The valuation of part-developed projects in the pharmaceutical sector

Transcription

1 050616ISP The valuation of part-developed projects in the pharmaceutical sector Masters of BioScience Enterprise Sangeeta Puran Fitwilliam College June 2005 Cambridge Healthcare & Biotech Ltd 18 Duxford Road Whittlesford Cambridge CB2 4ND, UK Tel: Fax (Europe): (0) (0) Registered in England number registered office as above

2 Masters of BioScience Enterprise Sangeeta Puran Fitzwilliam College The valuation of part-developed projects in the pharmaceutical sector Sponsored by: Cambridge Healthcare and Biotech Company Supervisor Martyn Postle University Supervisor Jochen Runde Date of Submission 16 th June 2005 Word Count (Excluding all material prior to page 9, figures, tables, references, and appendices) Permission is given for the following people to have access to this dissertation: Available for viewing by personnel bound by Confidentiality only Available to all Cambridge University staff and students Available only after consultation with Student/Company Do you give permission for the dissertation to be made available on the MBE website as a PDF file Do you give permission for a PDF version to be placed on the Intranet, where only members of the Institute of Biotechnology can download it? Student On behalf of the Company Signed by Student: Signed on behalf of Company

3 Preface The study reported in this dissertation was undertaken from the offices and under the supervision of Cambridge Healthcare and Biotech of 18 Duxford Road, Whittlesford, Cambridge, CB2 4ND, UK, from the period of 11 April to 27 May None of the work contained in this dissertation has been submitted for any other degree. This dissertation expressly acknowledges those who have helped with or sponsored this study. The work submitted under the auspices of this dissertation consists of the author s own work. Confirmed by the author Sangeeta Puran 2

4 Acknowledgements relating to this study I would like to thank Martyn Postle from whom the idea of this study originated and to the rest of the Cambridge Healthcare and Biotech team. I am extremely grateful for all your support and for making me feel so at home on Duxford Road. I would also like to thank Jochen Runde, my academic supervisor on this study. To Dr Geraldine Rodgers, thank you for all your hard work in organising the internship programme. I would also like to express my utmost gratitude to all those who participated in this study, including Chandu Ammini, John Aston, Stewart Atkins, Dr Sally Bennett, Sally Brashears, Anthony Gemmell, Alla Grabovska, Enda Gribbon, Paul Hodgkinson, Tom Kassberg, Stuart Kynoch, Alan Lamont, Jonathan Lane, Richard Mason, Richard Parkes, Alex Pasteur, Jonathan Pearce, Dr Cathy Prescott, Stephen Thompson, Richard Wehby, Rupert Winckler and the rest of the participants who could not be mentioned for confidentiality reasons. 3

5 Acknowledgments relating to the MBE programme I would like to thank Professor Chris Lowe, Dr Jim Murray, Dr Geraldine Rodgers and Ms Sabine Deering and the lecturers involved in the programme. You have provided a truly unparalleled learning experience. To my peers on the programme, it has been an honour. I look forward to hearing all about your exciting careers and lives from this moment onwards. Above all, I would like to thank Drazen for making me tea when I needed it most, and patiently waiting for me to finish. 4

6 Abstract This study considers the approach used by biotech business development to value partdeveloped projects (as defined by this study), and the extent to which such differs from the approach of pharma business development, analysts and venture capitalists. A combined quantitative and qualitative approach was employed, using interview (face to face and by telephone) and survey techniques. Our findings indicate the risk-adjusted net present value, discounted cashflow and comparables comprise the methodologies most used by all groups to value part-developed projects. Notwithstanding the application of common methodologies by all groups, differences were found between biotech business development and other groups in valuation proficiency. Differences were also observed between biotech business development and the other groups in respect of the application of specific valuation parameters (e.g. discount rates, forecast periods, expenditure assessment) and the qualitative factors and subjective criteria which its relevant partners and investors considered when determining the worth of part-developed projects. On the basis of these findings, recommendations are made to biotech firms on valuation issues to be considered when dealing with commercial partners and investors. 5

7 Executive Summary This study was conducted from 11 April 2005 to 27 May 2005 in conjunction with Cambridge Healthcare and Biotech. The broad objective of this study was to consider the methods currently used to quantitatively value drug development programmes falling within part-developed projects as defined by this study; to examine the different ways in which these methods are applied and hence understand from the perspective of biotech firms, how different values of a part-developed project can be arrived at. The study involved participants, comprising biotech business development (biotech), pharma business development (pharma), financial analysts (analysts) and venture capitalists (VCs) involved in the valuation of part-developed projects. A combined quantitative and qualitative approach using interview (face to face and by telephone) and survey techniques was employed. Risk-adjusted net present value was the method most used by all groups (including biotech) except for VCs. However, unlike other groups, biotech s preference was driven largely by a lack of familiarity with less conventional methods such as real-options and Monte Carlo simulation. Biotech adopted the same approach (i.e., the bottom-up approach) as the other groups in terms of assessing sales potential. Biotech, together with pharma and VCs, differed from analysts in the period over which cashflows were assessed. In particular, the analyst period did not commence until launch, consistent with the finding that analyst interest primarily lies in projects in advanced clinical development stage. Biotech also differed from pharma insofar as the detail expended in assessing downstream expenditures (e.g., marketing and promotional expenditures). In this regard, pharma hold informational advantages enabling more detailed assessments of such expenditures. In terms of discount rates, biotech used higher rates than pharma and analysts, and lower rates than VCs. Economic parameters (i.e., market size and pricing) posed the greatest sensitivities insofar as NPV modelling by biotech, suggesting greater scope for the application of real-options, which the literature describes as having been specifically developed to address economic sensitivities. Based on the key findings of this study, recommendations are provided to biotech firms in respect of developing proficiency across a wider range of methods and understanding the quantitative parameters and qualitative factors that commercialisation partners and investors are likely to be most influenced by when negotiating deal values. 6

8 Table of Contents Page Chapter 1 Outline Definition of part-developed project Objectives and structure.. 9 Chapter 2 Introduction to valuation methods Discounted cashflows Forecasting cashflows (a) Forecast period 15 (b) Investments.. 16 (c) Revenues Discounting DCF critiqued DCF does not properly account for technical risk Limitations of rnpv DCF does not account for different outcomes Scenario analysis Discount-tree modelling Monte Carlo simulation Limitations DCF does not properly account for the value of managerial flexibility in the face of economic uncertainty Valuation of options Limitations of real-options NPV modelling is theoretical Chapter 3 Methodology Participants Analysis 43 Chapter 4 Findings Methods used to value part-developed projects NPV parameters Revenue Forecast of sales potential Forecast period Investments Forecast of expenditure Discount rate Probability of success Sensitivities. 64 Chapter 5 Conclusion.. 67 References.. 72 Appendix 1 75 Appendix 2 76 Appendix 3 79 Appendix 4 81 Appendix 5 83 Appendix

9 List of Figures and Tables Figures Page Figure 1 A summary of the stages of drug development 10 Figure 2 A project lifecycle curve from a cashflow perspective.. 14 Figure 3 Example of standard sales evolution curves. 20 Figure 4 Example of DCF calculation for Project X.. 24 Figure 5 Example of rnpv calculation for Project X. 27 Figure 6 Outline of decision-tree modelling of the impact of technical and success.. 30 Figure 7 Outline of Monte Carlo simulation.. 32 Figure 8(a) Example of Real-options calculation for Project X: Step 1 Construction of binomial 36 tree.. Figure 8(b) Example of Real-options calculation for Project X: Step 2 Calculation of the value 37 of end states Figure 8(c) Example of Real-options calculation for Project X: Step 3 Calculation of the value 38 of internal states... Figure 9 Distribution of participants by reference to the following groups: biotech, pharma, 42 analyst and VC. Figure 10 Methods used by biotech, pharma, analysts and VCs to value part-developed projects 44 Figure 11 Approach used by biotech, pharma, analysts and VCs to forecast sales potential.. 50 Figure 12 Forecast periods used by biotech, pharma, analyst and VCs in NPV based methods 52 Figure 13 Level of detailed assessment by biotech of expenditures relevant to NPV modelling 56 Figure 14 DRs used by biotech, pharma, analysts and VCs in NPV modelling. 59 Figure 15 The NPV methods (rnpv and DCR) and the discount rates used by biotech. 62 Figure 16 Sensitivities relevant to NPV modelling by biotech. 64 Figure 17 The effects of discount rate changes on capitalised pre-clinical, clinical and total costs 80 per approved new drug. Figure 18 Example of detailed cashflow. 82 Figure 19 Probability of success rates used in NPV modelling. 96 Tables Table 1 Assumptions relating to post-clinical marketing costs.. 17 Table 2 Advantages of NPV based methods methods 91 Table 3 Disadvantages of NPV based. 92 Table 4 Advantages of Comparables 93 Table 5 Disadvantages of Comparables 93 Table 6 Probability of success rates 96 8

10 Chapter 1 Outline Drug development follows a sequence of distinct stages, each of which aims to generate economically valuable specific knowledge about the drug candidate in question (Arojarvi, 2001). In this way, the implementation of drug development 1 programmes generates intellectual assets capable of transfer (Rzakhanov, 2004). Determining the monetary worth of such intellectual assets, though complicated, is central to internal research programme prioritisation, licensing negotiations, investor funding decisions and equity analysis in the pharmaceuticals sector. A range of methods, each with differing computational complexities and limitations, can be used to assign a value to development programmes. However, complex science, long development times, high risk of technical failure and changing regulatory and market conditions, make valuing development programmes a significant challenge. The principal objective of this study is to consider the methods currently used to value development programmes comprising part-developed projects as defined in Section 1.1 below; to examine the different ways in which these methods are applied and hence understand how different values can be arrived at. 1.1 Definition of part-developed project For the purposes of this study, a part-developed project is defined as a development programme that is yet to complete all the following stages of the well defined paradigm of drug development: discovery, pre-clinical, phase I, phase II, phase III, regulatory review and commercial launch. Once the product in development is launched, the programme is taken to fall outside the definition of a part-developed project. Each of the relevant stages is summarised in Figure Drug development in this context is taken to include both drug discovery and development. 2 Figure 1 is based on Charnes et al (2000) and DiMasi et al (2003). A more detailed overview is provided in Appendix 1 9

11 Figure 1 - A summary of the stages of drug development Early stage Clinical trials Discovery Pre-clinical Phase I Phase II Phase III Regulatory review Commercial launch Years Duration difficult to estimate Total population Laboratory studies Laboratory and animal studies 20 to 80 healthy volunteers 100 to 300 patient volunteers 1000 to 3000 patient volunteers Purpose Discover drug candidate Assess safety and efficacy profile Establish safe dosages and assess metabolic effects Test drug candidate in patients. Verify safety and obtain preliminary efficacy data Establish statistically significant efficacy and monitor possible adverse reactions Obtain regulatory approval to supply in territories of interest Launch commercial sales of drug product

12 To this end, the focus of this study is on projects involving the development of prescription pharmaceutical products. Projects involving the development of medical devices and platform technologies are not considered. 1.2 Objectives and structure Most valuation methods stem from the traditional net present value method, commonly referred to as the discounted cashflow (DCF) method. This study considers DCF and the following other methods stemming from DCF, commonly cited 3 as relevant to valuing partdeveloped projects: risk-adjusted net present value method (rnpv); scenario analysis; decision-tree modelling; Monte Carlo simulation; real-options and comparables. The Economist 4 recently reported that biotech firms in 2004 accounted for more new drug approvals from the U.S. Food and Drug Administration than did large pharmaceutical firms, highlighting the importance of part-developed projects sourced from the biotech sector. There are now more than 4,400 biotech firms in the world, the vast majority still privately owned 5. The specific objectives of this study are to consider: The methods used by biotech firms to value part-developed projects; The application of parameters relevant to the predominant methods used by biotech firms to value part-developed projects; and The key differences in the valuation approaches between biotech firms and large pharma firms, financial analysts and venture capitalists. The findings of the study will contribute to the information currently relied upon by investors to assess the value proposition offered by funding drug development programmes, and by senior management of biotech firms to prioritise projects that are most likely to 3 Bennett et al 2004, Borissiouk, 2001, Charnes et al, 2000, Stewart, The Economist June 4 th 2005, p.79 5 Refer to the footnote above.

13 maximise firm value. These findings will also be relevant to understanding the commercial value attributed to part-developed projects in licensing and other commercialisation arrangements with commercial partners, and to financial analysts assessing the implications of part-developed projects on equity prices and public flotations. Following this outline, Chapter 2 introduces the relevant valuation methods. The presentation below commences with an introduction to DCF, followed by a critique of its limitations and the alternate methods that have been developed to overcome such limitations. Chapter 3 describes the methodology of this study. Chapter 4 presents the findings and Chapter 5 concludes by presenting recommendations to biotech firms based on such findings. 12

14 Chapter 2 Introduction to valuation methods DCF Valuation generally involves either a market, cost or income approach. By focusing on the cashflow opportunities of a project, DCF methodology incorporates the income approach. In short, the future revenues and investments 7 associated with the project are modelled into a net cashflow, which is then discounted in accordance with finance theory to derive the net present value of the project Forecasting cashflows Modelling cashflows involves forecasting investments and revenues over a forecast period. Figure 2 presents the typical project lifecycle from a cashflow perspective 8. During the early research stage, project cashflows tend to be negative. Early stage research can take several years, but is not as expensive as the clinical trial stages. The product is launched upon regulatory approval being issued, followed by relatively fast market penetration. A stable period of revenue generation follows. Finally, revenues start to decline as patent expiry approaches, with rapid decline on actual expiry. A consequence of the project lifecycle is that while the term of patent protection usually lasts 20 years, the effective period of monopoly is reduced to the patent term remaining following the issuance of regulatory approval 9. 6 The reader should note that the field of valuation is extensive and the presentation below is not intended to be an exhaustive discussion of valuation theory. 7 Investments in this regard are taken to mean all costs and expenditures associated with drug development. 8 Adapted from Arojarvi (2001) 9 Note that there is significant commentary arguing that the project lifecycle can be managed by evergreening practices (e.g., seeking additional patent protection on the basis of product enhancements to in effect extend patent term). 13

15 Figure 2 - A project lifecycle curve from a cashflow perspective Cashflows Patent period Patent expiry Years Early research Advanced clinical trials Launch Adapted from Arojarvi (2001)

16 (a) Forecast period Financial models can vary on how far into the project lifecycle they forecast. Arojarvi (2001) notes patent expiry is a common endpoint. The launch of generic versions after patent expiry is a key consideration in whether or not to use patent expiry as an endpoint. Recent history contains examples of rapid market share erosion of blockbuster drugs following generic launch. A famous example is the case of Eli Lilly whose US patent for its antidepressant Prozac R, expired in This paved the way for the regulatory approval of Barr Laboratories generic version, Fluoxetine. Prozac R reportedly lost 73% of market share within two weeks of generic launch (Tuttle et al, 2004) 10. The impact of generic competition can vary. Tuttle et al (2004) suggests market share erosion depends on product attributes (e.g., the degree of manufacturing difficulty) 11. Chandon (2004) suggests that a product may operate in a niche category that is too small, or with a brand presence too strong, to attract competition upon patent expiry. In the case of biologics, whether there can be generics that are directly substitutable for patented biologics is yet to be resolved. Bennett et al (2004) also considers the possibility of a product experiencing revenue decline due to competition before patent expiry. The endpoint of a forecast period may also be selected as the point beyond which information required to forecast is unavailable or unreliable (Frei et al, 2004). Selecting an endpoint ultimately remains arbitrary and at such endpoint, the question of residual value arises The following commentary, released at the end of 2004, contains even more recent examples: GlaxoSmithKline and AstraZeneca faced patent expiry dates for some of their biggest selling drugs and the subsequent generic competition has torn apart sales. AstraZeneca s blockbuster stomach ulcer drug Prilosec saw sales drop by 70% in The company lost $2.6 billion in sales to generic competition. Rival Glaxo is facing a difficult 2004 as the brunt of generic erosion of antidepressant Paxil bites. The company admitted 40% of Paxil sales were lost within weeks of the launch of generic competition. [Carter Nield of OrbiMed Advisors, accessed on 22 May 2005 from 11 Tuttle et al (2004) also refer to generic competition being dependent on therapeutic class (e.g., the extent to which there are a significant number of generic players already in a therapeutic class). 12 A terminal value may be calculated to assess such terminal value. Refer to Section below. 15

17 (b) Investments According to DiMasi et al (2003), the average capitalised out of pocket investment for developing a new product to the point of regulatory review is $US 802M 13, which is based on totalling the costs found by the authors to be incurred in pre-clinical development ($US 335M 14 ) and clinical development ($US 467M 15 ). The authors also highlighted potential investment variation depending on the therapeutic indication in question, with increased costs likely to be associated with chronic and degenerative diseases 16. Such variation is driven by the number of patients needed in a clinical trial and by the treatment costs per patient (e.g., outpatient versus intensive care treatment, cost of diagnostic procedures and co-medications, durations of treatment and requirements of follow-up). 13 The figure is provided in 2000 dollars and is based on information provided by firms, which at the time of the study, accounted for 42% of the US pharmaceutical industry research and development. The estimated average out of pocket costs per drug is $US 403M (2000 dollars). Capitalising out of pocket costs to the point of marketing approval at a real discount rate of 11% yields a total pre-approval cost of $US 802 M (2000 dollars). Appendix 2 contains a graph reflecting the effect of changes in discount rates. 14 Pre-clinical costs were taken to include all costs incurred from the commencement of a project up to completion of the pre-clinical stage. The authors found that assessment of pre-clinical costs is made difficult by the fact that pre-clinical costs are commonly incurred as part of wider research and development programme involving multiple projects. However, by construing a pre-clinical to clinical costs ratio, the authors inferred average pre-clinical out of pocket capitalised costs per approved drug of $US 335M. 15 Aggregating across all clinical phases, DiMasi et al (2003) found out of pocket capitalised clinical costs per approved drug to be in the order of $US 467M. Clinical costs include costs relating to trial design, patient recruitment 15, investigator and clinician costs, monitoring costs, data analysis, close out and reporting results, infrastructure costs, and costs related to the production of the clinical trial supplies and animal testing during the clinical period. 16 Bode-Greuel et al (2005) also observed that the costs for clinical development beyond phase I are difficult to estimate because they differ considerably depending on therapeutic indications. 16

18 The costs of regulatory submission 17 aside, the projection of post-clinical marketing costs tend to be based on conventional assumptions such as those outlined in Table 1. Nevertheless, it is necessary to consider the specificities of the target market (Bode- Greuel et al, 2005). For instance, hospital products are characterised by lower marketing costs than products promoted to specialists or primary physicians 18. Table 1 Assumptions relating to post-clinical marketing costs Item Cost of revenue Marketing expense Year 1 after launch Year 2 after launch Years 3 4 after launch Years 5 13 after launch General and administrative expenses Tax rate Working capital Assumption 25.5 % of revenue 100 % of revenue 50 % of revenue 25 % of revenue 20 % of revenue 11 % of revenue 35 % of profit 17 % of revenue Source: Charnes et al (2000) 17 These costs are determined by on a territorial basis. In general, most drugs will aim for U.S. Food and Drug Administration (FDA) approval. The costs of seeking FDA approval are estimated at $US M + $US300,000 for the Prescription Drug User Fee and the remainder for the preparation of the New Drug Application. Preparation costs are highly variable and depend largely on the amount and the quality of data to be presented. (Frei et al, 2004). 18 The increase is driven by the required number of physician contacts. Costs may also depend on the competitiveness of the market, such as in the case of the oncology market, where marketing expenses have considerably increased in the last few years due to increasing numbers of companies becoming active in the field. 17

19 (c) Revenues The discussion of revenue forecasting is based on the approach presented by Porter (1993). Porter s assessment of the potential revenues associated with a project comprises three limbs: size of target market, market share likely to be penetrated by the product in development and subsequent market growth. Starting with the assessment of market size, Porter outlines the bottom-up and topdown approaches. The bottom-up approach 19 calculates market size as follows: Market Size = Number of patients * Number of patients receiving treatment * Price of treatment per patient While incidence and prevalence information can usually be obtained from national health statistics, not all patients are necessarily diagnosed or treated. Price estimation, as will be considered shortly, is also complex. In contrast, the top-down approach 20 extrapolates from existing sales data of product(s) in the same therapeutic class as the product in development. Relevant sales data can be obtained from information services 21 and competitor firm reporting. The second limb of revenue forecasting considers market share. Porter identifies pricing, competition, dosage and formulation, efficacy, sales detailing and patient product loyalty amongst the factors influencing a new product s ability to penetrate the identified market. The distinction between volume market share (based on number of treatments) and value market share (based on sales value) is also relevant to the evaluation of market share. Bennett et al (2004) also notes that the most reliable way to estimate market share is to perform quantitative market research Also referred to as an epidemiological-based approach 20 Also referred to as a market-based approach 21 For example, IMS [refer to 22 The author states that this ascertains the likely demand for the product based on its attributes such as efficacy, dosing, side-effect profile and so on. The market research can range from fairly basic, such as interviewing a handful of clinicians, up to conducting several hundred physician interviews to power what is known as a conjoint model. 18

20 Pricing is a key consideration in revenue forecasting. Earl (2003) provides an illustration of the complexities relevant to pricing. The author begins with the observation that novel products, that are more efficacious than existing products, are typically priced at a premium. However, the number of participants who will switch to a more expensive product also needs to be determined. Further, during the forecast period, other products may lose patent protection and become subject to generic competition. Patient switch again needs to be considered. The preceding is by no means exhaustive of the factors relevant to pricing (e.g., reimbursement policies are often cited in pricing discussions). The third limb of revenue forecasting considers market growth. The current market growth will only be a guide to future growth prospects. The factors behind market growth need to be identified and again, the distinction between sales volume growth and sales value growth is relevant. Sales volume growth will be affected by changes in population growth, spread of an illness, frequency of occurrence, frequency of diagnosis, and treatment practice. Sales value growth will depend on changes in pricing and product mix (older products may have significantly lower prices than newer, more efficacious ones). 19

21 Falling outside of Porter s presentation is the use of standard sales evolution curves to forecast revenues. By looking at historical peak sales of drug products, the rate of rampup to peak sales and the rate of market erosion, different sales evolution curves have been developed. Figure 3 illustrates a sample of such curves. An obvious limitation of such approach is that the sales lifecycle of the product in development may not fit within the limited standard curve options offered. Figure 3 Example of standard sales evolution curves R Sales US$ million S1 F2 S2 T 100 F Year from launch F1: fast ramp fast erosion F2: fast ramp medium erosion S1: slow ramp fast erosion S2: Slow ramp medium erosion R: Tail product, growth T: Tail product, stable Source: Lehman Brothers Figure is provided as part of an internal confidential document. Accordingly, the relevant document containing such figure is not referenced in the Reference section. 20

22 2.1.2 Discounting An amount of money received today is worth more than the same nominal amount of money received in the future. Conversely, money received tomorrow is worth less than a dollar received today. Application of this principle to forecasted cashflows means that not only are future revenues worth less today than in the future, but also future investments will cost less today. Finance theory requires that a discount rate (DR) be used to translate the future cashflows of a project to today s value. From the perspective of investors financing a drug development project, the DR reflects the opportunity cost of capital. From the perspective of the entrepreneurs seeking to fund such project, the DR reflects the cost of capital. The textbook computations of the DR to be applied to a project (r Project ) use the capital asset pricing model (CAPM), which prescribes the following: r Project = r F + U Project * (r M - r F ) Where: r F is the risk free rate; U Project is the beta value of the project; and r M -r F is the difference in the expected return on the market and the risk-free rates. r F represents investor requirements that a project generate at least the same return as would be expected from investing in risk-free investments 24. CAPM also assumes that a risk premium (i.e., r M - r F ) is requested by investors for accepting the risk to invest in assets whose value is highly volatile. The risk premium in the case of the project in question is given an appropriate weight through U 25 Project. Although CAPM is not the only way of determining DRs 26, it is a framework allowing the DR of a project to be derived in an objective and replicable way. Notably, Randerson (2001) observed that none of the methods for calculating DRs would generate the DRs 24 In practice, this tends to be based on short term treasury rates. 25 The beta is company specific and describes the co-variance of the company s equity with the market 26 The weighted average cost of capital (WACC) is also cited in discussions of DR assessment. This model typically deals with firm DRs. If the firm has issued debt, an extension to CAPM is applied, reflecting the usually lower requirement of debt. Please refer to Bode-Greuel et al (2005) for further discussion. 21

23 (in the 30% - 70% 27 range) typically advocated by venture capitalists (VCs). VCs tend to use DRs that represent the internal rate of return (IRR) expected by their fund investors. While the IRR does not qualify as a DR in the strict textbook sense, their use by VCs is well established. The reasons provided by VCs for applying such high DRs include the high risk associated with drug development investments, the low liquidity of the underlying intellectual asset, and compensation for management services provided by VCs (Dittman et al, 2000). Once a DR has been identified, the present value of each time point of the net cashflow is calculated as follows: PV t = C t /(1 + r Project ) t Where: C t is the net cashflow at period t (usually yearly intervals). If residual value is considered to subsist in a project beyond the forecast period, a terminal value for the project can be calculated as follows 28 : Terminal value project = C EP / r Project (1 + r Project ) tep Where: C EP is the cashflow as valued at the period representing the endpoint of the forecast; and tep is the period at such endpoint. 27 Average DRs applied by VCs as published by the Harvard Business School: Start up: 33-46% First stage: 26-39% Second and third stage: 23-33% Fourth stage: 20-26% IPO: 16-23% Average DRs required by VCs in the UK as published by the London School of Business: Seed: Start up/early stage: Expansion/later stage: 20-34% Buy-out: 17-23% (Randerson, 2001) 28 Multiples methodology may also be used to calculate a terminal value. Given the limited scope of the present discussions, this methodology will not be considered here, however, relevant discussions can be found in most finance texts. 22

24 Finally, the net present value (NPV) of the project is the sum of the present values at each time point and, where relevant, the terminal value i.e.,: NPV project = C o + C 1 /(1 + r Project ) 1 + C 2 /(1 + r Project ) C EP / (1 + r Project ) tep + C EP /r Project (1 + r Project ) tep A positive NPV is usually taken to indicate that a project is likely to create value and is worth funding (with preference for higher NPV projects), with a negative NPV usually indicating the converse. As indicated above, the field of valuation is computation intensive. To avoid undertaking the present discussion completely in a theoretical vacuum, examples are provided. Figure 4 considers the application of DCF to a published example, which will hereafter be referred to as Project X (Villiger et al, 2005). 23

25 Figure 4 Example of DCF calculation for Project X The following example presents a very basic cashflow forecasted on a yearly basis for a pharmaceutical project right before the start of phase II, which is scheduled to last two years. The model assumes that the phase III trial will take three years, after which an application for regulatory approval can be submitted, with approval to follow after one year. Typically, a far greater level of detail will be presented. Elements such as depreciation, amortisation, tax and debt considerations can also be addressed 29. Note also that Project X is presented as a self conducted project. If, however, commercialisation of the project is to be by out-licensing, assumptions, for instance, relevant to licence fees, milestone payments and royalty rates further refine the projected revenues. The focus of the present discussion is on projected revenues attributable to the project as a whole. Accordingly, the impact of commercialisation strategies will not be considered. Following the computation of a net cashflow stream, a DR of 30% is used to calculate a net present value for Project X. In this regard, it is not unusual for the DR to be presented as part of a discount factor. No terminal value has been calculated. Forecast investments Investments ($US M) Ph II -15 Ph III -70 FDA -3 Launch -210 Forecast revenues Revenues Calculate net cashflow Net cashflow Forecast period Years from now Discount factor (1/(1 + r Project) t ) Discount factor M 29 Appendix 3 contains an example of a more detailed cashflow model. Net present value 24

26 2.2 DCF critiqued DCF has many advantages, including simplicity in application and wide acceptance 30. Villiger et al (2005) argue that DCF articulates each element of value in a form that can be adjusted to determine the impact of alternative views on key assumptions 31. Criticism of DCF however is equally prominent. The rest of this chapter will consider some of the commonly cited limitations of DCF, and the valuation methods that have been developed to address such limitations. 2.3 DCF does not properly account for technical risks The use of DRs in DCF methodology to simultaneously adjust for time and technical risks has been shown to penalise long term projects relative to short term projects (Randerson, 2001) 32. The risk-adjusted NPV method (rnpv) was developed specifically to deal with technical risk. More precisely, it takes technical risk 33 outside DRs, instead accounting for such risk by adjusting cashflows of each stage of development by probability rates based on the technical risks associated with such stage 34. In turn, a lower DR is also used, with the literature 30 Generally, DCF methodology is said to be used by virtually every major corporation doing a variety of business tasks, including capital investment decisions, operations analysis and business reporting. It is used in business in various forms under the names such as free cashflow, capital cashflow, shareholder value analysis, flows to equity and internal rate of return. In each form, minor technical differences exist; but the fundamental principles of the analysis remain the same and are consistent with the NPV principles as outlined above. The method is also used routinely by major management consulting firms, investment banks and government policy makers. It is taught in universities all over the world. Indeed the method is so central that some textbooks treat value as synonymous with discounted cashflows. The method has found considerable support in legal proceedings where it has been used widely to measure the value of lost opportunities. 31 Thus if one assumption or finding comes into question, the calculation can be modified with relative ease. 32 In this regard, the high rates applied by VCs are considered to invariably render research and development programmes to a negative value. 33 Also referred to as technological or scientific risk 34Example of industry average rates of success (phase to market) published by Bennett et al (2004): Pre-clinical -3%, Phase I 6%, Phase II 18%, Phase III 42%, Regulatory review 90%, Launch -100%. 25

27 publishing rates in the range of 9-15% (Randerson, 2001). Figure 5 applies rnpv to Project X, replicating an example recently published by Villeger et al (2005) The author has simply copied the example published by Villiger et al (2005) and has not undertaken an analysis of the correctness or otherwise of the calculations. It is beyond the current scope to undertake such analysis however, the principles employed in the relevant example mirror the principles employed by Borrissiouk et al (2001). An Excel sheet of the example can be found on 26

28 Figure 5 Example of rnpv calculation for Project X This refers to the probability of the project reaching launch from the start of the relevant stage. Standard rates can be sourced from literature, with noted sources including those published by The Tufts Centre for the Study of Drug Development and The Centre of Medical Research. In the case of Project X, at the start of phase II, there is a 50% probability of reaching launch. Project X is just about to enter into phase II, hence there is a 100% probability of achieving the relevant success rate state ( i.e., 50%). The net cashflow of each stage is adjusted by the corresponding probability to achieve state. Investments ($M) Revenues phase II phase III FDA -3 0 launch sum In the case of each stage following the initial stage, the probability to achieve state at the start of the stage is cumulative of the success rate and probability to achieve state of the preceding state. Net cashflow Success rate % % -3 90% -280 By way of example, in the case of phase III of Project X, the probability to achieve is calculated by: Success rate (phase II) *Probability to achieve state (phase II) 50% *100% = 50% The net cashflow of phase III is then adjusted by 50%. Probability to achieve state Years from now Discount factor 100% 0 100% % 2 83% % 5 62% % 6 56% A discount rate of 10% is used. The risk-free rate is not applied because there are risks in the development process in addition to technical risks. Risk adjusted net present value

29 2.3.1 Limitations of rnpv Bennett et al (2004) observes a continued practice of applying high DRs to rnpv modelling, indicating rnpv is susceptible to misuse. The calculation of probability rates remains problematic, particularly in respect of pre-clinical stages. Many unsuccessful pre-clinical projects are quietly discontinued. Additionally, probability rates tend to be presented as industry averages. The challenge of applying such rates to a specific therapeutic indication is well documented. For instance, Bennett et al (2004) explains that where the drug mechanism is well understood (such as in hypertension, diabetes or asthma) the relevant probabilities of technical success are likely to be higher than industry averages. Similarly, projects dealing with lesser understood diseases (such as cancer) may be associated with lower probabilities of technical success DCF does not account for different outcomes DCF valuation is in effect based on a single projection of inputs (Remer et al, 2001). The relevant inputs are, however, impossible to calculate with any certainty. Sensitivity analysis can be undertaken by flexing individual inputs in isolation (Bennett et al, 2004). By comparison, scenario analysis, decision-tree modelling and Monte Carlo simulation seek to deliver a range of NPVs based on likely variations to more than one input. 36 Randerson (2001) refers to the likelihood of rates between biologics and new chemical entities differing. 28

30 2.4.1 Scenario analysis Scenario analysis models the outcomes of events (i.e., scenarios) on NPV. For instance, Figure 5 is based on a revenue projection of $US 490 M. However, this is only one possible revenue scenario. Clearly, the product may attain different revenues 37. The relevant scenarios, based on the probabilities of such scenarios eventuating, can be modelled to examine effects on NPV Decision-tree modelling Decision-tree modelling considers the impact on project value of different scenarios at nominated decision points along the development path. Typically, decision points occur at the completion of each stage relevant to the development path. Figure 6 outlines decisiontree modelling to consider the impact of technical failure or success on project NPV. In this way, the relevant impact on value can also be pictorially represented, together with relevant pay-offs if the project is abandoned at any decision point in the event of technical failure. 37 In certain cases, the likely outcomes or scenarios have to be estimated as in the discussed example, while in others they are binary (such as product failure or success in clinical trials). Bennett et al (2004) outlines the following examples of scenarios that can be readily modeled: the outcome of a product failure or success, the entry of a new drug in the clinic, the entry of existing drugs into new indications, additional sales potential, the outcome of fundraising and the outcome of a royalty dispute. 29

31 Figure 6 Outline of decision-tree modelling of the impact of technical and success NPV of proceeding to launch NPV of proceeding to regulatory submission Succeeds 90% launch NPV of proceeding to phase III Begin phase II Succeeds 50% Fails 50% phase III Succeeds 60% Abandon regulatory submission Fails 40% Fails 10% Abandon Abandon

32 2.4.3 Monte Carlo Monte Carlo methodology simulates adjustments to multiple inputs (e.g., market size, expenditures, pricing, time to market) to produce an overall distribution of possible outcomes. This is achieved, in the first instance, by defining the statistical probability distribution of each uncertain input of interest. The type of distribution will depend upon the conditions surrounding that input. Software simulation is then used to repeatedly sample values from the probability distributions of each input. Each simulation generates a single NPV estimate. The end result of the repeated simulation is a range of possible NPVs and their respective probabilities of occurrence 38. Figure 7 outlines Monte Carlo simulation. 38 Another advantage is that the method allows the use of a lower DR because the uncertainties (failure possibilities) are incorporated into the valuation model. 31

33 Figure 7 Outline of Monte Carlo simulation Example of uncertain inputs Market size Probability Expenditures Probability Probability Pricing Simulation Repeated sampling from the probability distributions of each input of uncertainty Probability NPV project Time to market Probability

34 2.4.4 Limitations The methods essentially supplement DCF and rnpv, and to that extent propagate a number of the limitations associated with such methods. The value derived also depends on the choice of scenarios and the associated probabilities of occurrence, which in large part are determined subjectively. Further, although the methods are useful for assessing the spread of NPVs for a project depending on the various outcomes, they still do not assist in yielding a more reliable single NPV. 2.5 DCF does not properly account for the value of managerial flexibility in the face of economic uncertainty 39 In addition to the earlier mentioned technical risks, part-developed projects face economic uncertainty. Remer et al (2001) identify economic uncertainties that affect all projects (e.g., in respect of the size and scope of the market, the regulatory issues, intellectual property rights regimes, competition through generic drugs) and project specific uncertainties (e.g., lack of organisational and financial resources). Real-options methodology aims to address the impact of economic uncertainties on project value by applying financial options theory to the drug development process. The notion of value subsisting in a financial option is illustrated by the simple hypothetical presented in the footnote 40. By way of a basic overview, an option grants the option-holder a right (as opposed to an obligation) to exercise certain rights in respect of an asset. The relevant option derives its value from this right. The parameters relevant to financial options analyses 39 Market uncertainty is also referred to as economic uncertainty. 40 Consider a firm that has a present value of $100 M for operating an existing line of business into the future. There is the possibility that the firm can extend into a new opportunity with a present value of $50M if favourable conditions develop. Suppose that the likelihood of such conditions developing is 50%. According to Frei et al (2004), a business with this option is worth more than the $100M because rational investors will place a present value of $25M on this option. While the future outcome is uncertain either $50M or $0M the present value of $25M on this option is not uncertain. Investors will pay $25M today for a 50% risk neutral probability to receive a present value of $50M from the project. Once the value of the optional project is known by the market, the stock of this business would rise to $125M to reflect the full value of the firm. If the uncertainty over the project is resolved, the value of the business will change accordingly: either by rising to $150M if the favourable conditions materialise or by dropping to $100M if they do not. 33

35 include the price payable to be the option-holder (i.e., the exercise price) and the underlying asset to which the option provides rights to. Real-options theory models the drug development process as embedding a series of options 41 in the face of unpredictable economic developments. For example, Borissiouk et al (2001) explain that once a project has passed phase I, the option-holder has the option to invest in phase II. The start of phase II will require an investment outlay, which is the exercise price for that option. If the option-holder decides to invest, it will acquire the option to invest in phase III, together with an option on the future commercialisation of the project 42. However it may be that technical success in phase III is accompanied by unfavourable market conditions. The option-holder in such circumstances may abandon the project. Rational investors are assumed to be willing to place a value on such options 43. Consequently, under real-options methodology, the value of the project is linked not only to its cashflows but also to the presence of options Valuation of options Villiger et al (2005) explain, while there are alternative methods to value options 44 embedded in a part-developed project, the binomial tree 45 is accessible even to managers without math skills. The construction of a binomial tree is still formula driven and includes steps beyond succinct summation here. In the interests of simplicity, real-options valuation by binomial tree construction is considered in series in Figures 8(a) to 8(c) by replicating the Project X 41 The existing literature provides six categories of real-options based upon the types of managerial flexibility: option to defer, option to expand or contract, option to abandon or license, option to switch, compound option, and option to grow. 42 This option tends to be referred to as a compound option. 43 The option to wait to invest until relevant economic uncertainty resolves itself is also referred to as a learning option. Projects can be modelled to embed several other options of choice, including the options of expansion, deferral and licensing. 44 Villiger et al (2005) observe that the valuation of options is a daily business and there exist various methods for the valuation. Nevertheless, not all are suitable for the real option valuation. The often mentioned Black- Scholes formula, the standard valuation method for financial options, cannot be used in the context of realoptions, because of its inability to handle staged investments. Some alternative methods like finite difference are complicated, requiring sophisticated mathematics and computing. 45 The binomial model is also referred to as the true discrete binomial lattice option pricing model or the lattice algorithm methodology. 34