0092-861 5/2000 Copyright 0 2000 Drug Information Association Inc.

Drug fnformnfionJourMl, Vol. 34. pp. 1169-1194. 2000 Printed in the USA. All rights reserved.

NEW DRUG INNOVATION AND PHARMACEUTICAL INDUSTRY STRUCTURE: TRENDS IN THE OUTPUT OF PHARMACEUTICAL FIRMS JOSEPH A. DIMASI,PHD Director of Economic Analysis, Tufts Center for the Study of Drug Development, Tufts University, Boston, Massachusetts

This study examines what is generally regarded to be the most important measure of innovation in the pharmaceutical industry-the extent to which new drugs are developed and marketed by pharmaceutical firms. Pharmaceutical industry output, as measured by new chemical entity (NCE) approvals in the United States since the 1962 Amendments to the Federal Food, Drug, and Cosmetic Act of 1938, is examined at the firm level. This long-term historical perspective permits us to observe the extent to which this industry has been concentrated with respect to innovative output and how stable company leadership positions have been over time. Databases containing detailed information on all NCEs approved in the United States from 1963 to 1999 and on a large sample of investigational NCEs taken into clinical testing since 1963 were utilized to examine productivity in developing new products at the firm level according to the following stratifications: period of approval, therapeutic class, whether the compounds were self-originated (ie, developed by one firm) or acquired (eg, licensed), and Food and Drug Administration ratings of therapeutic significance. The data indicate that innovation in the pharmaceutical industry is fairly widely dispersed and has become less concentrated over time. Turnover in company rankings based on the number of new drug introductions is substantial. Nonetheless, firms have differed in the productivity of their development programs and some firms have consistently maintained high ranks for their level of innovative output over a lengthy period. Key Words: N C E ; Innovation; Therapeutic class; Output concentration; Success rate

INTRODUCTION THE SUCCESS OF in the phmaceutical industry may be measured by many

Preliminary data from this study were presented at the DIA 35th Annual Meeting. Baltimore, MD, June 28, 1999. Reprint address: Joseph A. DiMasi, PhD, Director of Economic Analysis, Tufts Center for the Study of Drug Development,Tufts University, 192 South Street, Suite 550, Boston, MA 021 11. E-mail: jdimasi@info net.tufts.edu.

yardsticks. A number of factors that contribUte to the productivity of the pharmaceutical industry have already been quantified in the literature. For example, the costs, risks, and length of the development process are important considerations when assessing the vitality of the industry and its prospects for future innovation. However, these factors have been previously analyzed (1,2,3). Rates Of return on new drug development for the industry as a whole have also been investigated (4). Furthermore, trends in the number of drugs

I169

1170

Joseph A. DiMasi

entering clinical testing have been monitored on the determinants of innovation in the pharand analyzed in a series of reports by the maceutical industry in the modem era. Tufts Center for,the Study of Drug DevelopGeneral output statistics by company are ment (CSDD) (5). However, central to both presented herein for the modem era of the financial viability of pharmaceutical firms pharmaceutical development in the United and to advances in public health are the num- States (post-1962 amendments). The extent ber and quality of new drugs that are devel- to which firms have specialized in the maroped and made available to patients. keting of drugs in various therapeutic classes The number of new drugs approved in the is investigated, the degree to which some United States each year is a statistic that is firms have focused on internal development readily available, but to my knowledge no as opposed to licensing-in the innovations of systematic analyses of the long-term patterns others, and metrics related to the quality of of new drug approvals by firm have been output are considered. Measures of the dereported in the literature. This study fills that gree to which pharmaceutical output has void by utilizing Tufts CSDD databases of been concentrated by firm and trends in connew drug approvals and of investigational centration are also analyzed. In addition, the new drugs to examine the degree to which question of whether the R&D programs of output in the industry is concentrated in a some firms have been notably more successsmall number of firms, if and how concentra- ful than others in getting new products to tion levels have changed over time, the extent market is examined. to which firms with relatively high output levels in one period maintain leadership posiDATA AND METHODOLOGY tions in later periods, and variability in the success of firms in bringing drug candidates The Tufts CSDD Approved Drug Database through clinical testing to regulatory ap- contains proprietary and nonproprietary inproval for marketing. formation on new drugs approved in the The reasons that some firms have been United States since 1963. In the current particularly successful in discovering and study, I focus specifically on new chemical developing new drugs are complex and can entities. An NCE is defined as a new molecuinclude such factors as economies of scale lar compound not previously approved in the in research and development (R&D) (6), United States. Excluded are new salts and management practices (7), connectedness to esters of existing compounds, surgical and scientific networks (8,9), and serendipity diagnostic materials, vaccines and other bio(10). Establishing the relative importance of logics, certain externally used compounds these and other factors in explaining the data (such as disinfectants, antiperspirants, and on firm output is beyond the scope of this sunscreens), and nutritional compounds project. The results obtained from this study, (such as natural forms of vitamins and sweethowever, are useful in that they provide a ening agents); this definition differs from that comprehensive picture of how firms have of a new molecular entity (NME) as defined fared in bringing new products to market in by the Food and Drug Administration (FDA) the period following a major paradigmatic primarily in that it excludes diagnostic drugs. shift in the way that pharmaceutical firms Along with other information, the database conduct their business-the result of legisla- provides the name of the company that spontion that directed regulatory authorities to sored the compound’s new drug application require proof of efficacy before granting (NDA), the NDA approval date, the therapeumarketing approval for new drugs (ie, the tic class for the compound’s original ap1962 Amendments to the Federal Food, Drug proved indications, and the therapeutic rating and Cosmetic Act of 1938). This work can, that the FDA assigned to the compound at therefore, aid and encourage further research the time of approval.

Trends in New Drug Innovation by Pharmaceutical Firms

The merger and acquisition history of firms that have obtained NDA approvals was determined from Tufts CSDD databases and public sources. The public information sources included PharmaProjects, the NDA Pipeline, the Medical & Healthcare Marketplace Guide, and various industry trade publications. The primary unit of analysis is the sponsoring company. Because nonproprietary data were used, it was not necessary to aggregate our findings on approvals (ie, results can be shown at the firm level). Innovative output was measured as the number of NCEs approved in the United States from 1963 to 1999. The data were analyzed for both the entire study period and for subperiods. Results are presented both by company affiliations at the time of approval and by the current structure of affiliations. The output of joint ventures where two firms form and share equal interest in a new company was attributed to the new company. Since firms differ in their therapeutic focus, analyses were conducted at the therapeutic class level. To apply a quality filter, the data were also stratified by FDA-assigned therapeutic rating. The FDA initiated its therapeutic rating system in 1976 to prioritize its reviews, but it also applied its system retroactively back to 1963. We utilized these retroactive ratings. Until 1992, the FDA used a threetiered therapeutic rating system; NMEs were given a lA, lB, or 1C rating. The lA, lB, and 1C ratings were assigned to drugs that were thought to represent a significant gain over existing therapy, a moderate gain over existing therapy, and little or no gain over existing therapy, respectively. This system was compressed in late 1992 to a two-tiered scheme where NMEs were rated as either 1P for priority review or 1 s for standard review. For purposes of analysis, we combined NCEs that had received a 1A or a 1B rating with those that had received a 1P rating into a “priority-rated” group; we combined NCEs that had received a 1C or 1 s rating into a “standard-rated’ group. Since firms differ in the extent to which

1171

they rely on licensing-in development projects, the results were also stratified according to whether the compounds were self-originated (developed entirely under the auspices of the sponsoring firm) or acquired (licensedin or otherwise obtained). The source information for NCEs was obtained from surveys of pharmaceutical firms and supplemented where necessary by public sources. Finally, proprietary data from the Tufts CSDD Investigational Drug Database were used to determine clinical success rates by firm for NCEs that had an investigational new drug application (IND) first filed with the FDA during the 1980s. A clinical success rate is defined as the probability that an NCE with an IND filed will eventually obtain marketing approval from the FDA. Clinical success rates were estimated by firm and used to construct a technical productivity index, accounting for differences in therapeutic class and compound source in the investigational drug portfolios of the analyzed firms. Where needed, the survey data were supplemented by public information on the current status of drugs from PharmaProjects and the NDA Pipeline. Given that data for these estimations were collected primarily on a confidential basis, individual firms will be identified only by a code name. The objective of the analysis, then, is to examine whether, and to what degree, scientific productivity (broadly construed) has varied across firms.

RESULTS Pharmaceutical Innovation in the Modern Era The 1962 Amendments to the Federal Food, Drug, and Cosmetic Act of 1938 marked the beginning of a new era of drug development in the United States. The legislation changed the context within which firms in the pharmaceutical industry must compete. Premarketing regulatory requirements increased as firms were required to demonstrate the efficacy of their new products, and a positive review by the FDA of a marketing applica-

I172

Joseph A. DiMasi

tion under the new standards was required before these products could be marketed. As a result, the scope and expense of new drug development were substantially increased. Firms that were best equipped by structure, financial capability, and company culture to adapt to the new environment would have an advantage in the long run. The pharmaceutical output of the drug industry is varied, ranging from line extensions of existing products to new therapeutic compounds; the output includes over-the-counter, diagnostic, and veterinary drugs, as well as human-use prescription drugs. In general, the activity of the pharmaceutical industry that is most innovative and therapeutically significant is the development of new human-use therapeutic compounds. It is this type of output that I focus on here. From 1963 to 1999 pharmaceutical firms received marketing approval in the United States for 691 NCEs. Following enactment of the 1962 amendments, the number of NCE approvals declined substantially relative to preamendment levels ( 11,12). During the postamendments era, the number of approvals generally declined up to and including the early 1970s, but increased from then onward (Figure 1). The upturn is notable during the 1980s and 1990s (The number of NCE approvals were 13.6 and 13.7 per year for 1963

to 1969 and the 1970s, respectively. However, approvals per year increased 35% to 18.5 for the 1980s and an additional 48% to 27.4 for the 1990s). The turnaround in output likely reflects adjustments on the part of firms to both the new institutional rules and to new discovery opportunities arising from research advances in biochemistry and enzymology in the mid-1970s and molecular biology in the 1980s and 1990s (9). Since the 1962 amendments, a number of pharmaceutical firms have merged or been acquired. Several waves of merger and acquisition activity among firms that have engaged in new drug development were evident from my examination of company histories. In particular, activity was greatest in the early 1970s, the late 1980s, and the mid to late 1990s. The consolidation of firms makes any method of attributing output to companies problematic. However, for many purposes a categorization of output according to the identity of the sponsoring parent company at the time that FDA approval was attained is a more informative approach than one that uses current company structure. Classified in this manner, 138 firms sponsored the 691 NCEs approved from 1963 to 1999. The 39 firms that had obtained at least five approvals are listed in Table 1. The data presented in Table 1 are un-

50 40

I

. ..

0 " " " " " " " ' " " " " " '

'

- L L u

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999

FIGURE 1. New chemical entity (NCE) approvals in the United States from 1963 to 1999. The curve is a quadratic fit to the annual approval data. Source of data: Tufts CSDD Approved Drug Database.

Trends in New Drug Innovution by Pharmaceutical Firms

1173

TABLE 1 Firms with at Least Five New Chemical Entities (NCEs) Approved in the United States from 1963 to 1999 (parent company at the time of approval) ~~

Rank

1 2 3 3 5 6 7 8 8 10 11

12 13 14

14 14 14

18 19 20 21 21 21 24 24 26 26 28 28 30 30 30 30 34 34 34 34 34 34

Company

Number of NCEs

Percent of all NCEsO

36 33 31 31 27 25 22 21 21 19 18 17

5.2 4.8 4.5 4.5 3.9 3.6 3.2 3.0 3.0 2.7 2.6 2.5 2.0 1.9 1.9 1.9 1.9 1.9 1.7 1.6 1.4 1.4 1.4 1.3 1.3 1.2 1.2 I .o 1 .o 0.9 0.9 0.9 0.9 0.7 0.7 0.7 0.7 0.7 0.7

Merck Johnson & Johnson Lilly Roche Pfizer American Home Products Upjohn Bristol-Myers Schering-Plough Warner-Lambett Abbott Wellcome Sandoz American Cyanamid Bayer Bristol-Myers Squibb Hoechst SmithKline Beckman SmithKline Beecham Ciba-Geigy Akzo Nobel Glaxo Syntex Nestle Pharmacia & Upjohn Glaxo Wellcome Sterling Boehringer lngelheim Robins Astra Dow Squibb Zeneca Allergan BASF BOC Group Dupont ICI Rhone-Poulenc Rorer

14 13 13 13 13 13 12 11 10 10 10

9 9 8 8 7 7 6 6 6 6 5 5 5 5 5 5

"In total, 691 NCEs, sponsored by 138 firms, were approved in the United States from 1963 to 1999. Four firms had four approvals, 14 firms had three approvals, 18 firms had two approvals, and 62 firms had one approval from 1963 to 1999.

weighted (by medical or commercial significance) counts of the number of successful innovations. The leading four firms (Merck, Johnson & Johnson, Lilly, and Roche) averaged nearly one NCE approval per year, with only five approvals in the aggregate separating them. They can be said to represent a first tier. The leading three firms had stable

research organizations in that their current output count is unaffected by major acquisitions of or mergers with other companies during the study period. The other firms in the top 10 for which this is true are Pfizer and Schering-Plough (As of this writing, Pfizer had announced plans to acquire Warner-Lambert.). The rest of the firms in the

1174

Joseph A. DiMasi

top 10 merged with, acquired, or were acquired by other firms that had obtained NCE approvals prior to the merger or acquisition. Thus, a ranking by the number of approvals that is based on current industry structure can look quite different than the one shown in Table 1. Table 2 provides such a ranking. It lists firms ranked by the number of NCE approvals (for firms with at least five approvals), attributing to parent companies at the end of 1999 all of the past approvals of the companies that the firms had acquired. The acquisition of American Cyanamid and A.H. Robins by American Home Products, the merger of Bristol-Myers and Squibb, and the acquisition of Syntex by Roche lifts these firms to the top three positions in the ranking, with an average of 1.1 approvals per year over the

study period. When using the current parent as the basis for the approval count, the top 10 firms each have at least 32 NCE approvals (The count for Johnson & Johnson is lower by one in Table 2 than in Table 1 since one of its NCEs was developed by a subsidiary, Iolab, that was divested to Chiron in 1995.). None of the firms have a large share of the total number of NCEs. The top 10 firms, when considering the parent at the time of approval, have shares that range from 2.7% to 5.2%. When using the current parent as the basis for analysis, the shares of the top 10 firms are still low and even more uniform, ranging only from 4.6% to 6.5%. We would expect some degree of correlation between physical output measures and sales levels. The innovation shares of the leading firms are in fact similar to reported recent market

TABLE 2 Firms with at Least Five New Chemical Entities (NCEs) Approved in the United States from 1963 to 1999 (current parent company) Rank

1 2 3 4 4 6 6 8

9 10

11 12 13 13 15

15 17 18 19 20

20 20 20 20 25

Company American Home Products Bristol-Myers Squibb Roche Aventis Pharmacia & Upjohn Merck Novartis Glaxo Wellcome SmithKline Beecham Johnson & Johnson Lilly Pfizer Schering-Plough Warner-Lambert Abbott Astrazeneca Bayer Akzo Nobel Nestle Allergan BASF Boehringer lngelheim Dupont Merck Monsanto BOC Group

Number of NCEs

Percent of all NCEs"

45 43 41 38 38 36 36 35 34 32 31 27 21 21 18

6.5 6.2 5.9 5.5 5.5 5.2 5.2 5.1 4.9 4.6 4.5 3.9 3.0 3.0 2.6 2.6 2.0 1.4 1.3 1.2 1.2 1.2 1.2 1.2 0.7

18

14 10

9 8 8

8 8

8 5

"In total, 691 NCEs, sponsored by 94 firms, were approved in the United States from 1963 to 1999. One firm had four approvals, eight firms had three approvals, 12 firms had two approvals, and 47 companies had one approval from 1963 to 1999.

1175

Trends in New Drug Innovation by Pharmaceutical Firms

shares based on worldwide pharmaceutical sales (There is reason to suspect that the market shares of large firms would tend to be greater than their output shares. DiMasi et al. (6) found a tendency for the commercial significance of NCE output to increase with firm size for a sample of fully integrated pharmaceutical firms. However, other things being equal, the market shares of innovator firms will tend to be lower than output shares because of generic competition.). For 1998, the top 10 firms ranked by share of the worldwide pharmaceutical market (Table 3) had sales shares that ranged from 2.9% for Lilly to 5.1% for Merck (Although merged firms have gained market share, the share of worldwide sales obtained by leading firms has remained relatively stable. The combined worldwide market share of the top 30 firms has varied between 52% and 57% over the last decade [ 131. Whether the recent spate of mergers and acquisitions has been successful

along various dimensions is unclear at this time. A survey of the parties most affected by the merger and acquisition activity-customers, employees, and investors-revealed that the groups differed in their assessments of how successful individual mergers and acquisitions have been [14]. Detailed analysis, and perhaps more time, are needed to fully assess the impact of pharmaceutical industry consolidation.). Seven of the top 10 firms and 18 of the top 20 firms ranked by pharmaceutical market share are among the top 10 and top 20 developers of NCEs, respectively. Similar results are obtained if the NCE ranking is made for more recent approvals (1982 to 1999).

Interfirm Heterogeneity in the Nature of Pharmaceutical Innovation The aggregate data on NCE output are informative, but they can mask important differ-

TABLE 3 Worldwide Pharmaceutical Sales by Company" Rank

1 2 3 4 5 6 7 8 9 10 11

12 13 14 15 16 17 18 19 20

Company Merck Aventisb Glaxo Wellcome Astrazenecab Bristol-Myers Squibb Pfizer Novartis Roche American Home Products Lilly Johnson & Johnson SmithKline Beecham Schering-Plough Pharmacia & Upjohn Warner Lambert Abbott Sanofi-Synthelabob Bayer Takeda Boehringer lngelheim

Pharmaceutical Sales ($ million)

15,296.5 13,608.1 13,230.5 12,754.0 12,573.0 12,230.0 1 1 , I 74.8 9,921.5 8,901.8 8,590.4 8,562.0 7,701.6 6,695.0 6,127.0 5,604.0 5,602.0 4,832.7 4,823.7 4,567.8 4,487.3

"Source: Scrip's 1999 Phar'maCeUtiCal Company League Tables, Surrey: PJB 1999. bHoechstand Rhone-Poulenc Rorer, Astra and Zeneca, and Sanofi and Synthelabo merged during 1999. Sales during 1999 for each of the firms in a merger were combined for the table.

1176

ences across firms in the extent to which they have diversified their innovative efforts, in the degree to which they have been self-sufficient in innovation, and in the quality of their innovation. Although the causes and implications of such differences are difficult to fully establish we can, at least as a first step, examine the available data to determine if firms do indeed differ in these dimensions.

Joseph A. DiMasi

experienced substantial growth in relative output; 23.7% of the NCE approvals from 1982 to 1999 were in the cardiovascular area. From chemical and pharmacological perspectives, the types of compounds developed in a therapeutic class do, however, change over time. For example, in cardiovascular drug development, approvals of drugs within major pharmacologic classes proceeded (with considerable overlap) from diuTherapeutic Focus. The demand for pharma- retics to beta-blockers, followed by calcium ceutical products depends on the uses to channel blockers, ACE-inhibitors, and which drugs are put. Drugs in different thera- HMG-CoA reductase inhibitors. Antiinfecpeutic classes generally are not substitutable tive development has progressed with the in consumption. Even drugs within the same marketing of successive generations of antigeneral class are not necessarily interchange- biotics and, with the advent of the AIDS able. The relevant markets are, therefore, de- crisis, a heightened interest in recent years termined by the diseases or conditions that in the development of antiviral drugs. Even the drugs are used to treat. While output and though the modes of action for various commarket shares can appear quite low when pound types within a therapeutic class can considering pharmaceutical products as a differ substantially, many of them share comwhole, they may be much higher for markets mon indications. defined in therapeutically meaningful ways. 1 examined how NCE output is distributed Consequently, many firms may choose to fo- across firms for the five largest therapeutic cus their efforts on developing competencies classes (Table 4). In three of these five classes in specific therapeutic areas. Prior success in the leading firm has been much more prolific particular fields and program-specific exper- than any of its competitors. For antineoplastise resident at the firm because of historical tic NCEs, Bristol-Myers has far outpaced reasons can also result in trends in innovation other firms, with more than twice as many at the firm level that are at least partially approvals as its nearest competitor. The Brisdependent on the paths that individual firms tol-Myers advantage is more than three-fold have followed, whereby some firms may when one adds the four approvals of the achieve dominant positions in developing merged firm, Bristol-Myers Squibb. Merck drugs in one or several therapeutic catego- and Johnson & Johnson obtained approxiries. mately twice as many NCE approvals as their It is important, therefore, to examine how nearest competitors in the cardiovascular and new drug innovation is distributed across analgesic/anesthetic classes, respectively. firms at the therapeutic class level. Over the The advantages that the premier innovator post- 1962 amendments period, nearly half in a therapeutic area has over competitors (43.4%) of NCE output has been in two ther- need not persist over time. To compete more apeutic areas: antiinfective and cardiovascu- effectively in a given therapeutic area, firms lar (Figure 2). Advances in biomedical sci- that lag the leader can build stronger proence from the late 1970s onward conceivably grams through expansion and by strengthcould have resulted in a wave of pharmaceu- ening their internal capabilities. It is also tical innovation that was associated with ma- possible to acquire capabilities and diminish jor shifts in therapeutic focus. However, the competitive advantages in therapeutic areas distribution across the eight therapeutic through mergers and acquisitions. Thus, Taclasses for the 1980s and 1990s is very simi- ble 5 demonstrates that the output advantages lar to that for the whole period. The one that the leaders in the analgesidanesthetic exception is the cardiovascular class, which and cardiovascular areas have over other

Trends in New Drug Innovation by Pharmaceutical Firms

1177

23.4%

Antiinfective Cardiovascular Central Nervous System Analgesklhesthetic Antineoplastic Endocrlne Respiratory Gastrointestinal Percent of NCE Approvals

FIGURE 2. New chemical entity (NCE) approvals In the United States from 1963 to 1999 by therapeutic category. Source of data: Tufts CSDD Approved Drug Database.

firms are greatly diminished when approvals are allocated to the current parent. Additionally, on either a parent company at the time of approval or on a current parent basis, the leading firms in all therapeutic categories had relatively small shares of all innovative output in the category (Shares of drugs used for specific indications would necessarily be higher than shares of drugs in the broader class.). In-house versus Acquired Innovation. Quan-

tifying the number of NCEs that individual firms have sponsored for approval does not, in and of itself, indicate whether the sponsoring firms were completely, or even largely, responsible for the development of the drugs. A substantial number of new drugs approved in the United States have been licensed-in or otherwise acquired by the sponsoring firms. Firms may differ in their strategic approaches to licensing and, therefore, in the extent to which they engage in this activity or in their decisions about when in the development process they tend to acquire compounds. Therefore, the amount of development that the sponsoring firms conduct on these compounds is quite variable. One way to gauge the degree to which firms have relied on and have been productive at discovering and developing their own drugs is to analyze drug approvals by origin. Of the 691 NCEs approved in the United States from 1963 to 1999, 61.8% were self-

originated. The proportion of NCE approvals that were self-originated has declined over time, decreasing from 71.6% for NCEs approved from 1963 to 1969 to 60.9% for NCEs approved in the 1990s. Most of the decline occurred shortly after the 1960s. Eight of the top 11 firms ranked by the number of self-originated NCE approvals (Table 6) were in the top 10 for the total NCE count (Table 1). As a group, the top firms in terms of the number of self-originated NCE approvals are more reliant on internal development than is the industry as a whole. The share of approvals that were for self-originated NCEs ranged from 57.9% for Warner-Lambert to 100% for Hoechst. The self-originated share for the group of 11 firms is 72.7%. The degree to which innovator firms marketed their own products varied somewhat across the major therapeutic categories. The percent of self-originated NCEs varied from 55% for gastrointestinal NCEs to 75% for respiratory NCEs. Firms that were dominant within a therapeutic class in terms of total NCE approvals also tended to have a relatively large number of the self-originated approvals in the class. For example, in the cardiovascular class Merck had 11 self-originated approvals compared to five for the next highest competitor. A notable exception is the antineoplastic class, since many oncology drugs have been initially identified and developed by the National Cancer Institute.

9 8 8 6

Roche

Bristol-Myers Johnson & Johnson Bayer

4

4 4

3

3 3 3 3 3 3

Astra Dupont Glaxo Nestle Schering-Plough Syntex

Upjohn Wellcome

Pharmacia & Upjohn

Roche

Bristol-Myers

3 3

Lilly Abbott Boehringer lngelheim Sandoz Upjohn

AmericanHome Products Pfizer

4 3

Merck

9

NCEs Companyd

Cardiovascular

4 4

4 4

5

6

7

14

4 4

Johnson & Johnson Sandoz Ciba-Geigy Merck A.H. Robins Wellcome

3 3 3 3

5 5 5

5

7

7

NCEs

Lilly Pfizer Warner-Lambert

Abbott

American Home Products Roche

NCEs Company"

Central Nervous System

"Eleven companies had two approvals and 19 companies had one approval. bFivecompanies had four approvals, five companies had three approvals, four companies had two approvals, and 26 companies had one approval. "Twenty-four companies had one approval. ?en companies had three approvals, 11 companies had two approvals, and 36 companies had one approval. 'Nine companies had two approvals and 20 companies had one approval.

Merck Schering-Plough Warner-Lambert Wellcome American Cyanamid Bristol-Myers Squibb Glaxo Wellcome Upjohn

11

American Home Products Akzo Nobel BOC Group Abbott

Pfizer

5

13

Merck

Lilly

Antineoplastic NCEs Company'

9

NCEs Comanyb

Antiinfective

Johnson & Johnson

Company"

AnalgesiclAnesthetic

TABLE 4 Leading Firms in the Number of United States New Chemical Entity (NCE) Approvals from 1963 to 1999 by Therapeutic Category (parent company at the time of approval)

?

4

BOC Group

Roche Aventis Glaxo Wellcome Novartis

AstraZeneca

Pharmacia & Upjohn

Bristol-Myers Squibb

4 3 3 3

4

9

13

BASF Bristol-Myers Squibb Lilly Roche Abbott Boehringer lngelheim

Novartis Pharmacia & Upjohn Pfizer SmithKline Beecham

American Home Products Aventis

Merck

"Nine companies had three approvals, three companies had two approvals, and nine companies had one approval. Three companies had three approvals, four companies had two approvals, and 11 companies had one approval. 'Seven companies had two approvals, and 13 companies had one approval. "Three companies had three approvals, nine companies had two approvals, and 20 companies had one approval. "Three companies had three approvals, three companies had two approvals, and 14 companies had one approval.

12 12 11 10

Roche SmithKline Beecham Pfizer American Home Products Johnson & Johnson Aventis Pharmacia & Upjohn Warner-Lambert Merck Schering-Plough

5 5

5 4

13

Liily

6

American Home Products AstraZeneca Merck Roche Akzo Nobel

15 14

7

Glaxo Wellcome

Bristol-Myers Squibb

4 4

5 5 5 5

8 8 6 6

8

12

14

5

Warner-Lambert Glaxo Wellcome Johnson & Johnson SmithKline Beecham

4 4 4

7 5 5 5

8

8

12

Roche Abbott Liily Pfizer

Novartis

American Home Products Aventis

NCEs

Central Nervous System

NCEs Company"

Cardiovascular

NCEs Companyd

Antineoplastic NCEs Company'

Glaxo Wellcome

8

~~~

Antiinfective

NCEs Comanyb

Johnson & Johnson

Company"

AnalgesidAnesthetic

TABLE 5 Leading Firms in the Number of United States New Chemical Entity (NCE) Approvals from 1963 to 1999 by Therapeutic Category (current parent company)

1180

While Bristol-Myers Squibb had substantially more approvals in this class than did any other firm, all of its antineoplastic approvals were acquired. Quality of Innovation. The data on NCE approvals presented thus far have not been adjusted for or categorized by quality. One measure of the medical significance of the output of pharmaceutical company R&D programs is the therapeutic rating assigned by the FDA to NMEs approved in the United States. The FDA makes these assignments for the purpose of prioritizing its reviews. Although the ultimate medical significance of some drugs is unknown at the time of approval since some new uses are not discovered or established until the drugs are used in everyday medical practice or until postapproval clinical testing is conducted, there is likely a strong positive correlation between the FDA ratings and medical significance (A number of other measures of the medical, scientific, and commercial significance of new drugs have been used in studies of the drug development and approval processes. For example, aside from FDA therapeutic ratings, Dranove and Meltzer [15] examine measures based on the number of citations to a drug in medical textbooks, in medical journals, and in subsequent patent applications. They also examine the number of major markets in which a drug is introduced and the drug’s United States sales, but these measures correlate more with a drug’s commercial significance than with its medical significance.). Our data suggest a positive correlation between a firm’s total number of NCE approvals and the number of the firm’s NCE approvals that received a priority rating. Nine of the top 11 firms ranked by approval of NCEs with priority ratings (Table 6) are also among the top 10 firms ranked by total NCE approvals (Table 1). (Therapeutic ratings were not available for three NCEs approved in the 1960s. Lilly, Bristol-Myers, and ScheringPlough each sponsored one of the three. The Lilly NCE was self-originated, while the other two NCEs were acquired.). The rela-

Joseph A. DiMasi

tionship between the number and the proportion of approvals with a priority rating that a firm receives is much weaker. For the firms in Table 6 ranked by number of NCEs with a priority rating, the share of their approvals that had received a priority rating is only slightly higher than for the industry as a whole (55.0% compared to 49.1%). However, the averages do not reveal the substantial variability that existed among firms in their propensity to produce products with high therapeutic ratings. For the firms listed in Table 6, the share of approvals with a priority rating ranged from 37.0% for Pfizer to 88.2% for Wellcome. Although the hypothesis that research programs that produce proportionately more approvals with priority ratings tend to be more profitable is plausible, further analysis would be required to establish this for particular firms for at least four reasons. First, while medically significant products are often commercially significant, this is not always the case. For example, FDA-assigned therapeutic ratings tend to be higher for orphan drugs than for other drugs (16,17) (Under the US Orphan Drug Act (Public Law No. 97-414, 96 Stat 2049 [1983]) as amended, an orphan drug is one that targets a disease or condition that affects less than 200000 people in the United States, or one for which the manufacturer can demonstrate that there is no reasonable expectation that development and marketing costs can be recouped from sales in the United States.). The sales of the vast majority of orphan drugs are relatively small (1 6). Secondly, DiMasi et al. (1) found that clinical development costs tend to be higher for drugs that had received higher FDA therapeutic ratings. Third, most investigational drugs fail in testing and firms can differ in their approval success rates (6). Thus, other things being equal, some firms may get their drugs on the market at a higher R&D cost per approval (6). Fourth, firms that licensein drugs that receive high therapeutic ratings will have to share some of the returns with the originators. The extent to which firms have either fully or partially developed NCEs that have

Trends in New Drug Innovation by Pharmaceutical Firms

1181

TABLE 6 Leading Firms in the Number of United States New Chemical Entity (NCE) Approvals from 1963 to 1999 by Source and by Medical Significance (parent company at the time of approval) Self-Originated" Company

Self-originated Priority-Rated

Priority-Ratedb

NCEs Company

NCEs Company

Merck Johnson & Johnson Liily Pfizer Roche Hoechst Upjohn Sandoz Schering-Plough

28 25 25 21 20 13 13 12 12

Merck Roche Johnson & Johnson Wellcome Lilly Upjohn Bristol-Myers Pfizer Schering-Plough

25 22 17 15 12 12 10 10 9

Wellcome Warner-Lambert

12 11

Warner-Lambert SmithKline Beckman

9 8

Merck Roche Johnson & Johnson Lilly Wellcome Upjohn Pfizer Warner-Lambert BoehringerIngelheim Hoechst Sandoz

NCEs 20 14 13 11 10 8 7

7 5 5 5

aDevelopedentirely under the auspices of the company. bCompounds that received a 1A. 1B, or 1P therapeutic significance rating by the Food and Drug Administration at the time of marketing approval.

received priority ratings can also be measured with our data (Table 6). Although the order is slightly different, the six leading firms in terms of the number of self-originated NCE approvals with priority ratings are also the six firms with the most approvals of drugs with priority ratings. The firms that developed the most self-originated NCEs with priority ratings tended to have a higher than average propensity to originate those of their approved NCEs that receive a priority rating; while 60% of all NCEs with a priority rating were self-originated, the share for the 11 leading firms was 70% (Among these leading firms, the share of priority-rated NCEs that were self-originated varied from 65% for Roche to 100% for Boehringer Ingelheim, Hoechst, and Sandoz.). The data in Table 6 also suggest that some firms that have refrained from engaging in major merger and acquisition activity have fared well in terms of in-house R&D and in bringing medically significant products to market. Johnson & Johnson, Lilly, and Merck all fit this characterization. Even when the data are analyzed on a current parent

basis, these firms retain high ranks along these dimensions (Table 7).

Concentration of Pharmaceutical Output The industrial organization literature is replete with studies that utilize measures of the degree to which markets are dominated by relatively few firms. These measures are often used to describe and analyze industry structure, with putative implications for economic efficiency in these markets. Industry concentration measures are usually defined in terms of sales, but employment and assetbased measures have also been used (See Scherer and Ross [ 181 for a good discussion of the use and limitations of concentration measures. Grabowski and Vernon [19] review a number of studies that have used concentration measures in analyses of the competitiveness of the pharmaceutical industry.). The most commonly used measures are the concentration ratio (especially, the four-firm and eight-firm ratios) and the HefindahlHirschman index (HHI) (A concentration ratio is the aggregate share of the market held

Joseph A. DiMasi

1182

TABLE 7 Leading Firms in the Number of United States New Chemical Entity (NCE) Approvals from 1963 to 1999 by Source and by Medical Significance (current parent company) Self-Originateda Company

Self-originatedl Priority-Rated

Priority-Ratedb

NCEs Company

~

~~

Merck Roche Novartis Aventis Lilly Pharmacia & Upjohn SmithKline Beecham Johnson & Johnsonc Glaxo Wellcome ffizer

28 28 26 25 25 25 25 24 23 21

Bristol-Myers Squibb

20

NCEs Company ~

Roche Glaxo Wellcome Merck Pharmacia & Upjohn Aventis Bristol-Myers Squibb Novartis Johnson & JohnsonC SmithKline Beecham American Home Products Lilly

~

NCEs ~

26 25 25 21 19 18 17 16 16 13

Merck Roche Glaxo Wellcome Aventis Pharmacia & Upjohn Johnson & Johnsonc Lilly Novartis SmithKline Beecham ffizer

12

Warner-Lambert

20 17 15

14 14 12 11 11 11

7 7

"Developed entirely under the auspices of the company. bCompounds that received a 1A, 1B, or 1P therapeutic significance rating by the Food and Drug Administration at the time of marketing approval. 'Johnson & Johnson has one fewer approval in each category than when considered on a parent company at the time of approval basis since it divested a subsidiary that had sponsored and retained ownership of one approved NCE.

by a given number of the leading firms in an industry. Thus, the four-firm sales concentration ratio is the percent of industry sales earned by the four firms with the highest sales. The Herfindahl-Hirschman index (20) is an alternative summary measure that accounts for the sales shares of all of the firms in the industry. It is defined as H H I = Zy=,= Sz, where S, is firm i's share of industry sales and N is the number of firms in the industry. Higher values indicate greater concentration. Typically, the percentage shares are multiplied by 100 so that the index can assume values up to 10000 [the case of a pure monopolist]. The index decreases with more firms in the industry and increases with greater inequality in firm shares for a given number of firms.). The data on NCE approvals can be used to provide measures of industry concentration in innovation. The concentration measures that have been applied to sales are applicable, without change in format, to innovative output. While concentration measures based on company sales can, in most instances, be sen-

sibly applied to data from a single year, using the NCE output from any one year to compute a concentration measure makes little sense. The development process is lengthy and variable. Most firms engaged in new drug development will not have an NCE approved every year. Grouping of years is, therefore, appropriate and necessary. Concentration measures of any type for a given year or period are only rough static indicators of industry structure. In theory, some industries with relatively high measured concentration may be quite competitive and some with relatively low measured concentration may exhibit oligopolistic behavior (18). Institutional knowledge of the industry and of the nature of competition in that industry is often a prerequisite to understanding the structure, behavior, and performance of the industry. However, measurement of the degree of concentration in an industry at different points in time will often serve as an accurate indicator of a trend in concentration (The explanation for a change and the economic significance of a change in concentration can

Trends in New Drug Innovation by Pharmaceutical Firms

vary from case to case. A concentration or deconcentration trend might, for example, reflect changes in regulatory policy, shifts in demand, or the emergence of new technologies that are initiated by new entrants or that some existing firms are better suited to adopt than are others.). The four-firm and eight-firm innovation concentration ratios in Figure 3 are very similar to reported sales concentration ratios for the pharmaceutical industry. The United States Census Bureau has published fourfirm sales concentration ratios for pharmaceutical preparations (Standard Industrial Classification [ S.I.C.] 2834) that vary from 22% to 28% between 1947 and 1992; the eight-firm sales concentration ratios vary from 36% to 45% (21) (The Census Bureau has published manufacturing industry concentration ratios for 1947, for years between 1947 and 1977 at irregular intervals, and at five-year increments thereafter. Grabowski and Vernon 1221 have reported concentration ratios for US ethical drug sales to retail stores and hospitals for every year from 1958 to 1973. The ratios are generally within the range of Census values.). The pharmaceutical industry concentration ratios are moderate in comparison to the sales concentration

1183

ratios for other industries. (The Census Bureau has published four-firm sales concentration ratios for 456 four-digit S.I.C. industries for 1992 [21]. The ratio for the pharmaceutical preparations industry is 26% for 1992; examination of the Census data shows that 69.3% of the four-digit industries had a higher concentration ratio. For 1992,43% of the industries had a four-firm concentration ratio of at least 40%. Similarly, Scherer and Ross [ 181report that 44% of 448 manufacturing industries with four-digit S.I.C. codes had four-firm concentration ratios of 40% or higher for 1982. Some scholars interpret ratios this high, when the market is appropriately measured, to be an indication of the existence of market power. For example, Scherer and Ross [18, p. 821 state, “When the leading four firms control 40 percent or more of the market, oligopolistic behavior becomes likely.”). The output concentration ratios indicate a modest deconcentration trend for the 1980s and 1990s relative to the 1960s and the 1970s (Figure 3). The HHI values also suggest a deconcentration trend, albeit one that is more clearly continuous across periods (The HHI values for NCE approvals are 407, 385, 324, and 223 for 1963 to 1969, 1970 to 1979,

60

cn

8

50

88 “ z

b

30 20

8..

E

g

10

Q)

n

0 1963-1969

1970-1979

1980-1989

1990-1999

Period of Approval 04-Firm Ratio (parent at approval) n 4-Finn Ratio (current parent) &Firm Ratio (parent at approval) W &Firm Ratio (current parent)

FIGURE 3. Concentration of output In the pharmaceutical industry. Share of total United States new chemical entity (NCE) approvals by the leading four and eight firms in number of approvals for the period. Source of data: Tufts CSDD Approved Drug Database.

1184

1980 to 1989, and 1990 to 1999, respectively. The trends for either concentration ratios or HHI values are the same whether approvals are attributed to the parent companies at the time of approval or to the current parents. An ideal measure of output concentration should yield values between those given by the two approaches, but measurement on a parent at the time of approval basis should be closer to the ideal. When grouped on a current parent basis, some approvals would be attributed to firms that merged decades after the approvals were obtained.). In contrast, neither the sales concentration ratios nor the sales HHI values exhibit a trend (The US Census Bureau [21] has published sales HHT values [for the 50 largest firms in an industry] beginning with the 1982 Census of Manufacturers. The index values for pharmaceutical preparations are 318,273, and 341 for 1982, 1987, and 1992, respectively.) However, given lengthy product lifecycles in this industry, (For their estimations of rates of return for NCE development, Grabowski and Vernon [ 4 ] use a 20-year product lifecycle.) concentration trends in sales may lag concentration trends in new product introductions. Therapeutic Class Concentration. One may argue that concentration measures that are based on shares of the sales of all pharmaceutical products or on shares of all NCE approvals are seriously biased downward, since economically meaningful markets are more appropriately defined at narrow therapeutic class, or even at specific indication, levels. Given that drugs have disparate uses, this argument is generally valid for the demand side of the market. However, the scope of the relevant market is appropriately defined only when substitution possibilities on both the production and consumption sides of the market have been considered (23). If firms can allocate resources to new therapeutic programs or expand existing programs in a relatively short period without substantial retraining or new hiring, then concentration in innovation can be measured appropriately at a fairly broad level. Concentration measures applied to thera-

Joseph A. DiMasi

peutic classes are, as expected, higher than for all NCE approvals. The measures are not, however, markedly higher for the larger therapeutic classes, and the classes generally exhibit a downward trend in concentration, as was noted for all NCE approvals. For example, the four-firm concentration ratio for cardiovascular NCEs fell from 61.6% for the 1960s to 23.3% for the 1990s (The cardiovascular NCE four-firm concentration ratios NCEs were 44.5 and 27.7 for the 1970s and the 1980s, respectively. Similarly, the cardiovascular NCE HHI values were 1243, 864, 412, and 328 for the 1960s, 1970s, 1980s, and 1990s, respectively.). However, the period during which notable deconcentration is first manifested varied by therapeutic class. While innovation in the cardiovascular area experienced substantial deconcentration throughout the study period, it was not until the 1990s that this occurred for antineoplastic NCEs (The four-firm concentration ratios for antiinfective, antineoplastic, and central nervous system NCEs for the 1990s were 32.8%, 28.3%, and 25.7%, respectively. The four-firm concentration ratio for analgesic/ anesthetic NCEs fell modestly from 41.6% for the 1960s to 32.4% for the 1990s. The patterns of change over time by therapeutic class are identical if the data are grouped by current parent.). As noted above, the breadth of the market definition on which concentration in innovation is best measured depends on substitution possibilities in production. If leading firms can conduct successful R&D in a number of therapeutic programs, then substantial dynamic competition can exist in therapeutic markets even if static concentration measures suggest a relatively high degree of market power. Evidence of a market presence in a variety of therapeutic areas would suggest this kind of substitutability (Several studies have found that major pharmaceutical firms in the 1960s and early 1970s were well diversified across therapeutic categories in terms of patents, sales, and worldwide NCE introductions [24,25].). Although the data on NCE approvals do not directly measure the degree of research effort across therapeutic catego-

Trends in New Drug Innovation by Pharmaceutical Firms

ries, they can be used to examine the degree to which different firms have been successful in reaching the market in a variety of therapeutic areas. Table 8 indicates the number of NCE approvals that leading firms have had in eight major therapeutic categories since 1982. Of the 16 firms, one had approvals in all eight categories, one had approvals in seven categories, eight had approvals in six categories, and four had approvals in five categories. Excluding the catch-all “other” category, the mean number of categories with approvals for these firms is 5.6 (median = 6). The data in Table 8 suggest a substantial amount of diversification across therapeutic categories by leading firms. However, the number of categories in which a firm has obtained approvals does not necessarily indicate how concentrated its output is in those categories. For example, 4 1% of Merck’s approvals are in the cardiovascular class and 50% of Bayer’s approvals are in antiinfectives. The degree to which company approvals are diversified across categories can be measured by a “numbers equivalent index” (Ravenschraft and Scherer [26] uses a variant of the Herfhdahl-Hirschman index to measure the diversification of companies across different lines of business. The index is defined as llXIF,;, where Fvis the share of company i’s sales that arise from the company’s j’th line of business. In the context herein, sales is replaced by number of NCE approvals and line of business is replaced by therapeutic category. If the company has approvals in only one category, then the index value will be one. The index cannot exceed the number of categories in which the firm has approvals. The index will be higher the more equal are the firm’s shares [ie, the more diversified are the firm’s approvals].). Table 9 presents diversification index values for 15 of the 16 firms listed in Table 8 for 1963 to 1980 and 1982 to 1999 (with the “other” category included as a separate class). By this measure, for 1982 to 1999 the most diversified firms were Johnson & Johnson, Hoechst, and Schering-Plough. The least diversified firms were Bayer, American Home

1185

Products, and Wellcome (These data need to be interpreted with some care. In the case of Merck the number of categories in which approvals were obtained increased and nearly all classes were covered. The relatively low diversification index value is due to a high concentration of approvals in one category [cardiovascular]. The disparity between diversification as measured by the number of categories and as measured by the index is the result of a particularly successful program in one area. See Galambos and Sturchi0 [27] and Vagelos [28] for discussions of the history of Merck’s development of cardiovascular drugs.). The data in Table 9 indicate a general increase in diversification over time. Eleven of the 15 firms had a higher diversification index value for the later period. Similarly, 12 of the firms had approvals in more thera; peutic categories in 1982 to 1999 than in 1963 to 1980, while none had approvals in fewer categories. The mean diversification index value increased from 3.5 1 (median = 3.27) for 1963 to 1980 to 4.68 (median= 4.64) for 1982 to 1999. The number of categories per firm increased from 4.7 for 1963 to 1980 to 6.3 for 1982 to 1999. The increased diversification in approvals is likely associated with greater diversification in discovery programs (although we lack the data to demonstrate the relationship). If this is so, then the increased output of the pharmaceutical industry over time may be partially explained by knowledge spillovers across categories. Henderson and Cockburn (29) have, for example, found support for the hypothesis that the productivity of drug discovery research (with output measured by counts of important patents) can be substantially enhanced by the existence of such externalities (both within the firm and across firms). Firm Turnover. Product innovation is the primary mechanism by which dynamic competition in the pharmaceutical industry is exhibited. New drugs typically supplant old ones in medical use. Substantial returns can be earned on new products, and, by Schumpeterian logic (30), these potential returns

3 4 4 5 3 2 1 2 2

3

1 2 0 3 0

0 2 0

3 2 1

2

0

2 1 0

2

1

2

1 2

3

2 0

1 1 3

1

0 1

0 1

1

1

1

0 2 0

3

1

0 0 0

0

1

0 0 0

0

0

0 1 1

0

0

0 1

1

1 0 1

0

1

0 0

0

2

0

1 0

1

1 1

~~

Respiratory

2

Gastrointestinal

0

1

2 2 0

2

3

2

1

4 2

2

~

2 1 0

5

2

1 1

5

0

1

2

2

4

4

1

0

Central Nervous System Endocrine

3 3

Cardiovascular

9 2 2

Antineoplastic 0 2 1 0

4 4 6 6

3 5 1

Antiinfective

0 0

1

1

0 0

0 0

0

2 0

1

1 1 3 0

Other

9 9 9

12 11 10 10 10

13 13 12

13

16

17

22 21

Total

~

'Prior to their acquisition by American Home Products, American Cyanamid had four approvals in three of eight categories and A.H. Robins had one approval. bPriorto their merger in 1989, Bristol-Myers had six approvals in four of eight categories and Squibb had one approval. 'Prior to their merger, SmithKline Beckman had five approvals in three of eight categories (and one approval in the "other" category) and Beecham had two approvals in one category. Sterling, which was acquired by SmithKline Beecham in 1994. had four approvals in three categories prior to the acquisition during the 1982 to 1999 period. '%'elcome was acquired by Glaxo in 1995. "Prior to their merger in 1995, Pharmacia had two approvals in two of eight categories, and Upjohn had six approvals in five categories.

Merck Johnson 8 Johnson Roche Pfizer American Home Productsa Bristol-Myers Squibbb Lilly Abbott SmithKline Beecham' Wellcorned Bayer Glaxod Hoechst Pharmacia 8 Upjohn' Schering-Plough Warner-Lambert

Company

Analgesic/ Anesthetic

TABLE 8 United States New Chemical Entity (NCE) Approvals from 1982 to 1999 by Therapeutic Category (parent company at the time of approval)

zg.0

?

2

L

4

9

10

12 17 6 21 13 6 2 2 15

10

13 11 12

3 6 7 3 2 2 5 5 4

6

6 5 7 4 5

4.83 3.90 5.14 2.78 3.60 2.60 2.57 3.27 5.45 3.00 2.00 2.00 4.67 3.57 3.24

Diversification Index

21 9 9

10 10

20 21 11

11

22 21 17 16 13 13

7 9 7 6 5 7 6 6 8' 5 3 7 7 6 5

Number of Number of Approvals Categories

4.17 6.78 4.74 4.13 3.60 5.12 5.14 4.35 4.64 3.90 2.63 6.25 5.09 5.40 4.26

Diversification Index

The firms listed are 15 of 16 firms with the most NCE approvals from 1982 to 1999. Glaxo is not included here since it had no U.S. NCE approvals during 1963 to 1980. 'Bristol-Myers Squibb, Pharmacia & Upjohn, and SmithKline Beecham were among the top 16 firms with approvals during 1982 to 1999, but they did not exist as the merged firms that they currently are in the earlier period. To provide a meaningful comparison over time, the figures listed for these three firms for both periods are computed on a current parent basis. The data include three approvals during 1982 to 1999 that were sponsored by the Allergan subsidiary of SmithKline Beckman. Allergan was divested when SmithKline Beckman and Beecham merged.

Merck Johnson & Johnson Roche Pfizer American Home Products Lilly Abbott Bristol-Myers Squibbb SmithKline Beechamb Wellcome Bayer Hoechst Pharmacia & Upjohnb Schering-Plough Warner-Lambert

Company

Number of Number of Approvals Categories

1963-1 980

1982-1 999

TABLE 9 Company Diversification of United States New Chemical Entity (NCE) Approvals by Therapeutic Category (parent company at the time of approval)"

s

3

E

0

n E, -.

5.

'tr

5

a

2. 0

2

3 a

a

E

P

$

5.

4

1188

should induce firms to conduct R&D on new drugs. Although not a necessary condition, this hypothesis can be supported by evidence of turnover in company rankings and shares of innovative output (See Cocks [24] for evidence of market share turnover for leading pharmaceutical firms in the 1960s and early 1970s.). Table 10 lists the 13 firms that ranked in the top 10 of NCE approvals for 1963 to 1969, and traces their changes in rank and share over time. The data indicate substantial turnover. Only five of the 13 firms had a top 10 ranking in the 1990s. Only four of the firms (Johnson & Johnson, Merck, Pfizer, and Roche) maintained a top 10 ranking in all periods. While the average rank of the 13 firms was 5.6 for the 1960s, the average ranks for the 1970s, 1980s, and 1990s were 10.2, 10.6, and 15.4, respectively. Similarly, the average output share for these firms was 4.8% in the 1960s, but 3.4%, 3.6%, and 2.3% for the 1970s, 1980s, and 1990s, respectively (These results are likely partly due to the general deconcentration trend noted above, but they are also due to the changes in rank for these particular firms.). These results suggest both that a firm’s success in developing new drugs in one period is no guarantor of its future success, and that some firms have been able to maintain relatively high levels of innovation over long periods. Further evidence of the dynamic nature of competition in the pharmaceutical industry can be ascertained from examining the full distribution of approvals by firm. The frequency distributions of firms by number of NCE approvals for three 10-year periods are shown in Figure 4. Increasing technological opportunities have not only been followed by increasing numbers of new drug approvals, but the evidence suggests that advances in biomedical science have fostered deconcentration in new drug development among existing firms and stimulated new entry. The number of firms with NCE approvals increased 84% from 1970-1979 to 19901999. The growth in the number of firms with approvals was driven by firms with only one approval in the period. Of the 50 firms that had one approval during 1990 to 1999,

Joseph A. DiMnsi

41 of them had their first ever NCE approval in this period (It should be noted that one of these 41 firms had acquired a firm that had had two approvals in an earlier period prior to acquisition and that two other firms were joint ventures of firms that have had NCE approvals. We should also stress that the deconcentration and new entry noted is for development of traditional chemical [small molecule] compounds. The biotechnology industry grew exponentially during the 1980s and 1990s, with the creation of hundreds of start-up firms and a number of biotechnology [large molecule] drugs reaching the marketplace. It is also the case, however, that the biotechnology revolution has had some impact on traditional drug development, as some of the new technologies have been used to help identify targets for small molecule drugs.).

Interfirm Variation in the Productivity of Development A fuller understanding of differential firm performance should be aided by data on the success of firms’ development programs. I used data from a Tufts CSDD database on the development history of investigational drugs, supplemented by data found in Pharmaprojects Plus (CD-ROM version 2.1) and the NDA Pipeline, to examine clinical success rates for NCEs that had an IND first filed during the 1980s.The Tufts CSDD database provides information on all investigational NCEs taken into clinical testing by a large sample of firms. The clinical development process is variable and very lengthy for some drugs. Thus, a number of compounds that first entered United States clinical testing during the 1980s have neither received FDA approval nor had research abandoned. The PharmaProjects database and the NDA Pipeline were used to determine the current status (through December 1999) of the NCEs with 1980 to 1989 IND filings in the Tufts CSDD database that were listed as still active as of the last Tufts CSDD survey. The dataset used for analysis contains 762 NCEs, 21 of which

1 3 4 4 6 6 8 8 8 8 8 8

1

Rank

8.4 8.4 7.4 5.3 5.3 4.2 4.2 3.2 3.2 3.2 3.2 3.2 3.2

Share of NCEs(%)

13 5 13 5 20 3 30 2

-

11

11 4 5

Rank

Rank

9 3 1 5 9 16 7 21 1 23 3 34 6

2.9 5.8 4.4 2.9 0.0 2.2 3.6 2.2 4.4 1.5 6.6 0.7 7.3

3.2 5.4 5.9 4.9 3.2 2.2 3.8 1.6 5.9 1 .I 5.4 0.5 4.3

Share of NCEs(%)

1980-1 989

Share of NCEs(%)

1970-1979

0.0

3.3 1.1 0.4

6 23 45

2.6 1.8 4.4 1 .I 2.2 0.7 4.0 3.3 4.7

Share of NCEs(%)

-

11 17 2 23 14 32 5 6 1

Rank

1990-1999

"Wellcome was acquired by Glaxo in 1995. bAmericanCyanamid was acquired by American Home Products in 1994. 'AH. Robins was acquired by American Home Products in 1989. %earle was acquired by Monsanto in 1985. Monsanto had no prior NCE approvals. Subsequent NCE approvals have been attributed to Searle. eUpjohn and Pharmacia merged to form Pharmacia & Upjohn in 1995. During the 1990s. Upjohn obtained one approval, Pharmacia obtained two approvals, and Pharmacia & Upjohn obtained nine approvals. The total of twelve approvals represents 4.4% of all NCE approvals during the 1990s.

American Home Products Lilly Merck Schering-Plough Wellcome a American Cyanamidb Pfizer Abbott Johnson & Johnson A.H. Robins' Roche Searled Upjohn'

Company

1963-1 969

TABLE 10 Variation in Company Rank and Share of New Chemical Entities (NCEs) by Period of United States Marketing Approval (parent company at time of approval)

Y

2

3

E

n r,

I

s

sg.

0

3

3

00

9

$

s.

5P

Joseph A. DiMasi

1190

60

50

a

50 40

c

ti

30

n

5

20

z

10

0

1

3-4

2

5-9

1o+

Number of NCE Approvals ( 01970-1979 0 1980-1989 W 1990-19991

FIGURE 4. Distribution of United States new chemical entity (NCE) approvals across firms (parent company at the time of approval) by period of approval.

(2.8%) were still active through December 1999. Table 11 shows the current clinical success rates for the 15 firms in the Tufts CSDD investigational database that had at least 20 NCEs with INDs first filed during the 1980s (These firms accounted for 637 of the 762 NCEs in the Tufts CSDD database with INDs first filed during the 1980s. The number of NCEs for the firms varied from 24 to 64.The firms tended to be highly diversified across therapeutic categories. The average number of therapeutic categories covered by a firm's investigational NCEs was 8.13, with a diversification index value of 5.71. The firms are coded by letter since most of the data used were obtained on a proprietary basis.). In aggregate, the current success rate for these firms is 20.4%. The individual firms to date have differed substantially in their performance with regard to taking drug candidates through development to marketing approval. As noted, some of the compounds in the sample are active and potentially can be approved for marketing. I used the following probit specification to model the conditional probability of approval at a given number of years from IND filing:

where u is the time from IND filing to either abandonment of development or to marketing approval. The regression results can be used to predict the proportion of active NCEs that will eventually obtain approval (See DiMasi [3] for a discussion of a two-stage process for estimating clinical success rates when a significant number of compounds are still in active testing. The first stage involves estimating a survival curve, while the second stage involves estimating the conditional probability of approval. With so few compounds in the sample analyzed here still active, employing the first stage makes little difference to the results. The probit equation was estimated separately for self-originated NCEs and for acquired NCEs, and the results were highly significant. Going a step further and estimating separate equations for each therapeutic category in each of these two groups may give poorer results, as the data for the equations will be much sparser and differences in the conditional probability of approval at long times from the start of clinical testing [as opposed to differences in overall success rates, which are cumulations of differences in survival rates and conditional approval probabilities over time] are likely to be small across categories.). The outcomes are shown in the column of predicted final

Trends in New Drug Innovation by Pharmaceutical Firms

1191

TABLE 11 Technical and Commercial Productivity for Leading Pharmaceutical Firms

ComPanY A B

C D E F

G H I J K L

M N

0

Current Success Rate”

Predicted Final Success Rateb

(“w

Technical Productivity Index’

41.9 43.9 41.7 31.7 18.6 21.4 20.0 15.2 17.8 15.1 13.2 10.9 11.5 11.9 8.0

49.3 43.9 41.7 31.7 23.3 21.4 22.0 16.7 19.9 16.3 13.2 13.4 13.9 11.9 8.0

2.07 1.76 1.75 1.42 1.02 0.90 0.85 0.83 0.80 0.72 0.63 0.62 0.53 0.48 0.33

(“w

Company

C D 6 A H E

M F L

G J

I

N K

0

Mean Clinical Time (years)d

Commercial Productivity Index”

4.18 3.52 4.34 6.01 4.00 5.52 4.86 4.64 5.46 4.32 4.36 4.25 4.40 4.69 4.01

2.16 1.51 1.31 1 .I2 1.08 1.05 0.96 0.91 0.86 0.77 0.76 0.69 0.61 0.60 0.47

“Percent of NCEs with INDs first filed during 1980 to 1989 that have been approved by the FDA through December 1999. bPercentof NCEs with INDs first filed during 1980 to 1989 that are predicted to ultimately be approved by the FDA. Predictions for NCEs that were still active as of the end of 1999 are estimated from a probit regression. “Ratio of the predicted final success rate for a company to the success rate the company would, given the make-up of its portfolio of investigational NCEs by therapeutic class and compound source, be expected to have if it were as technically successful as the industry as a whole. “Mean time from first IND filing to either a decision to abandon development or to NDA submission. “Ratio of 1998 company pharmaceutical sales per NCE-year (product of the number of NCEs with an IND first filed during 1980 to 1989 and mean clinical time) to 1998 industry pharmaceutical sales per NCE-year.

success rates in Table 11. As with current success rates, the variability in predicted success rates is quite high. Differences in clinical success can reflect either heterogenous firm capabilities and performance in drug discovery or the efficiency of clinical development programs. DiMasi (3) found that clinical success rates can also vary by source (self-originated or acquired) and by therapeutic class. To control for differences in the make-up of development porfolios by source and therapeutic class, I constructed for each firm a weighted average success rate that uses estimates of success rates by source and therapeutic class for the industry (ie, all firms in the database) as weights applied to each firm’s portfolio of drug candidates. The resulting weighted av-

erage success rate indicates what the firm’s success rate could be expected to be if the firm performed in all categories and by source as did the industry as a whole. Dividing the firm’s predicted final success rate by this weighted average success rate yields the technical productivity index values in Table 11. By this measure, only five of the 15 firms were above-average performers, and variability was very high with performance ranging from 67% below average to 107% above average. It is also possible to construct a rough measure of commercial productivity for these companies. By the late 1990s, the sales of most pharmaceutical firms should be heavily dependent on drugs that began clinical testing during the 1980s. Data on pharmaceutical

1192

Joseph A. DiMasi

sales are available for 1998. The productivity of new drug development, though, is dependent on costs as well as returns. The number of NCEs with INDs filed and the time spent in clinical testing can be used to proxy for clinical development effort. The measure used here is NCE-years (that is, the product of the number of NCEs with INDs filed and the mean time that the NCEs spent in clinical testing). Productivity can then be measured as pharmaceutical sales per NCE-year. A commercial productivity index was defined as the ratio of pharmaceutical sales per NCEyear for a given firm to industry aggregate pharmaceutical sales per NCE-year. The industry values were estimated from data on all firms in the dataset. Thus, the commercial productivity index is a measure of how far above or below average a firm’s commercial performance has been (The measure is a first approximation to a true index of commercial productivity. Ideally, one would want to know the actual R&D costs for all of the firms’ investigational NCEs for the period analyzed, the actual returns of the firms’ NCEs that made it to the marketplace over their entire lifecycles, and the production, distribution, and promotion costs of these approved NCEs.). Relative to technical productivity, the results in Table 11 show somewhat less, but still high, variability in commercial productivity. Six of the 15 firms had above average commercial productivity, with the results ranging from 53% below average to 116% above average. With the exception of a few firms, the technical productivity and the commercial productivity index values are close to one another. Even with the exceptions included, the two indices are correlated at a statistically significant level (Pearson correlation coefficient = 0.760, p = 0.001).

DISCUSSION Innovation in the pharmaceutical industry can be affected by numerous factors, some of which, such as regulation, trends in R&D costs, scientific opportunities, and knowl-

edge spillovers can potentially impact all firms. A comprehensive explanation of how innovation in this industry has proceeded should, however, also include firm-specific factors, such as individual organizational structures and how effectively a firm reacts to changes in its environment. Research on the nature of pharmaceutical innovation is, therefore, best conducted with knowledge of how innovative individual firms have been over a lengthy period. This study has provided such information for the most significant type of pharmaceutical innovation-the development and marketing of NCEs. Since the 1962 amendments to the Federal, Food, Drug and Cosmetic Act of 1938, innovation, as measured by NCE approvals, initially declined but later increased. Much of the increase in new product introductions likely is linked to advances in biomedical research, although some portion of it may be attributable to knowledge spillovers. Our evidence on NCE approvals indicates that this type of innovation is fairly widely dispersed across firms, and has become more so over time. Substantial turnover in rank among the leading firms producing new drugs is also evident from the data. Nonetheless, some firms have been more consistent in maintaining relatively high rates of new product introductions than have others. Furthermore, the evidence on the development experience of investigational drugs indicates that firms have differed substantially in their approval success rates for compounds that advance to clinical testing. The data also suggest that firms have varied notably in the commercial success that they have achieved relative to their development efforts. If one assumes that the observed interfirm differences in innovation rates and drug development performance are not random, then the data strongly indicate a need to understand in some detail how different firms have structured their R&D programs historically and how they have adjusted to changes in economic conditions, regulatory policy, and scientific progress. In that regard, historical accounts of developments in the industry that

Trends in New Drug Innovation by Pharmaceutical Firms

place events within the context of a coherent theoretical framework are helpful (9,3 1,32). Perhaps even more useful are detailed histories of leading firms (27). Alternative, but not entirely inconsistent, theories of innovation have been advanced in recent decades. Some may be viewed as descendants of or related in some respects to Schumpeter’s vision (30). Hypotheses about the nature of industrial innovation include those developed in evolutionary economics (33,34). In this framework, firms differ in the set of behavioral “routines” that they use when learning and developing new products and processes. Since these routines have evolved as successful responses to problems that the firm has encountered, they confer competencies that are specific to the firm. The routines for an individual firm are numerous and complementary, and so they are difficult to imitate. This may explain why some firms are consistently more successful at innovation than are their competitors. This theoretical framework also has implications for understanding why firms differ in their ability to adjust to environmental changes and for understanding how the historical paths that they have followed can impact trends in pharmaceutical innovation at the firm and industry levels. Also noteworthy are views that technological progress and success in the marketplace are derived from organizational capabilities that first-movers develop from investing in a management structure, in marketing and distribution networks, and in research and production to the extent needed to take advantage of economies of scale and scope (31). Ultimately, facts and empirical analysis will be needed to sort out which theoretical constructs or syntheses of constructs provide the most useful insights into the processes by which innovation occurs in the pharmaceutical industry. Acknowledgments-I am grateful to Elaine Healy for her careful research assistance in compiling merger and acquisition histories for this project. All errors and omissions, of course, are my responsibility. This research was funded, in part, by a grant from Merck & Co., Inc.

1193

REFERENCES 1. DiMasi JA, Hansen RW, Grabowski HG, Lasagna

L. Cost of innovation in the pharmaceutical industry. J Health Econ. 1991;10(2):107-142. 2. DiMasi JA, Hansen RW, Grabowski HG, Lasagna L. Research and development costs for new drugs by therapeutic category: a study of the US pharmaceutical industry. P h a m c o E c o n . 1995;7(2):152169. 3. DiMasi JA. Success rates for new drugs entering clinical testing in the United States. Clin Pharmacol Ther. 1995;58(1):1-14. 4. Grabowski HG, Vernon J. Returns to R&D on new drug introductions in the 1980s. J Health Econ. 1994;13:384-406. 5. DiMasi JA, Seibring MA, Lasagna L. New drug development in the United States, 1963 to 1992. Clin Pharmacol Ther. 1994;55(6):609-622. 6. DiMasi JA. Grabowski HG, Vernon J. R&D costs, innovative output and firm size in the pharmaceutical industry. Int J Economics Bus. 1995;2(2):201-219. 7. Henderson R, Cockburn 1. Measuring competence? exploring firm effects in pharmaceutical research. Srrategic Manage J. 1994;15(Winter special issue): 63-84. 8. Gambardella A. Competitive advantages from inhouse scientific research: the US pharmaceutical industry in the 1980s. Research Policy. 1992;21:391407. 9. Lamoreaux N, Galambos L. Understanding innovation in the pharmaceutical industry. Presented at Understanding Innovation, The Johns Hopkins University, Baltimore, MD, June 6, 1997. 10. Goodwin JS. The empirical basis for the discovery of new therapies. Perspectives Biology Med. 1991; 35( 1):20-36. 11. Grabowski HG. Drug Regulation and Innovation: Empirical Evidence and Policy Options. Washington, DC: American Enterprise Institute; 1976. 12. Grabowski HG, Vernon J. Innovation and structural change in pharmaceuticals and biotechnology. I n dustrial and Corporate Change. 1994;3(2):435449. 13. Pharma Marketletter. The role of mega-mergers in the future of the pharma industry. Pharma Markerletter. 1997;Feb 10, 24(6):26-27. 14. Tompkins R. Analyzing the impact of mergers and acquisitions. Scrip Magazine. 1996;Oct. 50:6-8. 15. Dranove D, Meltzer D. Do important drugs reach the market sooner? RAND J Economics. 1994;25(3): 402-423. 16. Shulman SR, Bienz-Tadmor B, Seo PS, DiMasi JA, Lasagna L. Implementation of the orphan drug act: 1983-1991. Food Drug I a n , J . 1992;47(4):363403. 17. Shulman SR, Manocchia M. The U.S. orphan drug program: 1983-1995. PharmacoEcon. 1997;12(8): 312-326.

1194 18. Scherer FM, Ross D. Industrial Market Structure and Economic Performance. 3rd ed. Boston: Houghton Mifflin; 1990. 19. Grabowski HG, Vernon J. New studies on market definition, concentration, theory of supply, entry, and promotion. In Chien RI, ed. Issues in Pharmaceutical Economics. Lexington, MA: DC Heath; 1979. 20. Hirschman AO. The paternity of an index. Am Econ Review. 1964;54:761. 21. U.S. Bureau of the Census. 1992 Census of Manufacturers: Concentration Ratios in Manufacturing. [MC-92-S-2]. Washington, DC: US Bureau of the Census; 1996. 22. Grabowski HG, Vernon J. Structural effects of regulation on innovation in the ethical drug industry. In Mason RT, Qualls FT, eds. Essays on Industrial Organization in Honor ofJoe Bain. Cambridge, MA: Ballinger; 1976. 23. Stigler GJ. Introduction. In Business Concentration and Price Policy. Princeton, NJ: Princeton University Press; 1955. 24. Cocks DL. Product innovation and the dynamic elements of competition in the ethical pharmaceutical industry. In Helms RB. ed. Drug Development and Marketing. Washington, DC: American Enterprise Institute; 1975. 25. Schwartzman D. Innovation in the Pharmaceutical

Joseph A. DiMasi Industry. Baltimore, MD: Johns Hopkins University Press; 1976. 26. Ravenscraft DJ, Scherer FM. Mergers, Sell-offs, and Economic Eflciency. Washington, DC: Brookings Institution; 1985. 27. Galarnbos L, Sturchio JL. Sustaining innovation: critical transitions at Merck & Co.. Inc. Presented at Understanding Innovation, The Johns Hopkins University, Baltimore, MD. June 7, 1997. 28. Vagelos PR. Are prescription drug prices high? Science. 1991;252:1080-1084. 29. Henderson R, Cockburn I. Scale, scope, and spillovers: the determinants of research productivity in drug discovery. RAND J Econ. 1996;Spring. 27(1): 32-59. 30. Schumpeter JA. Capitalism, Socialism, and Democracy. 3rd ed. New York: Harper & Row; 1950. 31. Chandler AD. Scale and scope: The Dynamics of Industrial Capitalism. Cambridge, MA: The Belknap Press of Harvard University Press; 1990. 32. Drews J. In Quest of Tomorrow's Medicines. New York: Springer-Verlag; 1998. 33. Nelson RR. Winter SG. An Evolutionary Theory of Economic Change. Cambridge, MA: The Belknap Press of Harvard University Press; 1982. 34. Nelson RR. Recent theorizing about economic change. J Econ Literature. 1995;33( 1):48-90.