Drug Discovery and Development – Some Interesting Math
Duane B. Lakings, Ph.D., Principal, DSE Consulting
As the personnel at most pharmaceutical and biotechnology companies should be aware, the cost, both time and money, for discovering and developing a novel therapeutic product is huge, with the estimates being over 12 years and $600 million, respectively. About half the time and most of the money is expended during late clinical phase development where extensive phase 3 clinical trials are conducted and the necessary manufacturing facilities are defined and put into place to support product launch after regulatory authority approval has been obtained. The $600 million guestimate includes the cost of the ‘losers’ (i.e., those drug candidates that enter the drug development process and ‘die’ before finishing). The cost for discovering and developing a single new therapeutic product (i.e., a ‘winner’) has been estimated to be around $120 million.
The total discovery and preclinical/nonclinical effort required to support one project for 12 years is about $24 million (see estimated breakdown below) or about 20% of the $120 million estimated for a single compound. The other $96 million, or 80% of the total, is expended on clinical trials and for establishing the API and drug product manufacturing processes and facilities.
Present estimates are that less than 1% of the compounds entering drug development successfully traverse the preclinical/nonclinical and clinical/manufacturing processes and have a marketing application submitted to a regulatory authority. About 90% of these ‘losers’ are identified during preclinical/nonclinical development with the primary reasons for stopping development being an unacceptable toxicology profile, undesirable pharmacokinetics or metabolism, or insufficient delivery to the site of pharmacologic action. Of the 10% that enter clinical trials, 9 out of 10 drug candidates ‘die’ because of insufficient efficacy in the proposed disease indication, unacceptable safety in humans, and/or undesirable pharmacokinetics or metabolism in humans.
Assuming that the 20/80 cost ratio for the discovery/preclinical/nonclinical and clinical/manufacturing processes is about right for the overall cost estimate ($600 million) for putting a single new therapeutic product on the market and that 100 candidates are needed to produce one ‘winner’, the preclinical/nonclinical cost for identifying 90% (90 of the 100) of the ‘losers’ is $120 million or $1.33 million each. To identify the 9 out 10 drug candidate ‘losers’ that enter clinical trials, $480 million or $53.3 million each is necessary.
If the discovery/preclinical/nonclinical effort could identify 95 out of 100 ‘losers’ (95% instead of 90%), thus allowing 1 out 5 candidates that enter clinical trials to be a ‘winner’, the overall cost of identifying a single new therapeutic product could be reduced to about $400 million ($126 million for discovery/preclinical/nonclinical and $266 million for clinical/manufacturing). If the discovery/preclinical/nonclinical effort identified 98% of the ‘losers’, the overall cost could be reduced to around $240 million or about twice of the cost estimated for developing a ‘winner’ without ‘losers’.
How can the discovery/preclinical/nonclinical identification of ‘losers’ be increased so that the overall cost of development can be reduced? The following are some thoughts and ideas on how to achieve this goal. While implementation of all these ‘ideas’ for each drug development project is neither justified nor warranted, some of them may useful for any given project and thus ‘save’ the drug candidate sponsor both time and money that can be utilized for the development of ‘winners’.
First, many discovery efforts are designed to identify novel compounds with the highest biological activity against a target, commonly a protein that is a receptor or an enzyme involved in a biochemical process and thought to be important in the mediation of a human disease or disorder. However, the active sites of targets are usually highly lipophilic and thus compounds with the ‘best’ ability to agonize or antagonize these targets are also highly lipophilic. Thus, in vitro assessments or HTS efforts will identify lipophilic compounds as being the most potent. However, for a lipophilic compound to reach the active site (i.e., the target) in an in vivo animal model or in human patients, the compound has to be transported in an aqueous environment (i.e., blood, extravascular fluids) from the site of administration to the target. Consequently, a compound with at least some hydrophilic properties (i.e., some aqueous solubility) may have somewhat less biological potency for a given target but may be a better candidate for in vivo mediation of a disease process because the compound can more effectively and in higher concentration be delivered to the site of action.
For drug discovery leads to become successful drug candidates and not to be one of 99 out of 100 ‘losers’, drug discovery groups first need to change the paradigm for selecting compounds from one based solely on biological potency to one that includes ‘delivery potential’. Many highly successful pharmaceutical companies have initiated this change by requiring NCE drug discovery leads to meet Lipinski’s Rules of Five, which uses chemical structure to estimate the delivery potential of compounds across membranes, before a lead can become a preclinical drug candidate. Thus, new drug candidates from these companies have at least some drug-like characteristics, which should improve both the preclinical and clinical success rates.
However, a primary goal for many smaller firms, who need to impress ‘investors’, is to initiate clinical trials as quickly as possible. Thus, a lead, either a NCE or a macromolecule, with the desired biological activity is ‘rushed through’ preclinical development where the minimal number of ‘standard’ toxicology studies are conducted and little or no data is obtained on other safety aspects, such as delivery, PK, and drug metabolism. Even in this ‘hurry’ environment, the lack of drug-like properties can ‘kill’ some candidates in preclinical development but a number still reach the clinic where they ‘die’, sometimes in phase 3 (to the dismay of ‘investors’ as demonstrated by the substantial drop in stock prices for a firm with a phase 3 ‘failure’) and after millions of dollars have already been expended. By slowing down the ‘rush’ process a little, say 6 months to a year, and carefully designing preclinical studies to be data productive for characterizing both the development attributes and demerits of a drug candidate, many of these ‘losers’ could be identified before reaching the clinic.
Second and possibly better yet, by carefully designing and conducting some preliminary, relatively not expensive (both in time and money) studies, the drug-like properties of a discovery lead or group of leads can be assessed. These developability assessment or lead optimization studies can quickly evaluate a number of key parameters to ascertain if a lead, or which compound from a group of leads, has the necessary attributes without major demerits to become a successful drug candidate. Attributes could include, but are not limited to:
A solubility and stability profile that allows administration by the proposed clinical route and delivery to the site of action in sufficient concentration to mediate a disease process.
A metabolism profile that is similar in proposed toxicology animal species and humans and does not cause the compound to be metabolically cleared from the body so rapidly that it does not have a chance to be an effective pharmacological agent.
A pharmacokinetic profile that produces a desired plasma/serum concentration time profile so that the concentration and residence time of the compound at the active site are sufficient to effectively agonize or antagonize the target to produce the desired effect.
An acute toxicity profile that does not produce adverse effects (or undesirable pharmacological activity in organ systems other than the target organ) at pharmacologically active doses and sufficiently higher so that an acceptable safety margin can be established.
A major demerit could be the lack of any of the above attributes. If only discovery leads with desirable drug-like properties, which are usually compound-specific and target-specific and need to be defined on a case by case basis, were allowed to enter preclinical development, the ‘losers’ could be identified early and if desired, other leads could be evaluated until a compound with the desired attributes was ‘discovered’.
Third, even the ‘best’ lead selection and optimization process will not uncover all the ‘losers’ and some will enter preclinical development. To maximize the potential of finding and ‘killing’ these ‘losers’ prior to the submission of an IND and the initiation of clinical trials, preclinical assessments, which include subchronic toxicology, safety pharmacology, genotoxicity, animal pharmacokinetics, and drug metabolism and ADME, need to be carefully designed and the results critically evaluated. Generic study protocols are available for each of these study types and can be (and should be since the regulatory agency requirements for a given study type will have already been incorporated) used as templates for the generation of drug candidate specific protocols
Since the majority of ‘losers’ are identified during preclinical development and a goal for reducing the overall time and cost of drug development is to increase the effectiveness of this ‘loser’ identification process, the charge of preclinical groups needs to be to design research studies that will uncover demerits and/or problems that were missed in earlier studies. The evaluation and further characterization of potential concerns (e.g., some CNS activity for a candidate being studied for a cardiovascular indication, extensive hepatic metabolism that may reduce systemic availability, substantial accumulation of drug-related material in non-target organs or tissues, induction or inhibition of enzymes involved in biochemical processes other than the desired pharmacology, toxicity in an organ system or tissue at doses only slightly higher than those required for pharmacologic activity) should also be a primary goal. Once the ‘drug candidate killing’ efforts of preclinical development groups are recognized as time and money saving activities and the members of these groups are appropriately acknowledged for their expertises in identifying ‘losers’, these researchers will employ their knowledge and experience for designing studies that not only meet regulatory agency requirements but also can more fully characterize the attributes and discovery the demerits of preclinical drug candidates.
Fourth, even the appropriate use of preclinical development to ‘kill’ drug candidates will not prevent some ‘losers’ from entering into clinical development. These ‘losers’ need to be discovered as quickly as possible and hopefully before the start of phase 3 studies. Clinical groups who are designing the phase 1 and 2 protocols should critically evaluate the preclinical results and use this information to add tests and evaluations (at appropriate times and with appropriate frequency) in human subjects/patients to as fully as possible explore the safety, pharmacokinetic, and efficacy characteristics of the drug candidate in the target species (i.e., the human). While these earlier clinical studies are on going, nonclinical studies are conducted to evaluate chronic toxicology, reproductive toxicity, carcinogenicity, and tissue distribution and disposition. As with the preclinical assessments, these nonclinical studies should be carefully designed with the goals of extending the knowledge on the safety of the drug candidates and of identifying ‘losers’.
In summary, to reduce the cost of drug development, the goal of drug development companies, particularly the discovery and preclinical/nonclinical groups in those organizations, should be on designing and conducting research studies that identify the development attributes and demerits of discovery leads and drug candidates. Determining that a compound or class of compounds has a non-drug-like characteristic (or more than one) should not be considered a negative but a positive. Modifying the chemical structure of the compound to remove or minimize the undesirable characteristic will produce an analogue with a better chance of successfully completing the drug development process. If structural modifications to generate candidates with drug-like attributes destroy or substantially reduce the pharmacological activity of the compound class, a successful drug candidate from that compound class would, in high probability, not have been possible. Either way the company is a winner, either by identifying a drug candidate with a better chance of success or finding the ‘losers’ before they drain the resources of the company.
Discovery / Preclinical / Nonclinical Estimated Costs
Estimated cost for a single laboratory of one scientist and two associates:
Scientist salary and benefits – $150,000 per year
Associate salary and benefits – $75,000 per year or $150,000 for two associates
Laboratory space and equipment – $150,000 per year
Management support – $50,000 per year
Total per laboratory – $500,000 per year
Four discovery laboratories (two pharmacology and two chemistry) devoting half time to a given project – $1 million per year. Estimated number of years required – 6 years (prior to selection of a preclinical drug candidate and during early preclinical evaluations). Total – $6 million.
Toxicology, drug metabolism and ADME, drug delivery and formulation development, bioanalytical chemistry preclinical/nonclinical laboratories. Four preclinical/nonclinical laboratories devoting half time to a given project – $1 million per year. Estimated number of years required – 6 years (after selection of the preclinical drug candidate until submission of a marketing application to a regulatory authority). Total – $6 million.
Support groups such as basic research (genomics, proteomics, molecular biology), analytical chemistry (structure determination and characterization for novel compounds), central computer group, animal care and maintenance, clinical chemistry and pathology, quality assurance unit, biostatistics, patent and intellectual property, project management, regulatory affairs, etc. Each support group expending $100,000 per year or $1 million per year for 10 support groups. Estimated number of years required – 12 years. Total - $12 million.
