learn from the past experience and historical data to save time and innovate new ideas. Furthermore, it is important for the Development scientists who are tasked to interface with the chemists and discovery scientists to learn this area of Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
1
2
COMMENTARY
expertise because this will be very different from how development projects are managed and progressed. More recently Ritchie and Macdonald6 and Bickerton et al.7 suggested the use of quantitative estimate of drug-likeness score based on the calculated physicochemical properties. However, this score is not able to discriminate drugs with respect to metabolic profile and route of elimination. On the contrary, Braggio et al.8 introduced a drug efficiency concept that they claimed as the key lead optimization parameter that can select the best molecule with high in vivo potency with potentially lower therapeutic doses. They claimed that changes in physicochemical properties of a molecule or the effect of in vitro Absorption, distribution, metabolism and excretion (ADME) or changes in a chemical structure contribute less to drug efficiency. For example, an increase in lipophilicity can decrease aqueous solubility and increase metabolic clearance that reduces systemic exposure, whereas, in many cases, increasing lipophilicity can also increase permeability, favoring absorption and penetration into the target compartment.9 Perhaps the complement of these two approaches will be needed to allow for a realistic lead optimization. With all these process enhancements and streamlined welldefined drug-likeness criteria for improved candidate selection described so far in the literature, what else can be done? With the infrastructure supportable by upper management in place, it will be the science and innovations that will be needed to bring this to the next level.
Figure 1. The ideal Discovery team makeup.
cited in the literature that showed that drug-likeness does not correlate well with the fate of marketed drugs.19
WHAT CAN WE DO BETTER? Based on what has been shared in the last few decades on success and pitfalls in discovery, the rest of this commentary will be based on my 5 “must do” list for success from a pharmaceutical scientist’s perspective.
DRUG DISCOVERY OVERVIEW In 1997, Lipinski et al.9 had published the “rule of 5” (RO5) physical property guidelines for drug absorption. This paper became the leading measure of drug-likeness, with more than 1500 literature citations. They stated that based on database from clinical candidates reaching phase II, poor absorption are more likely when clog P > 5; molecular mass is >500 Da, the number of H-bond donors >5, and number of H-acceptor is >10. Since then, an increasing number of papers have emerged highlighting the importance of lipophilicity, which is measured by log of octanol–water partition coefficient (log P) on individual absorption, distribution, metabolism, elimination, and toxicology parameters and on overall compound developability during the lead optimization.10–12 However, several expert opinions have cautioned following these rules rigidly13,14 because it was recognized that many valuable marketed drugs were made at the margins or even outside the boundaries of these proposed drug’s rigid properties.15 It was reported recently that most marketed drugs have become increasingly larger in size and more lipophilic.10,16,17 Some reasons for this increase were: advances in synthetic chemistryrelated methodologies, improvements in biological assays, an unhealthy preoccupation with high potency (where in vitro potency is normally sitting at the top of the screening cascade and is viewed as a filter for compound progression)18 and the introduction of more challenging drug discovery targets with shallow, lipophilic, or hydrophilic binding pockets.19 For example, “Best-in-Class” like the follow-on statins have higher molecular weights than their “First-in-Class” predecessors but they possess higher oral bioavailability, which goes against the convention of RO5 and would have been rejected if RO5 had been strictly enforced as a filter. Many more examples have been Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
Operating Model As part of the entrepreneurial activities in drug discovery, it is important to consider the implication of the makeup of the discovery team as discussed earlier (Fig. 1). The team has to be cross-functional and less hierarchical, with the authority to decide what scientific data are required to make the right decision to go forward or not. Team members must understand the background and history of the target and candidates such that there is no “tossing of projects over the fence.” Ownership has to be from beginning to end. For small pharmaceutical companies that will have limited internal resources, the discovery team will consist of consultants or contractors or CROs that will represent these functional areas. This may not be a new concept with all the changes and re-shaping occurring in the enterprise; it is still worthwhile to compare this model to existing ones to make sure that teams are identified appropriately. A truly integrated team and not just an “on-call participant” is required for Discovery to be successful. The team has to function like a separate business unit by itself. They should concentrate on big picture evaluation and not functional silos. Infrastructure to support this team will also be needed, for example, information technology group that can help consolidate the information that will become the database for follow-on activities or the development of in silico or modeling tools. Collaboration with CRO to help manage the resources is also the key to success. Additional collaboration with top academic thinkers combined with expertise in industry will improve science and innovation in discovery. The team should be able to leverage other functional areas if more resources are needed. To be a nimble discovery functional team, each team member requires the following competencies: DOI 10.1002/jps.24294
COMMENTARY
r strong r r r
leadership qualities (e.g., speaking up, educating each other on functional expertise, effective decision-making); knowledgeable in every aspects of discovery including understanding of the targets, biology, biochemistry, and so forth; shared mindset of a discovery scientists (e.g., discovery area constraints: API sparing, time limited activities, not restricted by SOPs, ability to engineer and conclude simple “Killer” experiments); familiar with innovation in the field that can enable their discovery and prevent show stoppers.
Perhaps the Collins philosophy of “getting the right people on the bus” needs to be taken seriously20 with the assumption that the right people can generate the right data to make the right decisions. “Go/no-go” decisions have to be clearly established for the team. Furthermore, upper management needs to clarify how much risk this team can manage to bring a candidate forward to the development. This should be clearly a “playbook”-free zone. Begin with the End in Mind A desirable Target Product Profile (TPP), that is, input from clinicians, key opinion leaders, and marketing for product research including patient-focused needs, needs to be identified Table 1.
3
before the decision to pre-invest in a target. On the contrary, if a project will be “First-in-Class,” filters for the candidate may be not as stringent. The following are a few examples of desired attributes of TPP:
r oral dosing, r low dose, r small-unit dosage form, r cost affordable, r once a day dosing, r no food effect, r no discernible taste, r immediate onset, r determine whether compound is “First-in-Class” or Bestin-Class,” and
r “Best-in-Class” in a therapeutic area will require the evaluation of comparator products. Lego Bricks Clearer data means clearer decisions. Table 1 summarizes the staged assays, complexity of each stage, data that is being generated using the appropriate type of testing. This commentary will not go in details as to how each dataset is generated. There are several reviews over the past few years that have captured the details of generating appropriate data very well.2–4,16,22–24 It is very important to note that from all of these reviews
Stage Activities, Data, and Complexity Lead Identification
Complexity
r >100 Leads r Low degree of chemical purity r Not characterized r ∼1 g materials
Lead Optimization
r 2–10 Leads r Some control of purity r Not characterized but may have r
information of physical form Scale up to 10 g
r 1–2 Candidates r Better purity and analytical methods
r Optimization of synthesis r Crystallization/salt/co-crystal screen
r Scale to ∼100 g Properties
r High potency (in vitro/in vivo) r r r r r r
Assay27
(1–10 nM) clog P/Mw ∼>0.66 (Ref.25) clog P 30 :M); counterscreen-off target activity r eADMET (metabolic stability, protein binding, minimal CYP inhibition and transporters, covalent binding, and no bioactivation) r pKa r pH solubility (pH 2/5.5/6.8) r Limited solubility screen in common solubilizers Medium throughput and specialized focused screens; miniaturized systems
r Establishing risk assessment using developability index
r Preclinical formulation optimized for exposure and acceptable safety
Manual data generation, some low throughput screen, use of platform methods, free “GxP” zone
hERG, human ether-a-go-go related gene; HTS, high throughput screening; PD, pharmacodynamics; eADMET, early/electronic computational models designed to predict the absorption, distribution, metabolism, excretion, and toxicity of molecules; GxP, for example good manufacturing practices, good laboratory practices that are covered under regulated environment. DOI 10.1002/jps.24294
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
4
COMMENTARY
and references, some salient features of generating relevant data are shared. For example, in the design stage, it would make sense to use predictive tools for physicochemical properties leveraging commercial software packages to screen all the thousands of leads in the earlier stage. Often in silico tools such as Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) predictor are being used to predict the risks associated with the candidate in development. High-throughput (HT) measurements, which are both cost and time effective, can generate actual experimental data. HT allowed for low amounts of materials usage while generating data that will help correlate physical properties to key elements such as toxicity or ADME. However, HT screen also have limitations such as range of measurements and will still need to have some human intervention to determine the true meaning of the results. The information generated at this stage will hopefully help make decisions as to the SAR and to assess correlation between structural features and physicochemical and ADMET properties, similar to building a structure using Lego bricks. An example of a good approach is described by Hill and Young,22 where interrelationship between physical properties can generate a solubility forecast index. Another example, which is my preference, is the method used by Yusof et al.,28 where they defined a desirability function for each property. The desirability of multiple properties can then be combined to obtain an overall “score” for the compound. The desirability function will depend on the uncertainty in the underlying data, the importance of this property to the target project objective, and so forth. This will make it clear when “the compound” can be confidently distinguished and avoids missed opportunities caused by inappropriate rejection based on some fixed rules. This selection approach is appropriate for the team knowing that the candidates did not die in vain. Once the leads are narrowed down to a few and other functional areas have a better understanding of the target, ADME properties and so forth, specialized medium throughputs and focused screens are used to generate critical datasets as listed in Table 1. It is also important to keep in
Table 2.
mind that the eventual lead/candidate has to meet the TPP (#2 above) that had been carefully set for each project during the inception of the team. Once the team selects a candidate for the development, data obtained using more reliable methods specific for the candidate will need to be generated in anticipation of results used in regulatory filing. However, the methods and testing will be in a non-Good Manufacturing Practice (GMP) environment. Developability Risk Assessment Before a candidate is considered for the development, we need to satisfy the next “must do” criteria, which is to confirm that this is truly a developable candidate. Table 2 summarizes the information that will need to be generated or compiled to allow for the evaluation of the Developability Indicators. These compound-specific properties will help assess the risk of developing the candidate that can satisfy the TPP. Moving molecule along the optimal discovery cascade can help speed up and lessen the complexity of the clinical development phase. Data generated from Table 2 will provide the team a dry run of what to expect when this candidate is developed. Physicochemical characterization such as evaluating the crystallinity, stability, potential for salt formation or cocrystallization, and polymorphism are important characteristics that most medicinal chemists are not driven to consider for the development. The pharmaceutical scientists and Process Research chemists will need to step up to take charge of this area to ensure that the active pharmaceutical ingredient (API) will be acceptable for scale-up within acceptable impurity limits, stable under the handling conditions and no genotoxicity alerts. Having both pharmaceutics scientists and Process Research chemists working together also allows for the sufficient API supplies at this stage of the evaluation. At the same time, the potential to formulate the compound for two species to provide the 10–50× exposure margin and dose proportionality are also important developability criteria to assess. If toxicity
Establishing Pharmaceutics Developability Indicators23
r Physicochemical characterization of the selected form о Salt/co-crystal screen о Polymorph screen о Scalability о Stability [chemical and physical including hygroscopicity (using dynamic vapor sorption (DVS)] о Solubility in physiological pH range including Fed State Simulated Intestinal Fluid (FESSIF / Fasted State Simulated Intestinal Fluid (FASSIF) о Bulk density, Flowdex, and particle size distribution (PSD)
r In vivo bioavailability in two species at relevant doses using suspension initially r Dose linearity to define high-dose limit with potential for 50-fold margin of exposure r High-dose Pharmacokinetic-Pharmacodynamic (PK–PD) using enabling formulation with sustained-release profile r Exploratory toxicology studies [hepatotoxicity alerts, cardiovascular (CV) safety, hERG (IC50 >30 :M] r ADME—metabolic stability; accumulation ratio, t1/2 , clearance, Biopharmaceutical Drug Disposition Classification System (BDDCS), efflux transporter substrate, CYP450 inhibition, and induction
r Predicted dose/maximum absorbable dose (MAD)a r Achievement of target product profile (TPP)
a MAD = S × K a × SIWW × SITT where S, solubility at pH 6.5 (mg/mL); Ka , permeability or trans-intestinal absorption rate constant (min−1 ); SIWW, small intestinal water volume ∼250 mL; SITT, small intestine transit time ∼270 min.
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
DOI 10.1002/jps.24294
COMMENTARY
cannot be assessed properly at this stage, this risk will be carried forward and can become a costly show stopper in the clinic. A strategy to get an estimation of the efficacious dose in human even at this early stage will be useful and will provide a good starting point to determine what dose strengths to supply for the first-in-man studies. It is obvious that low-dose-strength candidates will be at lower risk than going into the clinic with higher doses >300 mg. Formulation development, the potential for adverse effects, and risk of idiosyncratic drug reactions may be potential issues with high doses. Estimation of the maximum absorbable dose (MAD) from the formulation point of view will also help in identifying molecules with severely limited absorption resulting in the underestimation of development risk. This value in combination with the efficacious clinical dose estimation can be used to assess the complexity of the formulation and establish formulation strategies if MAD is low.29 More in depth description of each of the physicochemical properties are covered in other comprehensive reviews.4,23 Preclinical Formulation Development Information Gathering Formulations have always been the center of controversial discussions whenever the Discovery team gets excited about a lead or a candidate. In this commentary, the author will not dwell into the details of what formulations are considered or the conduct of how to support the formulation for toxicity studies. References with exhaustive lists of allowable vehicles for nonclinical oral safety studies are available since 2006. A few of these are provided in the reference section.30–33 Other references capture the challenges for formulating compounds that have been otherwise classified as brick dust solubility. Approaches to formulation for various delivery routes, goals of formulations, and physicochemical characteristics that are needed to select the appropriate vehicles have been covered in the literature.34–36 In most cases, formulation approaches do somehow provide the solutions that the team needs to increase exposure. However, an understanding of the properties and ADME profile of the candidate can help to ease the endless formulation screens that can still result in failure to achieve adequate exposure. For this reason, rather than repeating what is found in the references, the author would like to comment on what information should be collected that will allow for proper formulation development (Table 3). For example, what is the solubility of Table 3.
5
the candidate in FASSIF or FESSIF? What is the pH–solubility profile? These will determine whether pH control is required or whether there is sufficient solubility to manage the exposure at higher doses. Another example, in previous PK studies at low dose, what solid state form was used? Was there any crystallization or dissociation (if a salt) occurring during dosing? Often, simple optical microscopy can be used to answer some of these questions. If the form is crystalline and the exposure is adequate, then the formulation used for the low-dose study may be adequate. Another important point is the PK profile. Scientists always have a habit of just comparing Area under the curve (AUC) with Peak concentration (Cmax) and not take the initiative to look at the time–concentration curve. PK profiles can provide a wealth of knowledge. This can provide one with an evaluation of how accurate the AUC estimation was. If there are not enough time points to determine the Cmax , the estimation of the AUC will be flawed. Other information the profiles can provide are: (1) potential for accumulation during repeated dosing, (2) location of gut absorption, (3) relationship of Cmax to adverse effects, (4) variability, and (5) desired PK profile based on PK–PD studies. Is Cmax , AUC, or maintaining a plasma concentration above a targeted plasma trough concentration (Cmin) important for efficacy? Would a low Cmax /Ctrough ratio a desired profile? These parameters listed in Table 3 all play a key role in the selection of appropriate vehicles and API properties, solubilizers, surfactants, cosolvents, optimum pH for the formulation, and the needs for an enabling formulation.
CONCLUSIONS Looking ahead, the author think it will be great to apply a different approach of selecting development candidates, as outlined in this commentary. Note that this commentary made the assumption that a good therapeutic target had been selected. Working closely with medicinal chemists in designing molecules that are potent, bioavailable, and safe is a must. Rather than reinventing and overhauling current practices, stop and think of what can be improved. Get out of the comfort zone and keep challenging existing approaches. Because the Pharma laboratories are obviously generating candidates that are developed and are in clinical studies, it will be important to cross-check what can be performed better and what worked well. Create a measurement of what success looks
Critical Information Requirements for Formulation Selection for Oral Nonclinical Safety Studies
Physicochemical Properties
ADME Properties
Formulatability
API stability in solution Native pH in solution/pKa
MAD Desired PK profile Plasma sampling time points for PK studies Permeability (CaCo2 ) Transporter (substrate/inhibitor)
Duration of administration PK in suspension (properties—PSD/form) Acceptable excipient(s)/salt for tox study Assessment of scalability–viscosity, process and sedimentation Maximum formulatability
log D (pH 7) Solubility in FASSIF/SGF Known impurities—any alerts, are these also metabolites? Form of existing API used so far
BDDCS Accumulation factor Enzyme inducer Target exposure margin (10× or 50×) PK profile
Dose range
SGF, simulated gastric fluid. DOI 10.1002/jps.24294
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
6
COMMENTARY
like. Make sure that scientists are still current on literature and try to publish what they learned. This will help and force them to challenge their scientific knowledge and compare what they have toward new existing methodology. It will be appropriate for the discovery community to create an open forum similar to the International Consortium on Innovation and Quality in Pharmaceutical Development (IQ Consortium, an association of pharmaceutical and biotechnology companies aiming to advance innovation and quality in the development of pharmaceuticals through scientifically driven best practices and standards) to discuss best practices and also not only share success stories but also failures of commercializing efforts. CROs have to really bridge the gaps to support the industry with the appropriate services that the academics also have shared or even partner with them to commercialize these technologies. A huge effort is needed to make preclinical data translational to human clinical outcome especially in toxicology and should be the future research direction in discovery. The Roundtable of Toxicologists Consultants (RTC) should leverage their experiences and can collaborate with academia to continue efforts identifying relevant models to predict toxicity. Computational sciences will need to partner closely with discovery scientists to fill the gaps. This should be made more accessible for the community (both academia and industry) and challenged by applying the mathematical models to real cases. Hopefully, applicability of these in silico tools in discovery can be encouraged when the community understands this better. Last, there are still gaps in formulation of poorly soluble drugs/-lead candidates and technologies to increase exposure. Despite some complexity, nanoformulations, solid dispersions through spray drying, and hot melt are making huge strides. Lately, drug delivery/formulation summits, conferences, and forums are being organized that can hopefully address the gaps and allow for translation of exposure from toxicology studies to clinical studies. On the contrary, if the Discovery team considered the 5 “must do” list properly, enabling formulations may not be needed to advance drug discovery.
ACKNOWLEDGMENTS The author would like to thank John Higgins for his helpful critique and edits of this manuscript. In addition, the author is also indebted to Leanne Bedard of Bedard ADME-Tox Solutions and Karine Khougaz for their helpful assistance, discussions, and insight into the science of drug discovery and development.
REFERENCES 1. Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lindborg SR, Schacht AL. 2010. How to improve R&D productivity: The pharmaceutical industry’s grand challenge. Nat Rev/Drug Discov 9:203–214. 2. Palucki M, Higgins JD, Kwong E, Templeton AC. 2010. Strategies at the interface of drug discovery and development: Early optimization of the solid state phase and preclinical toxicology formulation for potential drug candidate. J Med Chem 53:5897–5905. 3. Caldwell GW, Ritchie DM, Masucci JA, Hageman W, Yan Z. 2001. The new pre-preclinical paradigm: Compound optimization in early and late phase drug discovery. Curr Topics Med Chem 1:353–366. 4. Saxena V, Panicucci R, Joshi Y, Garad S. 2009. Developability assessment in pharmaceutical industry: An integrated group approach for selecting developable candidates. J Pharm Sci 98(6):1962– 1979. Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
5. Mayweg A, Hofers U, Schnider P, Agnetti F, Galley G, Mattei P, Lucas M, Boehm H-J. 2011. ROCK: The Roche medicinal chemistry knowledge application—Design, use and impact. Drug Discov Today. 16(15–16):691–696. 6. Ritchie TJ, Macdonald SJF. 2014. How drug-like are “ugly” drugs: Do drug likeness metrics predict ADME behaviour in humans? Drug Discov Today 19(4):489–495. 7. Bickerton GR, Paolini GV, Besnard J, Moresan S, Hopkins AL. 2012.Quantifying the chemical beauty of drugs. Nat Chem 4:90–98. 8. Braggio S, Montanari D, Rossi T, Ratti E. 2010. Drug efficiency: A new concept to guide lead optimization programs towards the selection of better clinical candidates. Expert Opin Drug Discov 5(7):609–618. 9. Lipinski CA, Lombardo F, Dominy BW, Freeney PJ. 1997. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 23:3–25. 10. Lesson PD, Springthorpe B. 2007. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat Rev Drug Discov 6:881–890. 11. Veber DF, Johnson SR, Cheng H-Y, Smith BR, Ward KW, Kopple KD. 2002. Molecular properties that influence the oral bioavailability of drug candidates. J Med Chem 45:2615–2623. 12. Wenlock MC, Austin RP, Barton P, Davis AM, Leeson PD. 2003. A comparison of physicochemical property profiles of development and marketed oral drugs. J Med Chem 46:1250–1256. 13. Abad-Zapatero C. 2007. A Sorcerer’s apprentice and the Rule of Five: From rule of thumb to commandment and beyond. Drug Discov Today 12:995–997. 14. Leeson PD, Davis AM, Steele J. 2004. Drug-like properties: Guiding principles for design- or chemical prejudice? Drug Discov Today Technol 1(3):189–195. 15. Fotouhi N, Gillespie P, Goodnow R Jr. 2008. Lead generation: Reality check on commonly held views. Exp Opin Dug Discov 3:733– 744. 16. Leeson PD, Davis AM. 2004. Time related differences in the physical property profiles of oral drugs. J Med Chem 47:6338– 6348. 17. Walters WP, Green J, Weiss JR, Murcko MA. 2011. What do medicinal chemists actually make? A 50-year retrospective. J Med Chem 54:6405–6416. 18. Hann MM. 2011. Molecular obesity, potency and other addictions in drug discovery. Med Chem Commun 2:349–355. 19. Zhao H. 2011. Lead optimization in the nondrug-like space. Drug Discov Today 16:158–163. 20. Collins JC. 2001. Good to great: Why some companies make the leap- and others don’t. 1st ed. New York, New York: Harper-Business, pp xii, 300. 21. Hughes JP, Rees S, Kalindjian SB, Philpott KL. 2011. Principles of early drug discovery. Br J Pharm 162:1239–1249. 22. Hill AP, Young RJ. 2010. Getting physical in drug discovery: A contemporary perspective on solubility and hydrophobicity. Drug Discov Today 15(15–16):648–655. 23. Hageman MJ. 2010. Preformulation designed to enable discovery and assess developability. Comb Chem High Throughput Screen 13:90– 100. 24. Balbach S, Korn C. 2004. Pharmaceutical evaluation of early development candidates “the 100mg-approach”. Int J Pharm 275:1–12. 25. Ritchie TJ, Macdonald SJF, Peace S, Pickett SD, Luscombe CN. 2013. Increasing small molecule drug developability in sub-optimal chemical space. Med Chem Commun 4:673–680. 26. Waring MJ. 2010. Lipophilicity in drug discovery. Expert Opin Drug Discov 5(3):235–247. 27. Di L, Fish PV, Mano T. 2012. Bridging solubility between drug discovery and development. Drug Discov Today 17 (9-10):486– 495. 28. Yusof I, Shah F, Hashimoto T, Segall MD, Greene N. 2014. Finding the rules for successful drug optimization. Drug Discov Today 19(5):680–687. DOI 10.1002/jps.24294
COMMENTARY
29. Ding X, Rose JP, Van Gelder J. 2012. Developability assessment of clinical drug products with maximum absorbable doses. Int J Pharm 427:260–269. 30. Gad SC, Cassidy CD, Aubert N, Spainhour B, Robbe H. 2006. Nonclinical vehicle use in studies by multiple routes in multiple species. Int J Toxicol 25:499–521. 31. Strickley RG. 2006. Solubilizing excipients in pharmaceutical formulations. Encycl Pharm Technol 1:3334–3366. 32. Thackaberry EA. 2013. Vehicle selection for nonclinical oral safety studies. Expert Opin Drug Metab Toxicol 9(12):1635– 1646. 33. Gopinathan S, O’Neill E, Rodriguez LA, Champ R, Philips M, Nouraldeen A, Wendt MA, Wilson GE, Kramer JA. 2013. In vivo tox-
DOI 10.1002/jps.24294
7
icology of excipients commonly employed in drug discovery in rats. J Pharmacol Toxicol Methods 68:284–295. 34. Neervannan S. 2006. Preclinical formulations for discovery and toxicology: Physicochemical challenges. Expert Opin Drug Metab Toxicol 2(5):715–731. 35. Higgins J, Cartwright ME, Templeton AC. 2012. Foundation review: Progressing preclinical drug candidates: Strategies on preclinical safety studies and the quest for adequate exposure. Drug Discov Today 17(15–16):828–836. 36. Wuelfing WP, Kwong E, Higgins J. 2012. Identification of suitable formulations for high dose oral studies in rats using in vitro solubility measurements, the maximum absorbable dose model, and historical data set. Mol Pharm 9(5):1163–1174.
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
Advancing Drug Discovery: A Pharmaceutics Perspective ELIZABETH KWONG Kwong Eureka Solutions, Quebec, Canada Received 9 July 2014; revised 25 September 2014; accepted 4 November 2014 Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24294 ABSTRACT: Current industry perspective of how discovery is conducted seems to be fragmented and does not have a unified overall outlook of how discovery challenges are being addressed. Consequently, well-defined processes and drug-likeness criteria are being viewed as “broken” and will not maintain future R&D productivity. In this commentary, an analysis of existing practices for defining successful development candidates resulted in a 5 “must do” list to help advance Drug Discovery as presented from a Pharmaceutics perspective. The 5 “must do” list includes: what an ideal discovery team model should look like, what criteria should be considered for the desired development candidate profile, what the building blocks of the development candidate should look like, and how to assess the development C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci risks of the candidate. Keywords: drug-like properties; operating models in discovery; physicochemical properties; Drug Design; ADMET properties; insilico modeling; in-vitro models;target therapeutic profile (TPP); developability risk assessment; MAD; rule of five
INTRODUCTION Despite today’s challenging environment, pharmaceutical industry players continue to strive to deliver scientific excellence, meet unmet medical needs, and drive state-of-the-art innovation. At the same time, the industry and its stakeholders are cognizant that the current model is not sustainable for long term. To fix this situation, one of the analyses on R&D productivity1 has pointed toward the reduction of costs and cycle time. At the same time, shifting compound attrition to earlier during lead optimization before the first-in-human stage of clinical development will be a part of the cost-saving features of this enhanced productivity. In other words, the “preclinical stage” investment is now becoming the new “Phase IIb proofof-concept study” for key “go/no-go” decisions. By their estimation, this paradigm shift will increase the overall probability of technical success in late-stage phases II and III. This analysis therefore suggests a refocus of resources to discovery research and early translational medicine. Finally, for any successful R&D, there is no substitute for good people and good science. Investment in talent and getting the right people at the bench and in decision-making roles is critical.1 Since the publication from Paul et al.,1 there has been a longrange planning in the pharmaceutical industry to focus on process efficiencies in the preclinical stage. This has been achieved by optimizing decision-making in discovery research, which includes faster “go/no-go” decisions about progressing programs into clinical development. In addition, improvements have been implemented in research capabilities based on the newest scientific insights and the optimal use of resources to balance multiple projects, cost, priorities, and productivity. It will be increasingly important for scientists to not only think from the left or right side of their brain but rather thinking using one brain. As a consequence, a conscious effort to re-organize discovery groups evolved, which included the integration of the Correspondence to: Elizabeth Kwong (Telephone: +514-694-6475; E-mail: ec [email protected]) Journal of Pharmaceutical Sciences C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association
preclinical/development functions such as drug metabolism and pharmacokinetics, pharmaceutics, process chemistry research, and toxicology into drug discovery,2,3 in hopes of improving the efficiency of generating candidates for the development. At the same time, this shift of resources from early development to the discovery space allowed integrated teams such as Developability Assessment Groups4 to help drive the optimization of well-balanced “drug-like” properties of the candidate. Continued improvement on this holistic approach and reorganization of discovery team resulted in identifying liabilities early, and enabled more rationale candidate selection decisions. It was also suggested that drug delivery studies, drug–drug interaction assessment, and safety pharmacology assays can be used to provide preclinical information necessary to select a drug candidate with the best overall pharmaceutical profile. Higher sample throughput screens, better in vitro cell models, and computational models have been employed to manage resources, costs, and time.3 It is also worth noting that in the Discovery space, medicinal chemists are usually the leads and the drivers of the program. This element of discovery programs include many fields of research such as biology, pharmacology, biomarker identification, functional genomics, metabolism, pharmacokinetics, toxicology, pharmaceutics, and so forth and most often chemists in partnership with the biologists are at the heart of the team and will progress the program based on structural–activity relationship (SAR). The success of the discovery program will depend in part on how well the different cross-functional scientific information and historical data are captured to ensure the delivery of an effective and safe drug candidate. For example, medicinal chemists from Hoffmann-La Roche had developed ROCK, which is a wiki-based application to capture, browse, and search information, key discoveries, and property effects related to a chemical structure to aid their drug design.5 These types of tools help Discovery scientists to avoid reinvention, but rather learn from the past experience and historical data to save time and innovate new ideas. Furthermore, it is important for the Development scientists who are tasked to interface with the chemists and discovery scientists to learn this area of Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
1
2
COMMENTARY
expertise because this will be very different from how development projects are managed and progressed. More recently Ritchie and Macdonald6 and Bickerton et al.7 suggested the use of quantitative estimate of drug-likeness score based on the calculated physicochemical properties. However, this score is not able to discriminate drugs with respect to metabolic profile and route of elimination. On the contrary, Braggio et al.8 introduced a drug efficiency concept that they claimed as the key lead optimization parameter that can select the best molecule with high in vivo potency with potentially lower therapeutic doses. They claimed that changes in physicochemical properties of a molecule or the effect of in vitro Absorption, distribution, metabolism and excretion (ADME) or changes in a chemical structure contribute less to drug efficiency. For example, an increase in lipophilicity can decrease aqueous solubility and increase metabolic clearance that reduces systemic exposure, whereas, in many cases, increasing lipophilicity can also increase permeability, favoring absorption and penetration into the target compartment.9 Perhaps the complement of these two approaches will be needed to allow for a realistic lead optimization. With all these process enhancements and streamlined welldefined drug-likeness criteria for improved candidate selection described so far in the literature, what else can be done? With the infrastructure supportable by upper management in place, it will be the science and innovations that will be needed to bring this to the next level.
Figure 1. The ideal Discovery team makeup.
cited in the literature that showed that drug-likeness does not correlate well with the fate of marketed drugs.19
WHAT CAN WE DO BETTER? Based on what has been shared in the last few decades on success and pitfalls in discovery, the rest of this commentary will be based on my 5 “must do” list for success from a pharmaceutical scientist’s perspective.
DRUG DISCOVERY OVERVIEW In 1997, Lipinski et al.9 had published the “rule of 5” (RO5) physical property guidelines for drug absorption. This paper became the leading measure of drug-likeness, with more than 1500 literature citations. They stated that based on database from clinical candidates reaching phase II, poor absorption are more likely when clog P > 5; molecular mass is >500 Da, the number of H-bond donors >5, and number of H-acceptor is >10. Since then, an increasing number of papers have emerged highlighting the importance of lipophilicity, which is measured by log of octanol–water partition coefficient (log P) on individual absorption, distribution, metabolism, elimination, and toxicology parameters and on overall compound developability during the lead optimization.10–12 However, several expert opinions have cautioned following these rules rigidly13,14 because it was recognized that many valuable marketed drugs were made at the margins or even outside the boundaries of these proposed drug’s rigid properties.15 It was reported recently that most marketed drugs have become increasingly larger in size and more lipophilic.10,16,17 Some reasons for this increase were: advances in synthetic chemistryrelated methodologies, improvements in biological assays, an unhealthy preoccupation with high potency (where in vitro potency is normally sitting at the top of the screening cascade and is viewed as a filter for compound progression)18 and the introduction of more challenging drug discovery targets with shallow, lipophilic, or hydrophilic binding pockets.19 For example, “Best-in-Class” like the follow-on statins have higher molecular weights than their “First-in-Class” predecessors but they possess higher oral bioavailability, which goes against the convention of RO5 and would have been rejected if RO5 had been strictly enforced as a filter. Many more examples have been Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
Operating Model As part of the entrepreneurial activities in drug discovery, it is important to consider the implication of the makeup of the discovery team as discussed earlier (Fig. 1). The team has to be cross-functional and less hierarchical, with the authority to decide what scientific data are required to make the right decision to go forward or not. Team members must understand the background and history of the target and candidates such that there is no “tossing of projects over the fence.” Ownership has to be from beginning to end. For small pharmaceutical companies that will have limited internal resources, the discovery team will consist of consultants or contractors or CROs that will represent these functional areas. This may not be a new concept with all the changes and re-shaping occurring in the enterprise; it is still worthwhile to compare this model to existing ones to make sure that teams are identified appropriately. A truly integrated team and not just an “on-call participant” is required for Discovery to be successful. The team has to function like a separate business unit by itself. They should concentrate on big picture evaluation and not functional silos. Infrastructure to support this team will also be needed, for example, information technology group that can help consolidate the information that will become the database for follow-on activities or the development of in silico or modeling tools. Collaboration with CRO to help manage the resources is also the key to success. Additional collaboration with top academic thinkers combined with expertise in industry will improve science and innovation in discovery. The team should be able to leverage other functional areas if more resources are needed. To be a nimble discovery functional team, each team member requires the following competencies: DOI 10.1002/jps.24294
COMMENTARY
r strong r r r
leadership qualities (e.g., speaking up, educating each other on functional expertise, effective decision-making); knowledgeable in every aspects of discovery including understanding of the targets, biology, biochemistry, and so forth; shared mindset of a discovery scientists (e.g., discovery area constraints: API sparing, time limited activities, not restricted by SOPs, ability to engineer and conclude simple “Killer” experiments); familiar with innovation in the field that can enable their discovery and prevent show stoppers.
Perhaps the Collins philosophy of “getting the right people on the bus” needs to be taken seriously20 with the assumption that the right people can generate the right data to make the right decisions. “Go/no-go” decisions have to be clearly established for the team. Furthermore, upper management needs to clarify how much risk this team can manage to bring a candidate forward to the development. This should be clearly a “playbook”-free zone. Begin with the End in Mind A desirable Target Product Profile (TPP), that is, input from clinicians, key opinion leaders, and marketing for product research including patient-focused needs, needs to be identified Table 1.
3
before the decision to pre-invest in a target. On the contrary, if a project will be “First-in-Class,” filters for the candidate may be not as stringent. The following are a few examples of desired attributes of TPP:
r oral dosing, r low dose, r small-unit dosage form, r cost affordable, r once a day dosing, r no food effect, r no discernible taste, r immediate onset, r determine whether compound is “First-in-Class” or Bestin-Class,” and
r “Best-in-Class” in a therapeutic area will require the evaluation of comparator products. Lego Bricks Clearer data means clearer decisions. Table 1 summarizes the staged assays, complexity of each stage, data that is being generated using the appropriate type of testing. This commentary will not go in details as to how each dataset is generated. There are several reviews over the past few years that have captured the details of generating appropriate data very well.2–4,16,22–24 It is very important to note that from all of these reviews
Stage Activities, Data, and Complexity Lead Identification
Complexity
r >100 Leads r Low degree of chemical purity r Not characterized r ∼1 g materials
Lead Optimization
r 2–10 Leads r Some control of purity r Not characterized but may have r
information of physical form Scale up to 10 g
r 1–2 Candidates r Better purity and analytical methods
r Optimization of synthesis r Crystallization/salt/co-crystal screen
r Scale to ∼100 g Properties
r High potency (in vitro/in vivo) r r r r r r
Assay27
(1–10 nM) clog P/Mw ∼>0.66 (Ref.25) clog P 30 :M); counterscreen-off target activity r eADMET (metabolic stability, protein binding, minimal CYP inhibition and transporters, covalent binding, and no bioactivation) r pKa r pH solubility (pH 2/5.5/6.8) r Limited solubility screen in common solubilizers Medium throughput and specialized focused screens; miniaturized systems
r Establishing risk assessment using developability index
r Preclinical formulation optimized for exposure and acceptable safety
Manual data generation, some low throughput screen, use of platform methods, free “GxP” zone
hERG, human ether-a-go-go related gene; HTS, high throughput screening; PD, pharmacodynamics; eADMET, early/electronic computational models designed to predict the absorption, distribution, metabolism, excretion, and toxicity of molecules; GxP, for example good manufacturing practices, good laboratory practices that are covered under regulated environment. DOI 10.1002/jps.24294
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
4
COMMENTARY
and references, some salient features of generating relevant data are shared. For example, in the design stage, it would make sense to use predictive tools for physicochemical properties leveraging commercial software packages to screen all the thousands of leads in the earlier stage. Often in silico tools such as Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) predictor are being used to predict the risks associated with the candidate in development. High-throughput (HT) measurements, which are both cost and time effective, can generate actual experimental data. HT allowed for low amounts of materials usage while generating data that will help correlate physical properties to key elements such as toxicity or ADME. However, HT screen also have limitations such as range of measurements and will still need to have some human intervention to determine the true meaning of the results. The information generated at this stage will hopefully help make decisions as to the SAR and to assess correlation between structural features and physicochemical and ADMET properties, similar to building a structure using Lego bricks. An example of a good approach is described by Hill and Young,22 where interrelationship between physical properties can generate a solubility forecast index. Another example, which is my preference, is the method used by Yusof et al.,28 where they defined a desirability function for each property. The desirability of multiple properties can then be combined to obtain an overall “score” for the compound. The desirability function will depend on the uncertainty in the underlying data, the importance of this property to the target project objective, and so forth. This will make it clear when “the compound” can be confidently distinguished and avoids missed opportunities caused by inappropriate rejection based on some fixed rules. This selection approach is appropriate for the team knowing that the candidates did not die in vain. Once the leads are narrowed down to a few and other functional areas have a better understanding of the target, ADME properties and so forth, specialized medium throughputs and focused screens are used to generate critical datasets as listed in Table 1. It is also important to keep in
Table 2.
mind that the eventual lead/candidate has to meet the TPP (#2 above) that had been carefully set for each project during the inception of the team. Once the team selects a candidate for the development, data obtained using more reliable methods specific for the candidate will need to be generated in anticipation of results used in regulatory filing. However, the methods and testing will be in a non-Good Manufacturing Practice (GMP) environment. Developability Risk Assessment Before a candidate is considered for the development, we need to satisfy the next “must do” criteria, which is to confirm that this is truly a developable candidate. Table 2 summarizes the information that will need to be generated or compiled to allow for the evaluation of the Developability Indicators. These compound-specific properties will help assess the risk of developing the candidate that can satisfy the TPP. Moving molecule along the optimal discovery cascade can help speed up and lessen the complexity of the clinical development phase. Data generated from Table 2 will provide the team a dry run of what to expect when this candidate is developed. Physicochemical characterization such as evaluating the crystallinity, stability, potential for salt formation or cocrystallization, and polymorphism are important characteristics that most medicinal chemists are not driven to consider for the development. The pharmaceutical scientists and Process Research chemists will need to step up to take charge of this area to ensure that the active pharmaceutical ingredient (API) will be acceptable for scale-up within acceptable impurity limits, stable under the handling conditions and no genotoxicity alerts. Having both pharmaceutics scientists and Process Research chemists working together also allows for the sufficient API supplies at this stage of the evaluation. At the same time, the potential to formulate the compound for two species to provide the 10–50× exposure margin and dose proportionality are also important developability criteria to assess. If toxicity
Establishing Pharmaceutics Developability Indicators23
r Physicochemical characterization of the selected form о Salt/co-crystal screen о Polymorph screen о Scalability о Stability [chemical and physical including hygroscopicity (using dynamic vapor sorption (DVS)] о Solubility in physiological pH range including Fed State Simulated Intestinal Fluid (FESSIF / Fasted State Simulated Intestinal Fluid (FASSIF) о Bulk density, Flowdex, and particle size distribution (PSD)
r In vivo bioavailability in two species at relevant doses using suspension initially r Dose linearity to define high-dose limit with potential for 50-fold margin of exposure r High-dose Pharmacokinetic-Pharmacodynamic (PK–PD) using enabling formulation with sustained-release profile r Exploratory toxicology studies [hepatotoxicity alerts, cardiovascular (CV) safety, hERG (IC50 >30 :M] r ADME—metabolic stability; accumulation ratio, t1/2 , clearance, Biopharmaceutical Drug Disposition Classification System (BDDCS), efflux transporter substrate, CYP450 inhibition, and induction
r Predicted dose/maximum absorbable dose (MAD)a r Achievement of target product profile (TPP)
a MAD = S × K a × SIWW × SITT where S, solubility at pH 6.5 (mg/mL); Ka , permeability or trans-intestinal absorption rate constant (min−1 ); SIWW, small intestinal water volume ∼250 mL; SITT, small intestine transit time ∼270 min.
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
DOI 10.1002/jps.24294
COMMENTARY
cannot be assessed properly at this stage, this risk will be carried forward and can become a costly show stopper in the clinic. A strategy to get an estimation of the efficacious dose in human even at this early stage will be useful and will provide a good starting point to determine what dose strengths to supply for the first-in-man studies. It is obvious that low-dose-strength candidates will be at lower risk than going into the clinic with higher doses >300 mg. Formulation development, the potential for adverse effects, and risk of idiosyncratic drug reactions may be potential issues with high doses. Estimation of the maximum absorbable dose (MAD) from the formulation point of view will also help in identifying molecules with severely limited absorption resulting in the underestimation of development risk. This value in combination with the efficacious clinical dose estimation can be used to assess the complexity of the formulation and establish formulation strategies if MAD is low.29 More in depth description of each of the physicochemical properties are covered in other comprehensive reviews.4,23 Preclinical Formulation Development Information Gathering Formulations have always been the center of controversial discussions whenever the Discovery team gets excited about a lead or a candidate. In this commentary, the author will not dwell into the details of what formulations are considered or the conduct of how to support the formulation for toxicity studies. References with exhaustive lists of allowable vehicles for nonclinical oral safety studies are available since 2006. A few of these are provided in the reference section.30–33 Other references capture the challenges for formulating compounds that have been otherwise classified as brick dust solubility. Approaches to formulation for various delivery routes, goals of formulations, and physicochemical characteristics that are needed to select the appropriate vehicles have been covered in the literature.34–36 In most cases, formulation approaches do somehow provide the solutions that the team needs to increase exposure. However, an understanding of the properties and ADME profile of the candidate can help to ease the endless formulation screens that can still result in failure to achieve adequate exposure. For this reason, rather than repeating what is found in the references, the author would like to comment on what information should be collected that will allow for proper formulation development (Table 3). For example, what is the solubility of Table 3.
5
the candidate in FASSIF or FESSIF? What is the pH–solubility profile? These will determine whether pH control is required or whether there is sufficient solubility to manage the exposure at higher doses. Another example, in previous PK studies at low dose, what solid state form was used? Was there any crystallization or dissociation (if a salt) occurring during dosing? Often, simple optical microscopy can be used to answer some of these questions. If the form is crystalline and the exposure is adequate, then the formulation used for the low-dose study may be adequate. Another important point is the PK profile. Scientists always have a habit of just comparing Area under the curve (AUC) with Peak concentration (Cmax) and not take the initiative to look at the time–concentration curve. PK profiles can provide a wealth of knowledge. This can provide one with an evaluation of how accurate the AUC estimation was. If there are not enough time points to determine the Cmax , the estimation of the AUC will be flawed. Other information the profiles can provide are: (1) potential for accumulation during repeated dosing, (2) location of gut absorption, (3) relationship of Cmax to adverse effects, (4) variability, and (5) desired PK profile based on PK–PD studies. Is Cmax , AUC, or maintaining a plasma concentration above a targeted plasma trough concentration (Cmin) important for efficacy? Would a low Cmax /Ctrough ratio a desired profile? These parameters listed in Table 3 all play a key role in the selection of appropriate vehicles and API properties, solubilizers, surfactants, cosolvents, optimum pH for the formulation, and the needs for an enabling formulation.
CONCLUSIONS Looking ahead, the author think it will be great to apply a different approach of selecting development candidates, as outlined in this commentary. Note that this commentary made the assumption that a good therapeutic target had been selected. Working closely with medicinal chemists in designing molecules that are potent, bioavailable, and safe is a must. Rather than reinventing and overhauling current practices, stop and think of what can be improved. Get out of the comfort zone and keep challenging existing approaches. Because the Pharma laboratories are obviously generating candidates that are developed and are in clinical studies, it will be important to cross-check what can be performed better and what worked well. Create a measurement of what success looks
Critical Information Requirements for Formulation Selection for Oral Nonclinical Safety Studies
Physicochemical Properties
ADME Properties
Formulatability
API stability in solution Native pH in solution/pKa
MAD Desired PK profile Plasma sampling time points for PK studies Permeability (CaCo2 ) Transporter (substrate/inhibitor)
Duration of administration PK in suspension (properties—PSD/form) Acceptable excipient(s)/salt for tox study Assessment of scalability–viscosity, process and sedimentation Maximum formulatability
log D (pH 7) Solubility in FASSIF/SGF Known impurities—any alerts, are these also metabolites? Form of existing API used so far
BDDCS Accumulation factor Enzyme inducer Target exposure margin (10× or 50×) PK profile
Dose range
SGF, simulated gastric fluid. DOI 10.1002/jps.24294
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
6
COMMENTARY
like. Make sure that scientists are still current on literature and try to publish what they learned. This will help and force them to challenge their scientific knowledge and compare what they have toward new existing methodology. It will be appropriate for the discovery community to create an open forum similar to the International Consortium on Innovation and Quality in Pharmaceutical Development (IQ Consortium, an association of pharmaceutical and biotechnology companies aiming to advance innovation and quality in the development of pharmaceuticals through scientifically driven best practices and standards) to discuss best practices and also not only share success stories but also failures of commercializing efforts. CROs have to really bridge the gaps to support the industry with the appropriate services that the academics also have shared or even partner with them to commercialize these technologies. A huge effort is needed to make preclinical data translational to human clinical outcome especially in toxicology and should be the future research direction in discovery. The Roundtable of Toxicologists Consultants (RTC) should leverage their experiences and can collaborate with academia to continue efforts identifying relevant models to predict toxicity. Computational sciences will need to partner closely with discovery scientists to fill the gaps. This should be made more accessible for the community (both academia and industry) and challenged by applying the mathematical models to real cases. Hopefully, applicability of these in silico tools in discovery can be encouraged when the community understands this better. Last, there are still gaps in formulation of poorly soluble drugs/-lead candidates and technologies to increase exposure. Despite some complexity, nanoformulations, solid dispersions through spray drying, and hot melt are making huge strides. Lately, drug delivery/formulation summits, conferences, and forums are being organized that can hopefully address the gaps and allow for translation of exposure from toxicology studies to clinical studies. On the contrary, if the Discovery team considered the 5 “must do” list properly, enabling formulations may not be needed to advance drug discovery.
ACKNOWLEDGMENTS The author would like to thank John Higgins for his helpful critique and edits of this manuscript. In addition, the author is also indebted to Leanne Bedard of Bedard ADME-Tox Solutions and Karine Khougaz for their helpful assistance, discussions, and insight into the science of drug discovery and development.
REFERENCES 1. Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lindborg SR, Schacht AL. 2010. How to improve R&D productivity: The pharmaceutical industry’s grand challenge. Nat Rev/Drug Discov 9:203–214. 2. Palucki M, Higgins JD, Kwong E, Templeton AC. 2010. Strategies at the interface of drug discovery and development: Early optimization of the solid state phase and preclinical toxicology formulation for potential drug candidate. J Med Chem 53:5897–5905. 3. Caldwell GW, Ritchie DM, Masucci JA, Hageman W, Yan Z. 2001. The new pre-preclinical paradigm: Compound optimization in early and late phase drug discovery. Curr Topics Med Chem 1:353–366. 4. Saxena V, Panicucci R, Joshi Y, Garad S. 2009. Developability assessment in pharmaceutical industry: An integrated group approach for selecting developable candidates. J Pharm Sci 98(6):1962– 1979. Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
5. Mayweg A, Hofers U, Schnider P, Agnetti F, Galley G, Mattei P, Lucas M, Boehm H-J. 2011. ROCK: The Roche medicinal chemistry knowledge application—Design, use and impact. Drug Discov Today. 16(15–16):691–696. 6. Ritchie TJ, Macdonald SJF. 2014. How drug-like are “ugly” drugs: Do drug likeness metrics predict ADME behaviour in humans? Drug Discov Today 19(4):489–495. 7. Bickerton GR, Paolini GV, Besnard J, Moresan S, Hopkins AL. 2012.Quantifying the chemical beauty of drugs. Nat Chem 4:90–98. 8. Braggio S, Montanari D, Rossi T, Ratti E. 2010. Drug efficiency: A new concept to guide lead optimization programs towards the selection of better clinical candidates. Expert Opin Drug Discov 5(7):609–618. 9. Lipinski CA, Lombardo F, Dominy BW, Freeney PJ. 1997. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 23:3–25. 10. Lesson PD, Springthorpe B. 2007. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat Rev Drug Discov 6:881–890. 11. Veber DF, Johnson SR, Cheng H-Y, Smith BR, Ward KW, Kopple KD. 2002. Molecular properties that influence the oral bioavailability of drug candidates. J Med Chem 45:2615–2623. 12. Wenlock MC, Austin RP, Barton P, Davis AM, Leeson PD. 2003. A comparison of physicochemical property profiles of development and marketed oral drugs. J Med Chem 46:1250–1256. 13. Abad-Zapatero C. 2007. A Sorcerer’s apprentice and the Rule of Five: From rule of thumb to commandment and beyond. Drug Discov Today 12:995–997. 14. Leeson PD, Davis AM, Steele J. 2004. Drug-like properties: Guiding principles for design- or chemical prejudice? Drug Discov Today Technol 1(3):189–195. 15. Fotouhi N, Gillespie P, Goodnow R Jr. 2008. Lead generation: Reality check on commonly held views. Exp Opin Dug Discov 3:733– 744. 16. Leeson PD, Davis AM. 2004. Time related differences in the physical property profiles of oral drugs. J Med Chem 47:6338– 6348. 17. Walters WP, Green J, Weiss JR, Murcko MA. 2011. What do medicinal chemists actually make? A 50-year retrospective. J Med Chem 54:6405–6416. 18. Hann MM. 2011. Molecular obesity, potency and other addictions in drug discovery. Med Chem Commun 2:349–355. 19. Zhao H. 2011. Lead optimization in the nondrug-like space. Drug Discov Today 16:158–163. 20. Collins JC. 2001. Good to great: Why some companies make the leap- and others don’t. 1st ed. New York, New York: Harper-Business, pp xii, 300. 21. Hughes JP, Rees S, Kalindjian SB, Philpott KL. 2011. Principles of early drug discovery. Br J Pharm 162:1239–1249. 22. Hill AP, Young RJ. 2010. Getting physical in drug discovery: A contemporary perspective on solubility and hydrophobicity. Drug Discov Today 15(15–16):648–655. 23. Hageman MJ. 2010. Preformulation designed to enable discovery and assess developability. Comb Chem High Throughput Screen 13:90– 100. 24. Balbach S, Korn C. 2004. Pharmaceutical evaluation of early development candidates “the 100mg-approach”. Int J Pharm 275:1–12. 25. Ritchie TJ, Macdonald SJF, Peace S, Pickett SD, Luscombe CN. 2013. Increasing small molecule drug developability in sub-optimal chemical space. Med Chem Commun 4:673–680. 26. Waring MJ. 2010. Lipophilicity in drug discovery. Expert Opin Drug Discov 5(3):235–247. 27. Di L, Fish PV, Mano T. 2012. Bridging solubility between drug discovery and development. Drug Discov Today 17 (9-10):486– 495. 28. Yusof I, Shah F, Hashimoto T, Segall MD, Greene N. 2014. Finding the rules for successful drug optimization. Drug Discov Today 19(5):680–687. DOI 10.1002/jps.24294
COMMENTARY
29. Ding X, Rose JP, Van Gelder J. 2012. Developability assessment of clinical drug products with maximum absorbable doses. Int J Pharm 427:260–269. 30. Gad SC, Cassidy CD, Aubert N, Spainhour B, Robbe H. 2006. Nonclinical vehicle use in studies by multiple routes in multiple species. Int J Toxicol 25:499–521. 31. Strickley RG. 2006. Solubilizing excipients in pharmaceutical formulations. Encycl Pharm Technol 1:3334–3366. 32. Thackaberry EA. 2013. Vehicle selection for nonclinical oral safety studies. Expert Opin Drug Metab Toxicol 9(12):1635– 1646. 33. Gopinathan S, O’Neill E, Rodriguez LA, Champ R, Philips M, Nouraldeen A, Wendt MA, Wilson GE, Kramer JA. 2013. In vivo tox-
DOI 10.1002/jps.24294
7
icology of excipients commonly employed in drug discovery in rats. J Pharmacol Toxicol Methods 68:284–295. 34. Neervannan S. 2006. Preclinical formulations for discovery and toxicology: Physicochemical challenges. Expert Opin Drug Metab Toxicol 2(5):715–731. 35. Higgins J, Cartwright ME, Templeton AC. 2012. Foundation review: Progressing preclinical drug candidates: Strategies on preclinical safety studies and the quest for adequate exposure. Drug Discov Today 17(15–16):828–836. 36. Wuelfing WP, Kwong E, Higgins J. 2012. Identification of suitable formulations for high dose oral studies in rats using in vitro solubility measurements, the maximum absorbable dose model, and historical data set. Mol Pharm 9(5):1163–1174.
Elizabeth Kwong, JOURNAL OF PHARMACEUTICAL SCIENCES
