1 1-29
`Phormccosconomlcs 2602: 20 Suppl. 3:
`1 l70-7690/02/0C03—001 l/S25:00/0
`
`ORIGINAL RESEARCH ARTICLE
`_
`'
`© AdLs Infernuflonol lelted. All Ilghts reserved.
`
`Returns on Research and Development
`for 1990s New Drug Introductions
`
`Henry Grabowski,1]ohn Vernon1 and Joseph A. Dz'Mczsi2
`
`1 Department of Economics, Duke University, Durham, North Carolina, USA
`2 Tufts Center for the Study of Drug Development, Tufts University, Boston, USA
`
`Abstract
`
`Background: Previously published research by the authors found that returns on
`research and development (R&D) for drugs introduced into the US mar—
`ket in the 1970s and 1980s were highly skewed and that the top decile of
`new drugs accounted for close to half the overall market value. In the 19905,
`however, the R&D environment for new medicines underwent a number of
`changes including the following:
`the rapid growth of managed-care or-
`ganisations; indications that R&D costs were rising at a rate faster than that of
`overall inflation; new market strategies of major firms aimed at simultaneous
`launches across world markets; and the increased attention focused on the
`pharmaceutical industry in the political arena.
`
`Objective: The aim of this study was to examine the worldwide returns on R&D
`for drugs introduced into the US market in the first half of the 1990s, given that
`there have been significant changes to the R&D environment for new medicines
`over the past decade or 50.
`
`Results: Analysis of new drugs entering the market from 1990 to 1994 resulted
`in findings similar to those of the earlier research — pharmaceutical R&D is
`characterised by a highly skewed distribution of returns and a mean industry
`internal rate of return modestly in excess of the cost of capital.
`
`Conclusions: Although the distribution of returns on R&D for new drugs con—
`tinues to be highly skewed, the analysis reveals that a number of dynamic forces
`are currently at wOrk in the industry. In particular, R&D costs as well as new drug
`introductions, sales and contribution margins increased significantly compared
`with their 19805 values.
`'
`
`Competition in the research—based pharma—
`ceutical industry centres on the introduction of new
`drug therapies. In this paper, we examine the re—
`turns on research and development (R&D) for new
`drug entities introduced into the US market in
`the first half of the 1990s. This research work
`
`builds directly on earlier analyses of returns on
`R&D for the 19705 and 19805 introductions per—
`formed by Grabowski and Vemonll’zl
`
`Our prior analyses indicate that this industry has
`exhibited very skewed distributionsrof returns. In
`this regard, several significant new classes of drug ,
`therapies have been introduced since the late
`19705. Early movers in these classes have obtained
`the highest returns on R&D. We found that the top
`decile of new drugs accounted for close to half of the
`overall market value associated with all the new drug
`introductions in our 19705 and 19805’ samples.
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 1 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 1 of 19
`
`

`

`12
`Grubowski et ul.
`
`
`The results of our prior analysis are also consis-
`tent with an economic model of rivalrous R&D
`
`competition. In particular, the promise of above-
`average expected returns produces rapid increases
`in industry R&D expenditures, as firms compete to
`exploit these opportunities until the returns be—
`come unattractive. From an industry perspective,
`our results indicate that mean returns on R&D are
`
`relatively close in value to the risk—adjusted cost of
`capital for drug industry investments. This rent—
`seeking model is also supported by a recent empir—
`ical analysis by Scherer, who finds a strong rela—
`tionship between industry R&D outlays and profits
`over the period 1962 to 1996.91
`An investigation into the drug returns in the
`1990s is timely on a number of grounds. First, this
`decade has been characteriSed by the rapid growth
`of managed-care organisations on the demand side
`of the market for pharmaceuticals?” This has led
`to greater access to and utilisation of pharmaceuti-
`cals, but also greater generic competition in the
`poSt-patent period. Second, a new study of R&D
`costs by DiMasi and colleagues indicates that the
`R&D costs for new drugs have continued to rise
`much faster than the rate of general inflation.[5]
`This reflects, among other factors, the increased
`size of clinical trials compared with those for ear—
`lier new drug introductions. Third, many firms are
`changing their market strategies and attempting to
`launch their products simultaneously across world
`markets, reflecting the higher R&D investment
`costs and more intensive competition from new
`molecules in the same product class.
`In addition to these economic developments,
`the industry continues to be the subject of consid—
`erable attention by policy makers. Recent policy
`initiatives in the US include a Medicare prescrip-
`tion drug benefit, the parallel importation of drugs
`from Canada and Mexico, and various state pro—
`grammes affecting drug costs and utilisation by the
`poor and elderly populations. The potential effects
`of these policy initiatives on R&D returns remain
`an important issue for research. Our past work on
`R&D returns has provided a framework for the
`Congressional Budget Office and other groups to
`
`consider the effects on R&D of the proposed
`Clinton Health Care Reform Act and the Waxman—
`Hatch Act of 1984497]
`
`In the next section of this paper, we describe the
`data samples and methodology for our analysis of
`the returns to 1990 to 1994 new chemical entities
`
`(NCEs). ‘Empirical Results’ presents the empirical
`findings on the distribution of returns and a sensitiv—
`' ity analysis involving the main economic parameters.
`‘Drug Innovation and Industry Evolution Since 1970’
`provides a discussion of the results and compari—
`sons with the historical findings from our prior work,
`which is based on the same methodology. The final
`section provides a brief summary and conclusions.
`
`Methodology and Data Inputs
`
`Overview
`
`This section explains the methodology and key
`data inputs used in estimating the returns to 1990
`to 1994 NCEs. Our sample includes ‘large—mole-
`cule’ biologics, in addition to traditional ‘small
`molecule’ chemical drugs. A detailed discussion of
`the general methodology is provided in our earlier
`papers on R&D returns.“’2] Our focus here is on
`the similarities and differences of the 1990s sample
`compared with our analysis of prior NCE cohorts.
`The basic sample comprises 118 NCEs intro—
`duced into the US between 1990 and 1994. This is a
`
`comprehensive sample of the NCEs originating from
`and developed by the pharmaceutical industry that
`were introduced into the US in the 1990 to 1994 time
`period. However, three drugs were omitted from our
`sample because they failed to appear in any year in
`the IMS sales data audits. These drugs were distrib—
`uted outside of normal sales channels and were
`
`likely to have nonrepresentative R&D costs be-
`cause of their special indications.
`.
`The number of NCE introductions increased
`
`significantly in the early 1990s compared with the
`1980s. The corresponding 1980 to 1984 sample
`was 64 NCEs. This increase in NCEs reflects the
`
`increased R&D expenditures for new entities by
`the traditional pharmaceutical industry as well as
`the growth of the independent biopharmaceutical
`
`© Adls International LImIted. All rlghts reserved.
`
`Phormocoeconomlcs 2002; 20 Suppl. 3
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 2 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 2 of 19
`
`

`

`13
`R&D Returns for 19905 New Drug Introductions
`WWWMWMWW
`
`industryls] The latter industry was in its infancy
`in the early 1980s, but by the early 1990s it had
`become a significant source of new drug introduc—
`tions. There is also a significant increase in the
`number of new drugs approved for orphan drug
`indications. As we have discussed elsewhere, there
`
`is a_high degree of overlap between the biopharma—
`ceutical and orphan drug sub—samples!“
`Our basic procedure is as follows: for each new
`- drug in our sample, worldwide sales profiles are
`constructed over the drug’s product life cycle.
`These sales values are converted to after—tax prof—
`its and cash—flow values using industry data on
`profit margins and other economic parameters.
`These data are combined 'with R&D investment in-
`
`formation, based on the recent analysis by DiMasi et
`al.[.5] Mean net present values (NPVs) and internal
`rate of return (IRRs) are then computed for this port-
`folio of new drug introductions. The distribution
`of returns is another major focus of our analysis.
`
`Cost of Capital
`
`In our earlier analysis of 1980 NCEs, we
`.
`utilised a 10.5% real cost of capital for the phar-'
`maceutical firms. This was based on an analysis of
`the industry using the capital asset pricing model
`(CAPM)
`that was performed by Myers and
`Shyum-Sunder.[9] Their study was commissioned by
`the Office of Technology ASsessment as part of a
`larger study on R&D costs, risk and rewards.[101
`They found that the real after—tax cost of capital on
`equity plus debt varied between 10 and 11% during
`the 1980s.
`.
`
`For our sample of 1990 to 1994 introductions,
`the relevant investment period spans the mid-
`1980s through the late 1990s. In their original ar—
`ticle, Myers and Shyum—Sunder provided esti-
`mates of the cost of capital for 1985 and 1990.
`Myers and Howe have subsequently provided a
`related analysis for 1994““ We also performed a '
`comparable CAPM for analysis for January 2000.
`The results of these CAPM-based studies are
`summarised in DiMasi et al.[51'
`
`Using these four CAPM—based analyses, occur—
`ring at roughly 5-year intervals, we found that the
`
`mean cost of capital for pharmaceuticals over this
`period was just over 11%. Consequently, 11% was
`selected as the baseline value for the cost of capital
`in this analysis of 1990 NCEs. This represents a
`small increase from the 10.5% cost of capital
`utilised for the 1980 NCEs.
`
`As Myers and Shyum—Sunder indicated in their
`original article, the CAPM approach provides
`somewhat conservative cost-of—capital values with
`respect to investment in new prescription drugs.
`One reason is that the equity market data on which
`the CAPM analysis is based pertain to all the dif—
`ferent functional areas and commercial activities
`
`of drug firms (which can include over—the-counter
`drugs, animal health, basic chemicals, etc.). An—
`other reason why the cost of capital may be under—
`stated is the fact that many pharmaceutical firms
`carry significant cash balances. Indeed, Myers and
`Shyum-Sunder found that many pharmaceutical
`firms have large positive cash balances and are ac-
`tually net lenders rather than net borrowers. Con-
`sequently, these firms have a negative debt ratio.
`Myers and Shyum—Sunder did a sensitivity analy-
`sis to gauge how this factor would affect their 1990
`value and they found it causes the nominal (and
`real cost) of capital to increase by almOst a full
`percentage point.[9]
`,
`Several surveys have been performed of the
`hurdle rates used by US companies. A general
`finding is that hurdle rates are typically greater
`than the weighted cost of capital computed by a
`CAPM analysislm One of the authors undertook
`an informal survey of six pharmaceutical firms in
`mid—2001' with respect to the hurdle rates that drug
`firms utilise in their R&D investment decisions.
`
`The survey of these firms yielded (nominal) hurdle
`rates from 13.5% to over 20%. If one takes 3% as
`
`the'long—run expected rate of inflation, then an
`11% real rate of return corresponds to a nominal
`rate of 14%. This 14% rate is within the range of
`hurdle rates utilised by the drug firms in their R&D
`investment decisions, but it is at the lower end of
`
`the range. This is consistent with the view that a
`CAPM analysis provides conservative estimates
`on the industry’s .cost of capital.
`
`© Adis lm‘ernotlonol lelted. All rlghts reserved.
`
`Phormocoeconomlcs 2002; 20 Suppl. 3
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 3 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 3 of 19
`
`

`

`Grabowskz' et al.
`14
`WWWWW
`
`Myers and Howe further indicate that the R&D
`decision process can be modelled as a compound
`option pricing model.[“] Under this model, at any
`point in the R&D decision-making process, future
`R&D serves as a form of leverage, or debt, assum—
`ing the firm decides to undertake further develop—
`ment and marketing. Since this ‘debt’ or leverage
`declines over the subsequent stages of the R&D
`process, so will the firm’s cost of capital. Imple—
`mentation of this model requires unobservable
`informational inputs compared with the standard
`CAPM approachvusing a weighted cost of capital.
`DiMasi et al.[51 performed a sensitivity analysis us-
`ing this option value approach, and showed that for
`reasonable values of the forward looking discount
`rates, the CAPM and option value models yield
`comparable results.
`
`Research and Development (R&D)
`investment Expenditures
`
`To obtain representative R&D investment ex-
`penditures for the new drug entities in our sample,
`we relied on the recently completed study by
`DiMasi 'et al.[5] This study obtained R&D cost‘
`data-for a randomly constructed sample of 68
`drugs first tested clinically between 1983 and
`1994. The DiMasi study is designed to measure
`the average cost of a new drug introduction and
`includes discovery costs as well as the costs as-
`sociated with failed candidates.
`
`The mean introduction of our sample NCEs is
`1992 while the mean introduction of drug candi-
`dates analysed in the DiMasi study is 1997. DiMasi
`and colleagues had previously undertaken an anal-
`ysis of the costs of 1980s introductions using
`the same methodology employed in their new
`study.“31 That study was centred on 1984. Given
`the availability of these two R&D cost studies
`centred around 1984 and 1997, we can utilise a
`
`linear extrapolation procedure to estimate the mean
`R&D costs for our sample cohort.l
`
`1 Since our sample is centred around 1992, we utilise the
`following linear extrapolation equation to derive R&D costs:
`R&Dgz = R&Dg4 + (8/13) R&D97.
`
`Using this extrapolation procedure, we esti—
`mated the mean out—of—pocket R&D expenditures for
`the drugs in our sample to be $US308.4 million. This
`is approximately double the estimated R&D expen—
`ditures (in $US, 2000 values) for the 1980 to 1984
`samples of NCEs. DiMasi also estimated a repre-
`sentative investment period of 12 years from initial
`drug synthesis to Food and Drug Administration
`(FDA) approval. We were able to allocate the out-
`. of-pocket R&D costs over this 12-year period us—
`ing weights derived from the DiMasi et a1. study.[5]
`Capitalising these costs to the date of marketing,
`at a real cost of capital of 11%, yields $US613
`million as the aVerage (pre-tax) capitalised R&D
`investment per 1990 to 1994 NCE introduction.
`Our analysis is performed on an after-tax basis.
`For-the time period under study,-we estimated a
`30% average effective tax rate for the pharmaceu—
`tical industry (see ‘Effective Tax Rates’). Since
`R&D expenditures can be expensed for tax pur-
`poses, we multiplied the pre-tax values by 0.7 to
`get an after-tax value. This is shown in the first row
`of table I. Utilising the 30% effective tax rate,
`$US613 million pre-tax capitalised corresponds to
`an after-tax value of $US429 million.
`
`In addition to these pre-launch R&D expendi—
`tures, firms also undertake R&D outlays in the
`post—approval period for product extensions such
`as new indications, formulations and dosage lev-
`els. Since these activities can be viewed as spillo—
`vers from the original NCE introduction, these on—
`going R&D investment expenditures, as well as
`any extra revenues that they generate, are appro-
`priately incorporated into the analysis. On the
`basis of the DiMasi et al. study,[51 we estimated
`
`Table I. Capitalised research and development (R&D) costs for the
`mean new chemical entity in the 199016 1994 sample
`R&D costs
`Pre-tax
`Alter tax
`($US millions; 2000 values)°vb
`$429
`$613
`Discovery and‘development
`$51
`$73
`Product extensions after launch
`$480
`$686
`Total
`a R&D costs include expenditures on product failures as well as
`successes.
`
`_
`
`b R&D costs are capitalised to the first year of marketing using
`an 11% cost of capital.
`
`© Adls International Limited. All rights reserved.
`
`Pharmacoeconomlcs 2002: 20 Suppl. 3
`
`WATSON LABORATORIES, INc.., IPR2017-01622, Ex. 1113, p. 4 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 4 of 19
`
`

`

`15
`R&D Returns for 19905 New Drug Introductions
`WW
`
`the average post-approval R&D costs per NCE in
`our sample period to be $USlO7 million (before
`tax).2 We allocated these costs equally over the
`first 8 years of an NCE’s market life, using a dis-
`count rate of 11% from the date of marketing. This
`yields a present value of $US73 million (before
`tax) .and $USSl million dollars (after tax).
`When the after—tax values (see column two of
`table I) are added, the mean capitalised value for
`both pre— and post—approval R&D for the drugs in
`our sample is estimated to be $US480 million. This
`is the baseline value that-we compare with the pres—
`ent value of net revenues for the mean NCE in our
`sample.
`0
`
`Global Sales
`
`In our prior analysis, we obtained US sales data
`on each NCE in the sample. We then estimated
`worldwide sales for these compounds using a
`worldwide sales multiplier common to all NCEs.
`One limitation of this .approach is that the ratio
`of worldwide sales to domestic sales varies signif—
`icantly, both over time and across drugs in our
`sample.
`In the current analysis, our approach was to ob-
`tain worldwide sales data directly on as large a
`group of the drugs as possible. We were generally
`successful in this endeavour, in that we were able
`to obtain worldwide sales data for a majority of the
`NCEs in our sample (66 NCEs) using several com—
`plementary data sources. These 66‘ drugs ac—
`counted for more than 90% of total , US sales
`
`realised by our sample of NCEs ’and presumably a
`similar, or even larger, share of its realised world—
`wide sales. With respect to the latter point, there is
`evidence that the larger selling US drugs diffuse
`across more countries and have larger sales glob-
`
`ally than US compounds with smaller domestic
`sales.l14]
`To obtain worldwide sales data, we collected
`
`sales data that firms provide in their annual reports,
`in the reports of financial analysts, and in publi—
`cations such as MedAdNews. The last-mentioned
`
`source has compiled an annual survey of world-
`wide drug sales, by product, since 1990 on an ex—
`panding basis over time. The compilation for 2000
`includes information on the 500 top—selling pre—
`scription drugs worldwide.”5]
`.
`A complementary sourCe of data that we also
`relied on was IMS data on worldwide sales, which
`is based on audit data sources from'a large number
`of countries. The IMS data source was available to
`
`us (from a prior project) for a sub—sample of drugs
`‘ consisting of the largest sellin'g global drugs in our
`sample. It provided a check on the sales informa—
`tion provided by the company sources. In most
`cases, the IMS sales values were less than the com-
`
`pany values. This reflected the fact that the IMS
`does not capture all the sales channels available
`across countries, while the company data do in—
`clude every channel.
`In about 25% of the overlapping observations,
`however, the IMS sales were greater than the com—
`pany—reported values. An analysis into why this
`was the case revealed that the sub—sample of drugs
`with higher IMS sales was marketed internation—
`ally under multiple names and by several differ—
`ent companies. Consequently, sources such as
`MedAdNews didn’t capture all of the sales that
`were licensed to different companies for a partic—
`ular molecule. For the sub—sample of drugs for
`which this was an issue, we utilised the larger IMS
`worldwide sales values because they better cap—
`tured the worldwide market.
`
`2 DiMasi et a1.[5] obtained data from all the firms partici-
`pating in his survey on pre-approval and post-approval R&D
`expenditures. On the basis of an analysis of thesedata, they
`estimated that out—of—pocket R&D expenditures for product
`extensions in the post-approval period were 34.8% of pre- '
`approval R&D expenditures. Applying this percentage to our
`estimate of $US308.4 million for pre-approval R&D yields
`an estimate of $USlO7 million (in $US, 2000 values) as the
`R&D cost for post-launch product improvements.
`
`Using} this approach and these complementary
`data sources, we assembled worldwide sales data
`
`for 66 of the NCEs over the period of 1990 to 2000.
`For the remaining (very small selling) drugs in our
`sample, we multiplied their US sales values by a
`representative global sales multiplier to obtain es— '
`timates of their worldwide sales. The value of the
`
`global sales multiplier was 2.19. As discussed, this
`
`© Adls lnlernallonal lelled. All rlghls reserved.
`
`Pharmacoeconomlcs 2002; 20 Suppl. 3
`
`WATSON LABORATORIES, INC. , IPR2017101622, Ex. 1113, p. 5 of 19'
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 5 of 19
`
`

`

`Grabowski et al.
`
` \
`
`16
`
`latter sub—sample of drugs accounts for a very
`small share of overall sales for the full sample.
`
`Life-Cycle Soles Profiles
`
`Since data were available for the years 1990 to
`2000, 7 to 11 years of worldwide sales values for
`the NCEs in our sample were provided, depending
`on their date of introduction into the US market.
`
`The next task was to estimate future. sales over the
`Complete market life .of these products. Twenty
`years was chosen as the expected market life. This
`is the same assumption that we utilised for 1980s
`new drug introductions. We believe this to be a
`reasonable time horizon for an IR analysis. Any
`sales remaining after 20 years of market life are
`likely to be very small, given the sales erosion ex—
`perienced by most products from generic competi-
`tion and product obsolescence. Furthermore, these
`sales will also be severely discounted by the cost
`of capital in an IRR analysis.
`We utilised a two-step procedure to project fu—
`ture sales values. These steps involve forecasting
`sales to the point of US patent expiry and then pro—
`jecting sales in the post—patent period. The two-
`step approach is illustrated in figure 1 for one of
`the products in our sample. This product was intro-
`duced into the US market in 1992. There are 9
`
`Projected values
`(dashed lines)
`
`800
`
`700
`
`600
`
`500
`
`400
`
`300
`
`200
`
`Patent expiry O
`
`
`
`Meansalesin$USmillions(2000values) 100
`
`01 2 3 4 5 6 7 8 91011121314151617181920
`
`Sales year
`
`Fig. _1. Actual and projected worldwide sales values for a- rep-
`resentative sample product.
`
`years of sales information and its US patent expires
`in year 12. By year 9, this product was in the mature
`portion of its product life cycle. By using a refer—
`ence life—cycle curve, the product was projected to
`have relatively stable sales (in constant dollar
`terms) until year 12. A significant decline is then
`projected in the period after US patent expiry be—
`cause of the entry of generic competitors and re—
`lated economic factors.
`
`The estimated sales decline after patent expiry
`is based on the experience of major commercial
`products coming off patent in the 1994 to 1997
`period. In particular, we examined worldwide sales
`losses for a sample of NCEs for a 4-year period
`following their US patent expiry. The average per-
`centage declines observed were 31, 28, 20 and
`20%, respectively. We utilised these percentages
`to project sales in the first 4 years after patent ex-
`piry and, thereafter, 'a 20% decline until the prod—
`uct’s market life is completed in year 20. In our
`prior work, we found that generic competition is
`focused on products with significant sales at the
`time of US patent expiry. Consequently, for the
`drugs concentrated in the bottom four deciles of
`our sample (with worldwide sales of less than
`$US40 million in year 10 of their market life), we
`assume that the probability of generic competition
`is very low. For these drugs we assume that sales
`losses in the mature phase of cycle will proceed at
`a more moderately declining rate based on the ref-
`erence curve uSed for the pre-patent expiry period.
`We should note that the percentage declines in
`sales from generic competition in the US market
`observed in prior studies are much greater than the
`worldwide losses in sales for major commercial
`products observed here.“6] Hence, the decline in
`worldwide sales in -the post—patent period is amel-
`iorated by the lower incidence of generic competi—
`tion and sales losses outside the US. This may
`change by the time this cohort actually reaches
`patent expiry during the current decade, because
`reference pricing and generic competition are on
`the rise in many European countries.“71
`Figure 2 provides a plot of the sales life—cycle
`profile (in $US, 2000 values) for the top two dec—
`
`© Adls International Limited. All rights reserved.
`
`Pharmacoeconomlcs 2002; 20 Suppl. 3
`s.
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 6 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 6 of 19
`
`

`

`17
`R&D Returns for 19905 New Drug Introductions
`mm
`
`billion. This may be compared with peak sales of
`near $US1.8 billion for the 1980s cohort. The peak
`sales for the 1990s cohort also occur later than for
`the 1980s cohort.
`
`Pre—Tox Contributions and Other
`Economic Parameters
`
`The next step in the analysis was to obtain rev-
`enues net of production and distribution costs (of—
`ten categorised in the economic literature as
`‘quasi-rents’). For this purpose, we analysed pre—
`tax contribution margins in pharmaceuticals dur—
`ing the 1990s. As in prior work, we utilised data
`derived from the income statements of the pharma—
`ceutical divisions of a number of major multina—'
`tional drug companies to obtain representative
`values on contribution margins over timellvz]
`Our analysis of the data on these firms indicated
`that average contribution margins gradually in—
`creased from 42% in the early part of the 1980s to
`approximately 45% at the end of the decade. On
`the basis of these data, we constructed a linear con—
`
`tribution margin schedule over time. In particular,
`the contribution margin is 42% in the first year of
`the product life and grows by increments of 0.3%
`
`500
`
`El 1990-1994
`0 1 980-1 984
`
`‘
`
` 400
`
`300
`
`200
`
`100
`
`
`
`Meansalesin$USmillions(2000values)
`
`
`
`
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`
`
`
`
`Salesin$USmillions(2000values)
`
`3000
`
`
`
`El 1st Decile
`A 2nd Decile
`0 Mean
`I Median
`
`
`
`
`
`1 2 3 4 5 6 7 8 91011121314151617181920
`
`Salesyear
`
`Fig. 2. Worldwide sales profiles of 1990 to 1994 new drug in-
`troductions.
`.
`
`iles as well .as the mean and median drug com—
`pounds in our 1990 to 1994 sample. The sales
`curves illustrate the highly skewed distribution of
`sales in pharmaceuticals that was observed for
`early cohorts. The peak sales of the top decile com—
`pounds are several times the peak sales of the
`second decile compounds. The mean sales curve
`is also significantly above the median.
`‘
`Figure 3 provides a plot of mean worldwide
`sales for the 19905 sample compared with that for
`the 1980s cohort (in $US, 2000 values). Mean
`sales have increased significantly in real terms,
`with peak sales increasing from $US345 million
`for the 19805 cohort to $US458 million for the
`
`1990s cohort. There is also the suggestion that
`sales curves have become somewhat steeper in the
`ascending sales growth stages of the life cycle,
`' with a longer plateau before generic competition
`and product obsolescence take hold.
`Figure 4 shows a corresponding plot of the
`mean worldwide sales for the top decile com—
`pounds in the 1990 to 1994 and 1980 to 1984 pe—
`riods. This is instructive, given that the prospective
`returns for top decile compounds are primary driv—
`ers of R&D investment activities in pharmaceuti—
`cals. For the 1990s cohort, the top decile com—
`pounds reached peak sales of more than $US2.5
`
`0
`
`rfi—i—r—r—r—r—r—r—i—r—r—i—r—r—r-T-I—I—l
`12 3 4 5 6 7 8 91011121314151617181920
`
`Sales year
`
`Fig. 3. Comparison of mean worldwide sales curves for new
`drug introductions in the 1990 to 1994 and 1980 to 1984
`samples.
`
`© Adls International Limited. All rights reserved.
`
`Pharmacoeconomics 2002: 20 Suppl. 3
`
`WATSON LABORATORIES, INO. , IPR2017-01622, Ex. 1113, p. 7 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 7 of 19
`
`

`

`Grabowskz‘ et al.
`18
`WW
`
`[I 1990-1994
`0 1980-1984
`
`3000
`
`IO01OO
`
`2000
`
`1500
`
`1000
`
`500
`
`
`
`
`
`Salesin$USmillions(2000values)
`
`1'2 3 4 5 6 7 8 91011121314151617181920
`Salesyear
`
`Fig. 4. Comparison of mean worldwide sales curves for top
`decile drugs in the 1990 to 1994 and 1980 to 1984 samples.
`
`per year. We also assume that contribution margins
`'will continue to rise at this same rate during the
`current decade. Hence, over the» full 20-year life'
`cycle, target contribution margins are expected to
`rise from 42% in year one, to 48% by year 20, with
`a mean contribution margin of 45% over the full
`life cycle.
`While we constrained margins to average 45 %
`over the life cycle, we also recognise, as in our
`earlier analyses, that promotion and marketing ex—
`penditures are concentrated in the launch phases of
`the life cycle. In our prior analysis, we developed
`the following allocation rule based on a regression
`analysis of promotional and marketing outlays:
`promotion and marketing is equal to sales in year
`1, declines to 50% in year 2, and falls to 25% in
`year 3. We retained this assumed pattern on mar-
`keting outlays in the present analysis. IntervieWs
`with industry participants indicated that the initial
`post-launch years continue to be the primary focus
`of marketing and promotion activities.
`An analysis performed by Rosenthal et al.“81
`indicates that the. drug industry’s marketing ex—
`penses to sales ratios have remained relatively
`stable around 14% in the 1996 to 2000 period. How—
`ever,
`there were some important compositional
`
`shifts over this period. The direct-to—consumer ad—
`vertising to sales ratio increased from 1.2% to 2.2%
`between 1996 and 2000, at the expense of physi—
`cian detailing and hospital medical journal adver-
`tisingllsl
`.
`For the current analysis, we did make one rela-
`tively minor change in the allocation and timing of
`marketing expenditures related to launch. In par—
`ticular, we estimated pre-marketing launch expen—
`ditures in the order of 5 and 10% of first year sales
`in the 2 years immediately prior to launch. These
`marketing expenditures are for activities such as
`pre—launch meetings and symposiums, pricing and
`focus group studies, and sales force training. Our
`assumptions concerning the size and timing of
`these expenditures were guided by a recent survey
`report on pre—launch marketing expenditures by in—
`dustry consultants as well as interviews with some
`of the participating companies.“91
`As indicated above, our model is structured so
`
`that margins average 45% over the full product life
`cycle. Given the assumed pattern of launch expen-
`ditures, contribution margins for each product
`are below representative industry values in the
`first 3 years of marketing. However, as a product
`matures, both promotional and administrative
`costs decline in relative terms, and contribution
`
`margins increase over average industry values in
`the later years of the life cycle.
`The model is also structured to provide for cap-
`ital expenditures on plant and equipment (P&E).
`As in our model for the 19805 cohort, we assumed
`overall capital expenditures for P&E to be equal to
`40% of tenth year sales. Half of these outlays are
`assumed to occur in the first 2 years before market-
`ing and the other half during the initial 10 years of ‘
`the product’s market life. These assumptions imply
`an average capital investment to sales ratio of 3.3%
`over the full product life cycle. This is generally
`consistent with data from pharmaceutical industry
`income statements.
`
`In particular, we checked the reasonableness of
`our assumptions by comparing this implied 3.3%
`capital investment to sales ratio with the corre—
`sponding ratios observed on industry income state-
`
`© Adls International lelted. All rlghts reserved.
`
`Pharmacoec‘onomics 2002: 20 Suppl. 3
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 8 of 19
`
`WATSON LABORATORIES, INC. , IPR2017-01622, Ex. 1113, p. 8 of 19
`
`

`

`19
`R&D Returns for 19905 New Drug Introductions
`
`
`ments during the 1990s. We found that the drug
`industry capital investment to sales ratio averaged
`about 7.0% during the 1990s. However, the latter
`value includes investment for R&D as well as
`
`production, marketing and administrative facil—
`ities. In our model, provisions for capital invest-
`ment in R&D facilities are included in the cost
`
`estimates prdvided by DiMasi et al.[5] Accord—
`ingly, we asked some industry members involved
`with strateg