ORIGINAL RESEARCH ARTICLE
`messengeasanana
`
`@Adls International Limltecl. All rights resewed.
`
`The Distribution of Sales Revenues
`
`from Pharmaceutical Innovation
`
`Henry G. Grabowski and john Vernon
`
`Department of Economics, Duke University, Durham, North Carolina, USA
`
`AbSITGCT
`
`Objective: This report updates our earlier work on the returns to pharmaceutical
`research and development (R&D) in the US (1980 to 1984), which showed that
`the returns distributions are highly skewed. lt evaluates a more recent cohort of
`new drug introductions in the US (1988 to 1992) and examines how the returns
`distribution is emerging for drugs with life cycles concentrated in the l9905
`versus the 1980s.
`
`Design and setting: Methods were described in detail in our earlier reports. The
`current sample included 110 new drug entities (including 28 orphan drugs), and
`sales data were obtained for the period 1988 to 1998, which represented between
`7 and l 1 years of sales for the drugs included. 20 years was chosen as the expected
`market life for this cohort, and a 2-step procedure was used to project future sales
`for the drugs — during the period until patent expiry and then beyond patent expiry
`until the 20-year time-horizon was completed. Thus, the values in the first half
`of the life cycle are essentially based on realised sales, while those in the second
`half are projected using information on patent expiry and other inputs.
`
`Main outcome measures and results: Peak annual sales for the top decile of
`drugs introduced between 1988 and 1992 in the US amounted to almost $US l .1
`billion compared with peak sales of less than $USl 75 million (1992 values) for
`the mean compound. In particular, the top decile accounted for 56% of overall
`sales revenue. Although the sales distributions were skewed in both our earlier
`and current analysis, the top decile in the later time-period exhibited more rapid
`rates of growth after launch, a peak that was more than 50% greater in real terms
`than for the 1980 to I984 cohort, and a faster rate of expected decline in sales
`after patent expiry. One factor contributing to the distribution of sales revenues
`becoming more skewed over time is the orphan drug phenomenon (i.e. most of
`the orphan drugs are concentrated at the bottom of the distribution).
`
`Conclusion: The distribution ofsales revenues for new drug compounds is highly
`skewed in nature. In this regard, the top decile of new drugs accounts for more
`than half of the total sales generated by the 1988 to 1992 cohort analysed.
`Furthermore, the distribution of sales revenues for this cohort is more skewed
`
`than that of the 1980 to 1984 cohort we analysed in previous research.
`
`
`WATSON LABORATORIES v. UNITED THERAPEUTICS, |PR2017-0‘l622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 1 of 12
`
`

`

`Grnbowski :9 Vernon
`
`
`In this study, we examine the distribution of
`sales revenues for a comprehensive sample of new
`drugs introduced into the US during the period
`1988 to 1992. In earlier research, we examined the
`
`second decile compounds, $USl 50 million for the
`mean compound, and only $USSO million for the
`median compound (1990 values). In our analysis,
`we also estimated the ‘quasi-profits’ for each entity
`
`returns to research and development (R&D) on US
`new drug introductions during the 1970s and early
`1980s.[152] One ofthe key findings was that the top
`decile of new drugs accounted for a large share of
`the total market value generated by these entities.
`In this regard, the returns to R&D projects in
`pharmaceuticals have properties similar to those of
`venture capital investments. This has important
`implications for both private and public deci-
`sion- makers.
`
`A new analysis of this issue is warranted by a
`number of important changes on both the demand
`and supply sides of the market for new drugs. In
`particular, there has been significant new entry and
`industry restructuring since our last analysis of the
`returns to R&D. In addition, managed care has
`grown dramatically during the 1990s, and now ac-
`counts for a dominant proportion of drug prescrip-
`tions. These factors can significantly affect the life
`cycles of sales and the distribution of revenues
`across new drug introductions.
`
`Background
`
`The 1980 to 1984 Cohort of New Drug
`Introductions
`
`— the surplus of global sales revenues over produc-
`tion and distribution costs — and discounted them
`
`to the date of market launch. The top decile, the
`most profitable 10% of the compounds, contrib-
`uted 48% of the quasi-profits realised by the full
`sample of NCE introductions during this period.
`By contrast, the bottom half of the distribution
`(deciles 6 through 10, encompassing the entities
`with peak sales below $US50 million) accounted
`in total for only 8% of the quasi-profits.
`
`Returns for Venture Capital Investments and
`Initial Public Offerings (IP05)
`
`Recent work by Scherer et al.[3'4] has shown that
`many other innovational activities are charac-
`terised by skewed outcome distributions. Ofpartic-
`
`ular interest are 2 of their data samples involving a
`large number of investments by US venture capital
`
`firms in start-up companies between 1969 and
`1988. The first sample was compiled by Venture
`Economics Incorporated and involved a portfolio
`of investments in 383 start-up companies made by
`13 venture capital firms. The second sample in-
`volved a similar data set assembled by Horsley-
`Keough Associates of 670 distinct investments
`made by 16 venture capital companies.
`
`In this section, we summarise some ofthe core
`
`Scherer’s analysis indicates that investment re-
`
`findings from our previous work on pharmaceuti-
`cals and relate them to recent work on the returns
`
`for venture capital investment. Our last analysis
`focused on a comprehensive sample of 64 new
`chemical entities (NCEs) introduced into the US
`market between 1980 and 1984“] In this regard,
`figure 1 shows the sales profiles over the marketing
`life cycle for the top 2 deciles of NCEs (ranked by
`tenth-year US sales) and the mean and median
`compound. The figure indicates that there is a high
`degree of variability in the sales performance of
`NCEs. In particular, the peak annual US sales were
`more than $US700 million for the top decile com-
`pounds, approximately $US300 million for the
`
`turns from venture-financed start-ups are highly
`skewed. As shown in table I, a relatively small
`number of start-up firms generate a large share of
`the total investment value, as measured by the cap-
`ital appreciation or loss at the time of investor exit
`from each investment. In the case of the Venture
`
`Economics sample, the most profitable decile of
`projects accounted for 62% of the total value gen-
`erated by all 383 investments. For the Horsley-
`Keough sample, 59% of the overall value was at-
`tributed to the top decile of start-up company
`investments. This can be compared with our sam-
`ples of 1980 to 1984 NCEs, where the top decile of
`NCEs accounted for 48% of the quasi-profits.
`
`© Adis International Limited. All lights reserved.
`
`Phorrnocoeconomics 2000: 18 Suppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-01622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 2 of 12
`
`

`

`Revenue Distribution from Pharmaceuticals
`
`
`30° ' l 15: Decile
`A 2nd Decile
`El Mean
`A Median
`
`600 -
`
`
`
`
`
`
`
`
`Salesin$USmillions(1990values)
`
`
`
` 4
`
`400 —
`
`200 '-
`
`5
`
`6
`
`7
`
`B 91011121314151617181920
`
`Fig. 1. US sales profiles for 1980 to 1984 new chemical enliliesi”
`
`Salesyear
`
`Scherer et al.[3] also examined the stock market
`
`performance of a comprehensive sample of 110
`
`venture-funded high-technology companies that
`had their lPOs between 1983 and 1986. A decade
`
`later, he examined the returns from an equal dollar
`investment in each of these companies at the time
`of their 1P0. An investment in a full bundle of these
`
`IPO companies would have slightly outper-
`formed a comparable dollar investment in the
`NASDAQ index over the same period.l However,
`the market performance of these 1P0 firms also
`exhibited the same tendency toward extreme val-
`ues as the samples involving venture-financed
`start-up investments discussed earlier in this sec-
`tion. As shown in table I, the I] firms that consti-
`
`tuted the most profitable decile ofthese [P0 com-
`panies accounted for 62% of the overall market
`
`value in 1995. Correspondingly, the other 99 high-
`technology firms in this sample accounted for the
`remaining 38%.
`
`Implications for R&D Investments
`
`The data shown in table I indicate that R&D
`
`investments in pharmaceuticals have much in
`
`1 Returns were based on the market values of these com-
`
`panies approximately I decade later (December 31. 1995).
`This analysis takes account of the market values of the
`surviving IPO companies,
`those that merged with other
`firms, and those deleted because of bankruptcies and failure
`to meet NASDAQ financial criteria.
`
`common with private investments by venture cap-
`ital firms in start-up companies as well as public
`market investments in high-technology 1P0 com-
`panies. All of these innovative investment activi-
`
`ties are characterised by a high degree of risk. This
`results from the fact that a few extreme values ac-
`
`count for a large share of the cumulative realised
`returns. As Scherer and others have observedJ‘” the
`law of large numbers doesn’t work very well when
`
`the probability distribution of outcomes is highly
`skewed. One important consequence for pharma-
`
`ceutical R&D is that considerable variability in
`portfolio outcomes can be expected, even for those
`
`pharmaceutical companies with large diversified
`portfolios of R&D pipeline drugs.
`In the case of pharmaceuticals, the blockbuster
`
`compounds, which constitute the top decile of
`
`NCEs in figure 1, generally represent significant
`therapeutic advances in treating a particular dis-
`
`Table I. Distribution of returns for selective innovative samples
`Data set
`Percent of value
`in top decile
`62
`
`Venture Economics (383 start-up
`investments) [Scherer et atm]
`Horsiey—Keough (are start-up investments)
`[Scherer et al.13]]
`Scherer et at.“ (1983-1986 lPOs: market
`value in 1995)
`43
`Grabowski and Vemonm
`
`{1980-1934 NCEs)
`IP05 = initial public offerings; NCEs = new chemical entities.
`
`59
`
`62
`
`© Adis International Limited. All fights reserved.
`
`Phormclcoeconomics 20(1): l8 Suppl. l
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-O1622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 3 of 12
`
`

`

`24
`Grnbowski 8 Vernon
`
`
`ease, usually one with significant market size. In
`most instances, these therapies are the first or sec-
`ond introductions in a new chemical class of com-
`
`pounds, and offer a novel approach to treating a
`particular disease.[51 The pharmaceutical industry
`has also been characterised historically by signifi-
`cant first-mover advantages.[6’7] Other things being
`equal, later market entrants tend to capture substan-
`tially lower market shares.
`
`The most novel compounds face the greatest
`risks — from a scientific, regulatory and commer-
`
`cial perspective. In this regard, the therapeutic pro-
`
`files of these compounds are the most difficult to
`
`predict on the basis ofpreclinical screens and leads.
`In addition, the long lag time and R&D activities
`ofcompetitors magnify these scientific and techni-
`
`cal risks. Accordingly, unforeseen clinical outcomes,
`
`the introduction of rival products and other changes
`
`in the market, and regulatory problems and lag
`
`times can dramatically affect a new drug’s eco-
`nomic prospects during the development process.
`
`These factors help to explain why so many of
`
`the compounds in figure 1 are marketed despite
`very small peak sales revenues and quasi-profits that
`are a small fraction ofmean R&D costs.2 lfsignifi-
`cant uncertainties surrounding a compound ’5 eco-
`nomic prospects are not resolved until clinical de-
`
`velopment is largely complete, most of the R&D
`costs are then sunk. At this point, as long as a com-
`
`pound’s expected revenues cover the incremental
`
`or variable costs on a prospective basis, it is ra-
`tional to market or license out the compound,
`even if this doesn’t cover any of the compounds
`
`2 We did not have R&D costs on an individual NCE basis.
`Another factor could be that R&D costs are also lower for
`
`drug entities with smaller sales and quasi—profits. While this
`may be the case, an analysis of R&D costs for a repre-
`sentative sample of NCEs at different stages of the R&D
`process by DiMasi et ails] indicated that there is much less
`variability in R&D costs than in revenues across NCEs. This
`is plausible, given the fact that all FDA approved drugs must
`meet stringent regulatory requirements. Approved drugs also
`share in common pre—project discovery costs and the costs of
`failures. These components account for more than 50% of
`the mean estimated R&D cost of $U8202 million in the
`mid—[9805.
`
`large fixed R&D costs. Of course, in the long
`run, the firm also must have its share of winners
`
`for its R&D programme to be profitable and remain
`viable.
`
`Recent Market Developments
`
`The basic sample to be investigated comprises
`1 10 new drug entities developed for the US market,
`approved by the FDA, and introduced into the US
`market between 1988 and 1992. This is a compre-
`hensive sample ofthe new drug entities introduced
`into the US market during this period. In this paper,
`we focus on the US sales performance of these en-
`
`tities. In future papers, we will examine the returns
`on R&D of these entities and integrate global sales
`
`and costs into the analysis.
`In our past work, we have found that differences
`in sales revenues constitute the major driving force
`
`underlying the skewed distribution of quasi-profits
`across NCESI'QI An analysis of sales performance
`in the US is therefore interesting in its own right. In
`this regard, the US is also the largest market for
`pharmaceuticals, accounting for roughly halfofthe
`sales relating to new drug introductions studied in
`past samples. We also found that sales ofthese new
`drugs in other major markets (Europe and Japan)
`were significantly positively correlated with their
`US sales revenues.
`
`Managed Care and Demand Side Changes
`
`As noted in the introduction, the demand side
`
`of the market for new pharmaceuticals has been
`undergoing substantial change during the past de-
`cade. Pharmacy benefit management firms (PBMs)
`have emerged as the main overseers of the prescrip-
`tion drug plans of employers and managed-care
`institutionslgdol PBMs have implemented drug
`formularies to encourage more price competition
`and incentive programmes for generic drug usage
`when brand products come off patent. At the same
`time, managed-care institutions have broadened
`insurance coverage for prescription drugs, and unit
`sales have grown as drug therapies and compliance
`have been encouraged as a way of avoiding more
`expensive medical treatments.
`
`© Adis lntemalional Limited. All rights reserved.
`
`Pharrnacoeconomics 201): IS Suppl. I
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, IPR2017-01622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 4 of 12
`
`

`

`25
`Revenue Distribution from Pharmaceuticals
`
`
`PBMs and health maintenance organisations
`(HMOs) can have differing effects on the sales rev-
`enue for a new drug introduction over the market-
`ing life cycle. New drugs that represent novel thera-
`
`In our sample, there is also a high degree of
`overlap between the biopharmaceutical and orphan
`drug sets. This phenomenon has been discussed
`elsewhere and is the result of several factors.[”]
`
`peutic interventions for particular diseases and
`conditions have generally received broad coverage
`
`First, many ofthe initial biotechnology drugs were
`recombinant versions of natural hormones with ap-
`
`and speedy approvals for inclusion on drug formu-
`laries. However, as follow-on drugs are introduced
`into the same class, price discounting and compe-
`tition usually occur in order to obtain formulary
`access. The growth of managed care has also been
`an important factor contributing to a more rapid
`erosion of sales when drugs come off patent.
`Therefore, as a new drug proceeds through its mar-
`keting life cycle, and as competition develops in a
`given therapeutic class, the influence of the PBMs
`of managed-care providers on sales revenues is
`subject to important shifts over time.
`
`Biapharmaceuticals. Orphan Drugs and
`Supply Side Changes
`
`There have also been important changes in the
`
`supply side of the market. In this regard, the num-
`ber of new drug entities introduced onto the US
`market during the 1988 to 1992 period is signifi-
`cantly larger than during the earlier 1980 to 1984
`period. This reflects some important industry de-
`
`velopments. First, the current sample includes new
`biopharmaceutical entities as well as NCEs. The bio-
`
`technology industry was essentially in its infancy
`in the early 1980s. However, by the early l990s, it
`
`had become a significant source of new therapeutic
`entities.
`
`Another important event was the passage of the
`Orphan Drug Act by Congress in 1983. This pro-
`vided incentives in the form of tax credits, market
`
`exclusivity, and regulatory assistance for the devel-
`opment of drugs targeted to diseases and condi-
`tions involving small patient populations.[“] In
`particular, a drug is eligible for orphan drug status
`under the law if it is approved for an indication
`involving a population of <200 000 patients.
`Roughly one-quarter of the drugs in our current
`sample were granted orphan drug status for at least
`one approved indication.
`
`proved indications for small patient populations.
`In addition, many biopharmaceutical firms sought
`the market exclusivity protection of orphan drug
`status, given the initial uncertainties surrounding
`biopharmaceutical patents.
`
`It is important to point out that there is wide
`variability in the sales revenues realised by orphan
`
`drugs in our sample. In particular, some of the
`novel biotechnology drugs granted orphan drug
`
`status were able to achieve blockbuster status by
`obtaining relatively large reimbursements per drug
`
`treatment. In addition, some of these drugs re-
`ceived orphan drug status for some indications as
`
`well as approval for other non-orphan indications.
`Conversely, many of the orphan drug approvals in
`
`the 1988 to 1992 period were for very rare condi-
`tions and, by historical standards, these drugs had
`very small sales (i.e. annual sales of only a few
`million dollars). Hence, the group of orphan drug
`compounds is very heterogeneous in nature.
`
`Data Samples and Methodology
`
`Annual drugstore and hospital sales in the US
`were obtained from IMS America for each of the
`
`l 10 new drug entities in our sample. The sales data
`covered the period 1988 to I998. This provided
`
`between 7 and 11 years of sales data for the drugs
`in our sample cohort, depending on a drug’s year
`of introduction.
`
`20 years was chosen as the expected market life
`for this cohort. We felt this was a reasonable value,
`
`since virtually all of the drugs in our sample had
`patent lifetimes of significantly less than 20 years,
`and products with substantial market sales would
`be expected to face strong generic competition and
`sales losses after patent expiry. While some prod-
`ucts may have positive sales after year 20, these
`sales would be expected to be small and to have
`
`© Adis International limited. All rights reserved.
`
`Pharrnacoeconomics 2WD: 16 mppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-01622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 5 of 12
`
`

`

`26
`Grabowski 8 Vernon
`
`
`very low weights in any type of discounted present
`value analysis.
`
`for these drugs will tend to be understated if current
`trends persist into the future.
`
`We used a 2-step procedure to project future
`sales values for the products in this sample. A key
`
`time-point in the life cycle of sales is the year of
`
`patent expiry. This is clustered between years 10
`
`and 14 for the current sample. In our approach, the
`
`first step involved projection to the year of patent
`expiry and the second step projection of the post-
`
`patent expiry values.
`
`To project sales to the point ofpatent expiry, we
`
`utilised an approach similar to our past analy-
`ses.['»21 In particular, we constructed a reference
`
`life cycle curve based on the sales of products in-
`
`troduced in the mid-1980s (i.e. the new drug cohort
`
`immediately preceding the current one). We used
`this as the basic framework to project sales values
`for most of the NCEs. However, to take account of
`
`recent market and competitive developments af-
`
`fecting demand for the leading compounds and
`therapeutic groups, we also utilised the sales fore-
`
`casts from a group of security analysts to make
`
`adjustments when there was a significant deviation
`from the reference case.
`
`The estimated sales for the period after patent
`expiry were based on an analysis of generic com-
`petition in the mid—l 9905.”:‘31 When this analysis
`
`was used, the percentage decline in average sales
`
`during the first 2 years after patent expiry for prod-
`ucts with annual sales of $USSO million or more at
`
`the time of patent expiry were computed to be 43
`and 42%, respectively.3 Thereafter, a 10% annual
`decline was utilised over the remaining years of
`
`market life. In our analysis of generic competition
`
`since the passage of the 1984 Waxman-Hatch Act,
`
`we have observed a strong trend over time toward
`an increased erosion ofsales after patent expiry.“3l
`Since most of the products in our sample will ex-
`
`perience patent expiry in the early part of this de-
`cade, the rates of erosion of sales after patent expiry
`
`3 The probability of generic competition is low for drugs
`with annual sales at the time of patent expiry that are below
`$USSO million. Accordingly, we assumed no generic compe-
`tition would occur in the case of these smaller selling drugs.
`
`Empirical Results
`
`Sales of Orphan versus Non-Orphan Drugs
`The first issue we examined was the sales per-
`formance of the orphan versus non-orphan drugs in
`our sample. Figure 2 shows a plot of the life cycle
`of sales profiles for the mean compound in these 2
`subsamples of drugs. As discussed in the previous
`section, Data Samples and Methodology, the val-
`ues in the first half of the life cycle are essentially
`based on realised sales, while those in the second
`
`half are projected using information on patent ex-
`piry and other inputs.4
`The non-orphan drugs exhibit the general char-
`acteristics observed in prior work: rapid growth af-
`ter launch, maturation about 10 to 11 years into the
`life cycle, and then a rapid decline in sales after
`patent expiry and generic entry. By contrast, the
`orphan compounds exhibit more moderate growth
`
`rates after launch, a much lower expected peak
`sales level, but also slower expected rates of de-
`clines in sales in the later stages of the life cycle.
`The last-mentioned phenomenon is due to longer
`average patent protection periods as well as less
`generic exposure for the orphan drug population,
`given their smaller average sales levels.
`
`As discussed in the section entitled Biopharma-
`ceuticals, Orphan Drugs and Supply Side Changes,
`there are different economic incentives in terms of
`
`the R&D and regulatory process for orphan drugs
`
`compared with non-orphan drugs. In future work,
`we plan to investigate their economic returns in a
`
`separate study. Nevertheless, the 28 orphan drugs
`in our eun'ent sample are very heterogeneous and
`
`include some of the leading biopharmaceutical
`products such as epoetin-Ot (erythropoietin) and
`
`4 Since our sample involves a basket of new drugs intro—
`duced between I988 and I992, years 8 to l
`l of the life cycle
`are a blend of actual and forecasted sales. For example, year
`8 involves the first year of forecasted sales for the 1992
`cohort and actual sales for the 1988 to 199] cohorts. Simi-
`
`larly, year ll involves actual sales for the 1988 cohort and
`projected sales for the other cohorts.
`
`© Adis lnlemulioncul Limited. All lights reserved.
`
`Pharmacoeconomics 2011: IS Suppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-01622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 6 of 12
`
`

`

`27
`Revenue Distribution from Pharmaceuticals
`
`
`200
`
`I Non-orphan NCEs
`El Orphan NCEs
`
`9
`
`100
`
`
`
` 8Salesin$USmillions(1992values)
`
`
`
`
`
`123456789 10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`1?
`
`13
`
`19 20
`
`Sales year
`
`Fig. 2. Mean US sales for 1988 to 1992 new chemical entities (NCEs). Orphan versus non-orphan drugs.
`
`human growth hormone. Because of this, we have
`chosen not only to retain these orphan drugs in our
`
`decile in the later time-period exhibits more rapid
`rates of growth after launch, a peak more than 50%
`
`sample but also to analyse the distribution of sales
`
`with and without these drugs present. The results
`
`do not change in a qualitative manner.
`
`Distribution of Series for was to 1992 introductions
`
`Figure 3 provides a plot of the expected sales
`
`profiles for the full sample of new drugs for the
`
`l 988 to 1992 period. This provides an exact counter-
`part of figure 1, and shows the life cycle of sales
`
`patterns for the top 2 deciles and the median and
`mean drug compounds. The main observed differ-
`
`ence between figures 1 and 3 is associated with the
`
`top decile of drugs. In particular, the ‘mountain’
`
`type profile of the top decile in figure 3 has grown
`taller and steeper compared with the other profiles
`
`displayed in these figures. In this respect, the top
`
`greater in real terms than for the 1980 to 1984 co-
`hort, and a faster rate of expected decline in sales
`
`after patent expiry.
`The definite impression from figure 3 is that the
`distribution of revenue has become more skewed
`
`over time. One factor contributing to this trend is
`the orphan drug phenomenon. Most of the orphan
`drugs are concentrated at the bottom of the distri-
`
`bution, with a few blockbuster drugs in the top
`decile. This tends to make the overall distribution
`
`more skewed. However, the basic findings are not
`altered in a qualitative manner when the orphan
`drugs are omitted from the sample.
`The movement toward more skewness over
`
`time, as indicated in figure 3, was confirmed by a
`more detailed analysis that we performed. This is
`
`_l NOD
`
`900
`
`0')8
`
`m8
`
`c:
`
`
`
`I 1st Decile
`A 2nd Decile
`El Mean
`:3 Median
`
`
`
`123456789101112
`
`13 14 15 16 17 18 19 20
`
`
`
`
`
`Salesin$USmillions(1992values)
`
`Fig. 3. US sales profiles of 1988 to 1992 new chemical entities.
`
`Salesyear
`
`© Adis lntemcttionol Limited. All rights reserved.
`
`Photmclcoeconomics 20(1): 18 Suppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-O1622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 7 of 12
`
`

`

`28
`Grnbotuski 8 Vernon
`
`
`presented in figure 4. In this figure, we plot the full
`distribution of sales by decile for the 1980 to 1984
`and the I988 to I992 cohorts. Here sales data are
`
`based on the seventh year after launch, so that this
`analysis is based completely on actual sales values.
`In particular, the top decile of new drugs for the
`
`1988 to 1992 period accounts for 56% ofthe over-
`
`all sales revenue for the full sample of 110 drugs.
`If we omit the 28 orphan drugs, the top decile ac-
`counts for 52% of the sales revenue. By contrast,
`
`the top decile ofNCEs for the I980 to 1984 period
`accounted for 48% of overall sales (and the same
`
`percentage of quasi-profits).
`The top decile in figures 3 and 4 is dominated
`
`by new drug introductions that are pioneers or early
`entrants in a new therapeutic class of compounds.
`
`In particular, this group includes the world’s largest
`selling drug in 1998, Prilosec® (omeprazole), the
`
`first drug in the proton pump inhibitor class, which
`is used to treat ulcers. It also includes the first 2
`
`selective serotonin re-uptake inhibitors, Prozac®
`(fluoxetine) and Zolofi‘i‘) (sertraline), used to treat
`
`depression. Also in the top decile of drugs are the
`
`2 largest selling biopharmaceutical therapies —
`Epogen® (epoetin-tx), which is used for treating
`anaemia, and Neupogen® (filgrastim), which is
`used as an adjunctive chemotherapeutic agent. In
`
`addition, the top-selling decile includes the follow-
`ing: Taxol® (paclitaxel), the leading chemothera-
`
`1000
`
`peutic drug for ovarian cancer; Norvasc‘E' (amlo-
`dipine), a new kind of calcium antagonist for treat-
`ing hypertension; Biaxin® (clarithromycin) and
`Zithromax® (azithromycin), 2 semi-synthetic mac-
`rolide anti-infective agents; and Pravachol® (prava-
`statin) and Zocor® (simvastatin), 2 leading statin
`drugs for cholesterol reduction.
`
`Changes in Mean Sales over Time
`In the case of skewed distributions, the revenue
`
`performance of the mean compound is dis-
`proportionately affected by the realised values in
`
`the upper tail of the distribution. Accordingly, we
`would expect the mean sales to be significantly
`greater in the I988 to 1992 cohort, compared with
`the earlier 1980 to 1984 cohort. In order to see how
`
`sales of the mean compound have changed over
`time, we plot the mean curves for the 2 time cohorts
`on a separate graph in figure 5. This graph is based
`on the entire sample of 1988 to 1992 drugs, includ-
`ing the orphan compounds. The case with the or-
`phan drugs excluded is shown in figure 6.
`
`In both cases, there is a significant upward shift
`
`in the mean sales curves through the period of prod-
`uct maturity. However, the faster rate of generic
`
`competition expected for the later time cohorts
`causes a projected convergence of the 2 curves af-
`
`ter year 15 of the life cycle. Nevertheless, the ex-
`pected present value of sales revenues will be
`higher for the more recent time cohort, given the
`
`El 1980-1984 NCES
`I 1988-1992 NCEs
`
`a:8
`
`
`
`m8
`
`.p.3
`
`
`
` N8Salesin$USmlllions(1992values)
`
` 6
`
`9
`
`10
`
`T
`
`8
`
`Fig. 4. Distribution of US sales by decile. Data reflect actual sales in the seventh year after marketing. NCEs = new chemical entities.
`
`Deciles
`
`© Adis International limited. All rights reserved.
`
`Phorrnucoeconomies 2000; 16 alppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-O1622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 8 of 12
`
`

`

`29
`Revenue Distribution from Pharmaceuticals
`
`
`I 1988-1992 NCEs
`[II 1930-1934 NCEs
`
`200
`
`150
`
`100
`
`ClC!
`
`
`
`
`
`Salesin$USmillions(1992values)
`
`
`123456789
`10
`11
`12
`13
`14
`15
`16 17
`18
`19 20
`
`Salesyear
`
`Fig. 5. Comparison of mean US sales for the different new chemical entity (NCE) cohorts, with orphan drugs included in the 1988
`to1992 cohort.
`
`positive differences in the earlier years of the life
`cycle. As noted, this is driven in large part by the
`
`sales performance of the top decile products.
`
`Sales of New Drug introductions versus R&D
`Ouflays by Company
`
`As discussed earlier, one important conse-
`quence of a skewed distribution is that even firms
`
`with sizeable portfolios of R&D projects can ex-
`pect considerable variability in portfolio out-
`
`comes. In order to gain some further insights into
`this issue, we aggregated each company’s sales (in
`
`the seventh year of market life) for all of its new
`drug introductions during the 1988 to 1992 period.
`We then plotted these portfolio outcomes against
`the company’s pharmaceutical R&D expenditures
`
`in the 1983 to 1985 period. We utilised an average
`
`lag time of6 years between R&D expenditures and
`new drug introductions to reflect the long gestation
`period in pharmaceutical R&D.[“]
`Figure 7 shows the resulting plot of new drug
`sales versus R&D expenditures for a total of 18
`
`firms for which R&D expenditure data were avail-
`able. There is a considerable range in the size of
`these firms, but all can be characterised as multi-
`
`national companies that are also vertically inte-
`
`grated across all types of pharmaceutical activities
`(i.e. R&D, manufacturing and marketing). The an-
`nual R&D expenditures of these 18 firms in the
`mid-19803 was between $USlOO million and
`
`$USSOO million (measured in 1992 dollars).
`
`
`
`
`
`Salesin$USmillions(1992values)
`
`200
`
`150
`
`100
`
`010
`
`I 1933-1992 NCEs
`El 1930-1934 NCEs
`
`
`
`123456739 10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`1?
`
`18
`
`19 20
`
`Sales year
`
`Fig. 6. Comparison of mean US sales for the different new chemical entity (NOE) cohorts, with orphan drugs excluded in the 1988
`to 1992 cohort.
`
`© Adis International Limited. All rights reserved.
`
`Phormucoeconomics 2030; 18 Suppl. 1
`
`WATSON LABORATORIES V. UNITED THERAPEUTICS, |PR2017-O1622
`
`UNITED THERAPEUTICS, EX. 2080
`
`Page 9 of 12
`
`

`

`Grnbowskr' 8 Vernon
`
`
`4000
`
`3000
`
`2000
`
`
`
`Companysalesfornewdrugs($USmillions)
`
`
`
`1000
`
`100
`
`200
`
`300
`
`400
`
`500
`
`600
`
`Company R&D expenditures (SUS millions)
`
`Fig. 7. Sales for 1988 to 1992 new drug introductions (data from the seventh year post-launch) plotted against 1983 to 1985 mean
`research and development (R&D} expenditures for 18 multinational pharmaceutical companies. Both sales revenue and R&D outlays
`are expressed in 1992 dollars.
`
`Figure 7 shows that there is a positive relation-
`
`ship between a company’s R&D expenditures and
`its subsequent sales from new drug introductions.5
`However, there is also much variation in the scatter
`
`of points around the best fitted least-squares re-
`
`gression line, as shown in figure 7