ORIGINAL RESEARCH ARTICLE
`
`Pharmacoeconomics 2002; 20 Suppl. 3: 11-29
`1170-7690/02/0003-0011/$25.00/0
`
`© Adis International Limited. All rights reserved.
`
`Returns on Research and Development
`for 1990s New Drug Introductions
`Henry Grabowski,1 John Vernon1 and Joseph A. DiMasi2
`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 1990s,
`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 so.
`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 1980s 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 1970s and 1980s introductions per-
`formed by Grabowski and Vernon.[1,2]
`
`Our prior analyses indicate that this industry has
`exhibited very skewed distributions of returns. In
`this regard, several significant new classes of drug
`therapies have been introduced since the late
`1970s. 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 1970s and 1980s’ samples.
`
`

`
`12
`
`Grabowski et al.
`
`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.[3]
`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.[4] 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 1984.[6,7]
`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.[1,2] 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
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`R&D Returns for 1990s New Drug Introductions
`
`13
`
`industry.[8] 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.[8]
`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.[10]
`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.[11] We also performed a
`comparable CAPM for analysis for January 2000.
`The results of these CAPM-based studies are
`summarised in DiMasi et al.[5]
`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 analysis.[12] 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 International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`14
`
`Grabowski et al.
`
`Myers and Howe further indicate that the R&D
`decision process can be modelled as a compound
`option pricing model.[11] 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 approach using a weighted cost of capital.
`DiMasi et al.[5] 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.[13] 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.1
`
`1 Since our sample is centred around 1992, we utilise the
`following linear extrapolation equation to derive R&D costs:
`R&D92 = R&D84 + (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 al. 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 $US429 million.
`In addition to these pre-launch R&D expendi-
`tures, firms also undertake R&D outlays in the
`post-approval period for product extensions such
`as new indications, formulations and dosage lev-
`els. Since these activities can be viewed as spillo-
`vers from the original NCE introduction, these on-
`going R&D investment expenditures, as well as
`any extra revenues that they generate, are appro-
`priately incorporated into the analysis. On the
`basis of the DiMasi et al. study,[5] we estimated
`
`Table I. Capitalised research and development (R&D) costs for the
`mean new chemical entity in the 1990 to 1994 sample
`
`Pre-tax
`
`After tax
`
`R&D costs
`($US millions; 2000 values)a,b
`$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.
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`R&D Returns for 1990s New Drug Introductions
`
`15
`
`the average post-approval R&D costs per NCE in
`our sample period to be $US107 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 $US51 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.
`
`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-
`
`2 DiMasi et al.[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 these data, 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 $US107 million (in $US, 2000 values) as the
`R&D cost for post-launch product improvements.
`
`ally than US compounds with smaller domestic
`sales.[14]
`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.[15]
`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 selling 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.
`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
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`16
`
`Grabowski et al.
`
`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.[16] 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.[17]
`Figure 2 provides a plot of the sales life-cycle
`profile (in $US, 2000 values) for the top two dec-
`
`latter sub-sample of drugs accounts for a very
`small share of overall sales for the full sample.
`
`Life-Cycle Sales 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 IRR 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)
`
`Patent expiry
`
`800
`
`700
`
`600
`
`500
`
`400
`
`300
`
`200
`
`100
`
`0
`
`Mean sales in $US millions (2000 values)
`
`0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
`Sales year
`
`Fig. 1. Actual and projected worldwide sales values for a rep-
`resentative sample product.
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`R&D Returns for 1990s New Drug Introductions
`
`17
`
`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-Tax 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 time.[1,2]
`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%
`
`1990-1994
`1980-1984
`
`500
`
`400
`
`300
`
`200
`
`100
`
`0
`
`Mean sales in $US millions (2000 values)
`
`1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
`Sales year
`
`Fig. 3. Comparison of mean worldwide sales curves for new
`drug introductions in the 1990 to 1994 and 1980 to 1984
`samples.
`
`1st Decile
`2nd Decile
`Mean
`Median
`
`3000
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`0
`
`Sales in $US millions (2000 values)
`
`1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
`Sales year
`
`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 1990s 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 1980s 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
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`Grabowski et al.
`
`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-
`tising.[18]
`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.[19]
`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 1980s 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-
`
`1990-1994
`1980-1984
`
`3000
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`0
`
`18
`
`Sales in $US millions (2000 values)
`
`1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
`Sales year
`
`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.[18]
`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
`
` Adis International Limited. All rights reserved.
`
`Pharmacoeconomics 2002; 20 Suppl. 3
`
`

`
`R&D Returns for 1990s New Drug Introductions
`
`19
`
`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 provided by DiMasi et al.[5] Accord-
`ingly, we asked some industry members involved
`with strategic planning for information on what
`percentage of their P&E expenditures was devoted
`to R&D, versus other firm activities. We obtained
`a range of 40 to 50% of total capital expenditures
`devoted to R&D. Given this range, the capital in-
`vestments to sales ratio for non-R&D activities
`implied by our model is consistent with the ob-
`served data from company income statements.
`For working capital, it was assumed that ac-
`counts receivables are equal to 2 months of annual
`sales and inventories are 5 months of sales (valued
`at manufacturing cost). These are also based on
`the analysis of balance sheet data of major phar-
`maceutical firms. Working capital is recovered at
`the end of the final year of product life.
`
`Effective Tax Rates
`
`Our analysis of returns is conducted on an af-
`ter-tax basis. In our prior studies of returns, we
`computed average effective tax rates based on
`analysis of income statement data from eight major
`pharmaceutical firms. The average effective rate
`was 35% for the 1970s cohort and 33% for the
`1980s cohort. A comparable analysis for the 1990s
`cohort yielded an effective tax rate of 30%. This is
`the rate used in our baseline case. The difference
`between the nomin