Drug fnformnfionJourMl, Vol. 34. pp. 1169-1194.

2000 0092-861 5/2000

Printed in the USA. All rights reserved. Copyright 0 2000 Drug Information Association Inc.

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 pharmaceuti-
cal 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 substan-
tial. 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 yardsticks. A number of factors that contrib-

THE SUCCESS OF in the phma-
ceutical 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.
Ute 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 impor-
tant considerations when assessing the vital-
ity 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 phar-
and analyzed in a series of reports by the maceutical industry in the modem era.
Tufts Center for,the Study of Drug Develop- General 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 mar-
oped 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 de-
reported 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 con-
new 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 success-
small 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 posi-
DATA 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 in-
proval 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 molecu-
include 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 sweet-
however, 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 spon-
tion that directed regulatory authorities to sored the compound’s new drug application
require proof of efficacy before granting (NDA), the NDA approval date, the therapeu-
marketing approval for new drugs (ie, the tic class for the compound’s original ap-
1962 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 Market-
place Guide, and various industry trade pub-
lications.
The primary unit of analysis is the spon-
soring company. Because nonproprietary
data were used, it was not necessary to aggre-
gate 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. Re-
sults are presented both by company affilia-
tions at the time of approval and by the cur-
rent 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 rat-
ing 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 three-
tiered 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 proj-
ects, the results were also stratified according
to whether the compounds were self-origi-
nated (developed entirely under the auspices
of the sponsoring firm) or acquired (licensed-
in or otherwise obtained). The source infor-
mation 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 mar-
keting approval from the FDA. Clinical suc-
cess rates were estimated by firm and used
to construct a technical productivity index,
accounting for differences in therapeutic
class and compound source in the investiga-
tional drug portfolios of the analyzed firms.
Where needed, the survey data were supple-
mented by public information on the current
status of drugs from PharmaProjects and the
NDA Pipeline. Given that data for these esti-
mations were collected primarily on a confi-
dential basis, individual firms will be identi-
fied 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 pharma-
ceutical industry must compete. Premarket-
ing regulatory requirements increased as
firms were required to demonstrate the effi-
cacy of their new products, and a positive
review by the FDA of a marketing applica-
I172
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 in-
dustry is varied, ranging from line extensions
of existing products to new therapeutic com-
pounds; 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 signif-
icant is the development of new human-use
therapeutic compounds. It is this type of out-
put 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 approv-
als 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 ap-
provals were 13.6 and 13.7 per year for 1963
50
40
I . ..
0 " " " " " " " ' " " " " " '
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.
Joseph A. DiMasi
to 1969 and the 1970s, respectively. How-
ever, 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 enzy-
mology in the mid-1970s and molecular biol-
ogy 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 acqui-
sition 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-
' - L L u
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 Company Number of NCEs Percent of all NCEsO

1 Merck 36 5.2
2 Johnson & Johnson 33 4.8
3 Lilly 31 4.5
3 Roche 31 4.5
5 Pfizer 27 3.9
6 American Home Products 25 3.6
7 Upjohn 22 3.2
8 Bristol-Myers 21 3.0
8 Schering-Plough 21 3.0
10 Warner-Lambett 19 2.7
11 Abbott 18 2.6
12 Wellcome 17 2.5
13 Sandoz 14 2.0
14 American Cyanamid 13 1.9
14 Bayer 13 1.9
14 Bristol-Myers Squibb 13 1.9
14 Hoechst 13 1.9
18 SmithKline Beckman 13 1.9
19 SmithKline Beecham 12 1.7
20 Ciba-Geigy 11 1.6
21 Akzo Nobel 10 1.4
21 Glaxo 10 1.4
21 Syntex 10 1.4
24 Nestle 9 1.3
24 Pharmacia & Upjohn 9 1.3
26 Glaxo Wellcome 8 1.2
26 Sterling 8 1.2
28 Boehringer lngelheim 7 I .o
28 Robins 7 1 .o
30 Astra 6 0.9
30 Dow 6 0.9
30 Squibb 6 0.9
30 Zeneca 6 0.9
34 Allergan 5 0.7
34 BASF 5 0.7
34 BOC Group 5 0.7
34 Dupont 5 0.7
34 ICI 5 0.7
34 Rhone-Poulenc Rorer 5 0.7
"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 signifi-
cance) counts of the number of successful
innovations. The leading four firms (Merck,
Johnson & Johnson, Lilly, and Roche) aver-
aged nearly one NCE approval per year, with
only five approvals in the aggregate separat-
ing 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 acquisi-
tions 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 War-
ner-Lambert.). The rest of the firms in the
1174 Joseph A. DiMasi

top 10 merged with, acquired, or were ac-
quired 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 approv-
als (for firms with at least five approvals),
attributing to parent companies at the end of
1999 all of the past approvals of the compa-
nies that the firms had acquired. The acquisi-
tion of American Cyanamid and A.H. Robins
by American Home Products, the merger of
Bristol-Myers and Squibb, and the acquisi-
tion 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 Company Number of NCEs Percent of all NCEs"

1 American Home Products 45 6.5

2 Bristol-Myers Squibb 43 6.2
3 Roche 41 5.9
4 Aventis 38 5.5
4 Pharmacia & Upjohn 38 5.5
6 Merck 36 5.2
6 Novartis 36 5.2
8 Glaxo Wellcome 35 5.1
9 SmithKline Beecham 34 4.9
10 Johnson & Johnson 32 4.6
11 Lilly 31 4.5
12 Pfizer 27 3.9
13 Schering-Plough 21 3.0
13 Warner-Lambert 21 3.0
15 Abbott 18 2.6
15 Astrazeneca 18 2.6
17 Bayer 14 2.0
18 Akzo Nobel 10 1.4
19 Nestle 9 1.3
20 Allergan 8 1.2
20 BASF 8 1.2
20 Boehringer lngelheim 8 1.2
20 Dupont Merck 8 1.2
20 Monsanto 8 1.2
25 BOC Group 5 0.7
"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.
Trends in New Drug Innovation by Pharmaceutical Firms 1175

shares based on worldwide pharmaceutical along various dimensions is unclear at this

sales (There is reason to suspect that the mar-
ket 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 world-
wide 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 world-
wide sales obtained by leading firms has re-
mained 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
time. A survey of the parties most affected
by the merger and acquisition activity-cus-
tomers, employees, and investors-revealed
that the groups differed in their assessments
of how successful individual mergers and ac-
quisitions 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 pharma-
ceutical market share are among the top 10
and top 20 developers of NCEs, respectively.
Similar results are obtained if the NCE rank-
ing is made for more recent approvals (1982
to 1999).
Interfirm Heterogeneity in the Nature of
Pharmaceutical Innovation
The aggregate data on NCE output are infor-
mative, but they can mask important differ-

TABLE 3
Worldwide Pharmaceutical Sales by Company"
Pharmaceutical
Rank Company Sales ($ million)

1 Merck 15,296.5
2 Aventisb 13,608.1
3 Glaxo Wellcome 13,230.5
4 Astrazenecab 12,754.0
5 Bristol-Myers Squibb 12,573.0
6 Pfizer 12,230.0
7 Novartis 1 1 , I 74.8
8 Roche 9,921.5
9 American Home Products 8,901.8
10 Lilly 8,590.4
11 Johnson & Johnson 8,562.0
12 SmithKline Beecham 7,701.6
13 Schering-Plough 6,695.0
14 Pharmacia & Upjohn 6,127.0
15 Warner Lambert 5,604.0
16 Abbott 5,602.0
17 Sanofi-Synthelabob 4,832.7
18 Bayer 4,823.7
19 Takeda 4,567.8
20 Boehringer lngelheim 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 Joseph A. DiMasi

ences across firms in the extent to which they experienced substantial growth in relative
have diversified their innovative efforts, in output; 23.7% of the NCE approvals from
the degree to which they have been self-suffi- 1982 to 1999 were in the cardiovascular area.
cient in innovation, and in the quality of their From chemical and pharmacological
innovation. Although the causes and implica- perspectives, the types of compounds devel-
tions of such differences are difficult to fullyoped in a therapeutic class do, however,
establish we can, at least as a first step, exam-
change over time. For example, in cardiovas-
ine the available data to determine if firms cular drug development, approvals of drugs
do indeed differ in these dimensions. within major pharmacologic classes pro-
ceeded (with considerable overlap) from diu-
Therapeutic 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. Antiinfec-
peutic classes generally are not substitutable tive development has progressed with the
in consumption. Even drugs within the same marketing of successive generations of anti-
general 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 com-
market shares can appear quite low when pound types within a therapeutic class can
considering pharmaceutical products as a differ substantially, many of them share com-
whole, 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 antineoplas-
tise 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 Bris-
dependent 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 approxi-
ries. 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 pro-
ence from the late 1970s onward conceivably grams through expansion and by strength-
could 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, Ta-
classes 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
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. Addition-
ally, 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 out-
put 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 sponsor-
ing 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 com-
pounds is quite variable. One way to gauge
the degree to which firms have relied on and
have been productive at discovering and de-
veloping 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-
1177
23.4%
originated. The proportion of NCE approvals
that were self-originated has declined over
time, decreasing from 71.6% for NCEs ap-
proved 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-origi-
nated NCE approvals are more reliant on in-
ternal 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 mar-
keted 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 rela-
tively large number of the self-originated
approvals in the class. For example, in the
cardiovascular class Merck had 11 self-origi-
nated 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 de-
veloped by the National Cancer Institute.
 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)

AnalgesiclAnesthetic Antiinfective Antineoplastic Cardiovascular Central Nervous System

Company" NCEs Comanyb NCEs Company' NCEs Companyd NCEs Company" NCEs

Johnson & Johnson 9 Lilly 13 Bristol-Myers 9 Merck 14 American Home 7

Products
Merck 5 Pfizer 11 Roche 4 AmericanHome 7 Roche 7
Products
American Home 4 Roche 9 Pharmacia & Upjohn 3 Pfizer 6 Abbott 5
Products
Akzo Nobel 4 Bristol-Myers 8 Upjohn 3 Lilly 5 Lilly 5
BOC Group 4 Johnson & Johnson 8 Wellcome 3 Abbott 4 Pfizer 5
Abbott 3 Bayer 6 Boehringer 4 Warner-Lambert 5
lngelheim
Astra 3 Merck Sandoz 4 Johnson & Johnson 4
Dupont 3 Schering-Plough Upjohn 4 Sandoz 4
Glaxo 3 Warner-Lambert Ciba-Geigy 3
Nestle 3 Wellcome Merck 3
Schering-Plough 3 American Cyanamid A.H. Robins 3
Syntex 3 Bristol-Myers Squibb Wellcome 3
Glaxo Wellcome
Upjohn
"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. ?
 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)
~~~

AnalgesidAnesthetic Antiinfective Antineoplastic Cardiovascular Central Nervous

System

Company" NCEs Comanyb NCEs Company' NCEs Companyd NCEs Company" NCEs

Johnson & Johnson 8 Bristol-Myers Squibb 15 Bristol-Myers Squibb 13 Merck 14 American Home 12
Products
Glaxo Wellcome 7 Glaxo Wellcome 14 Pharmacia & Upjohn 9 American Home 12 Aventis 8
Products
American Home 6 Liily 13 AstraZeneca 4 Aventis 8 Novartis 8
Products
AstraZeneca 5 Roche 12 Roche 4 Novartis 8 Roche 7
Merck 5 SmithKline Beecham 12 Aventis 3 Pharmacia & Upjohn 8 Abbott 5
Roche 5 Pfizer 11 Glaxo Wellcome 3 Pfizer 6 Liily 5
Akzo Nobel 4 American Home 10 Novartis 3 SmithKline Beecham 6 Pfizer 5
Products
BOC Group 4 Johnson & Johnson BASF 5 Warner-Lambert 5
Aventis Bristol-Myers Squibb 5 Glaxo Wellcome 4
Pharmacia & Upjohn Lilly 5 Johnson & Johnson 4
Warner-Lambert Roche 5 SmithKline Beecham 4
Merck Abbott 4
Schering-Plough Boehringer 4
lngelheim
"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.
1180
While Bristol-Myers Squibb had substan-
tially more approvals in this class than did
any other firm, all of its antineoplastic ap-
provals were acquired.
Quality of Innovation. The data on NCE ap-
provals presented thus far have not been ad-
justed 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 discov-
ered or established until the drugs are used
in everyday medical practice or until postap-
proval 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 applica-
tions. They also examine the number of ma-
jor markets in which a drug is introduced
and the drug’s United States sales, but these
measures correlate more with a drug’s com-
mercial significance than with its medical
significance.).
Our data suggest a positive correlation be-
tween a firm’s total number of NCE approv-
als and the number of the firm’s NCE approv-
als 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 approv-
als (Table 1). (Therapeutic ratings were not
available for three NCEs approved in the
1960s. Lilly, Bristol-Myers, and Schering-
Plough 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 propor-
tion 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%). How-
ever, the averages do not reveal the substan-
tial 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 pro-
grams that produce proportionately more ap-
provals 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 com-
mercially significant, this is not always the
case. For example, FDA-assigned therapeu-
tic 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 manufac-
turer can demonstrate that there is no reason-
able expectation that development and mar-
keting costs can be recouped from sales in
the United States.). The sales of the vast ma-
jority 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 thera-
peutic 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 license-
in 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
Self-Originated" Priority-Ratedb Priority-Rated

Company NCEs Company NCEs Company NCEs

Merck 28 Merck 25 Merck 20

Johnson & Johnson 25 Roche 22 Roche 14
Liily 25 Johnson & Johnson 17 Johnson & Johnson 13
Pfizer 21 Wellcome 15 Lilly 11
Roche 20 Lilly 12 Wellcome 10
Hoechst 13 Upjohn 12 Upjohn 8
Upjohn 13 Bristol-Myers 10 Pfizer 7
Sandoz 12 Pfizer 10 Warner-Lambert 7
Schering-Plough 12 Schering-Plough 9 Boehringer- 5
Ingelheim
Wellcome 12 Warner-Lambert 9 Hoechst 5
Warner-Lambert 11 SmithKline Beckman 8 Sandoz 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 mea-
sured with our data (Table 6). Although the
order is slightly different, the six leading
firms in terms of the number of self-origi-
nated 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 In-
gelheim, 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 re-
plete with studies that utilize measures of the
degree to which markets are dominated by
relatively few firms. These measures are of-
ten used to describe and analyze industry
structure, with putative implications for eco-
nomic efficiency in these markets. Industry
concentration measures are usually defined
in terms of sales, but employment and asset-
based 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] re-
view a number of studies that have used con-
centration measures in analyses of the com-
petitiveness of the pharmaceutical industry.).
The most commonly used measures are the
concentration ratio (especially, the four-firm
and eight-firm ratios) and the Hefindahl-
Hirschman index (HHI) (A concentration ra-
tio is the aggregate share of the market held
1182 Joseph A. DiMasi

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-originatedl
Self-Originateda Priority-Ratedb Priority-Rated

Company NCEs Company NCEs Company NCEs

~ ~~ ~ ~ ~

Merck 28 Roche 26 Merck 20

Roche 28 Glaxo Wellcome 25 Roche 17
Novartis 26 Merck 25 Glaxo Wellcome 15
Aventis 25 Pharmacia & Upjohn 21 Aventis 14
Lilly 25 Aventis 19 Pharmacia & Upjohn 14
Pharmacia & Upjohn 25 Bristol-Myers Squibb 18 Johnson & Johnsonc 12
SmithKline Beecham 25 Novartis 17 Lilly 11
Johnson & Johnsonc 24 Johnson & JohnsonC 16 Novartis 11
Glaxo Wellcome 23 SmithKline Beecham 16 SmithKline Beecham 11
ffizer 21 American Home 13 ffizer 7
Products
Bristol-Myers Squibb 20 Lilly 12 Warner-Lambert 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 concentra-
tion 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 ac-
counts 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 con-
centration. 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 approv-
als can be used to provide measures of indus-
try concentration in innovation. The concen-
tration measures that have been applied to
sales are applicable, without change in for-
mat, 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 com-
pute 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 ap-
proved every year. Grouping of years is,
therefore, appropriate and necessary. Con-
centration measures of any type for a given
year or period are only rough static indicators
of industry structure. In theory, some indus-
tries with relatively high measured concen-
tration may be quite competitive and some
with relatively low measured concentration
may exhibit oligopolistic behavior (18). In-
stitutional 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 de-
gree of concentration in an industry at differ-
ent points in time will often serve as an accu-
rate 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 1183

vary from case to case. A concentration or
deconcentration trend might, for example, re-
flect changes in regulatory policy, shifts in
demand, or the emergence of new technolo-
gies 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 simi-
lar to reported sales concentration ratios for
the pharmaceutical industry. The United
States Census Bureau has published four-
firm sales concentration ratios for pharma-
ceutical 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 con-
centration 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 pharmaceuti-
cal industry concentration ratios are moder-
ate in comparison to the sales concentration
ratios for other industries. (The Census Bu-
reau has published four-firm sales concentra-
tion ratios for 456 four-digit S.I.C. industries
for 1992 [21]. The ratio for the pharmaceuti-
cal 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 manufactur-
ing industries with four-digit S.I.C. codes had
four-firm concentration ratios of 40% or
higher for 1982. Some scholars interpret ra-
tios this high, when the market is appropri-
ately 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
50
8
88 “ 30
z
b
8..
20
E
g
Q)
10
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 prod-
uct lifecycles in this industry, (For their
estimations of rates of return for NCE devel-
opment, 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 pharmaceu-
tical products or on shares of all NCE approv-
als are seriously biased downward, since eco-
nomically 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 re-
training 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 ther-
apeutic classes, and the classes generally ex-
hibit a downward trend in concentration, as
was noted for all NCE approvals. For exam-
ple, the four-firm concentration ratio for car-
diovascular NCEs fell from 61.6% for the
1960s to 23.3% for the 1990s (The cardiovas-
cular NCE four-firm concentration ratios
NCEs were 44.5 and 27.7 for the 1970s and
the 1980s, respectively. Similarly, the cardio-
vascular NCE HHI values were 1243, 864,
412, and 328 for the 1960s, 1970s, 1980s,
and 1990s, respectively.). However, the pe-
riod 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 antineoplas-
tic 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 innova-
tion 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 dy-
namic 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 diver-
sified across therapeutic categories in terms
of patents, sales, and worldwide NCE intro-
ductions [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 thera-
peutic 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 approv-
als 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 catego-
ries with approvals for these firms is 5.6 (me-
dian = 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 indi-
cate how concentrated its output is in those
categories. For example, 4 1% of Merck’s ap-
provals are in the cardiovascular class and
50% of Bayer’s approvals are in antiinfec-
tives. The degree to which company approv-
als 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 mea-
sure the diversification of companies across
different lines of business. The index is de-
fined as llXIF,;, where Fvis the share of com-
pany 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 approv-
als and line of business is replaced by thera-
peutic category. If the company has approv-
als 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 di-
versified 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 di-
versified firms were Johnson & Johnson,
Hoechst, and Schering-Plough. The least di-
versified 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 rela-
tively low diversification index value is due
to a high concentration of approvals in one
category [cardiovascular]. The disparity be-
tween 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 Stur-
chi0 [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 cate-
gories per firm increased from 4.7 for 1963 to
1980 to 6.3 for 1982 to 1999. The increased
diversification in approvals is likely associ-
ated with greater diversification in discovery
programs (although we lack the data to dem-
onstrate 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 ex-
ample, found support for the hypothesis that
the productivity of drug discovery research
(with output measured by counts of impor-
tant patents) can be substantially enhanced
by the existence of such externalities (both
within the firm and across firms).
Firm Turnover. Product innovation is the pri-
mary mechanism by which dynamic compe-
tition in the pharmaceutical industry is exhib-
ited. New drugs typically supplant old ones
in medical use. Substantial returns can be
earned on new products, and, by Schumpet-
erian logic (30), these potential returns
 4

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

Central
Analgesic/ Anti- Anti- Cardio- Nervous Gastro- Respir-
Company Anesthetic infective neoplastic vascular System Endocrine intestinal atory Other Total

Merck 3 4 0 9 0 2 2 1 1 22
Johnson 8 Johnson 5 4 2 2 3 2 1 1 1 21
Roche 1 6 1 2 3 0 1 0 3 17
Pfizer 1 6 0 4 2 1 0 2 0 16
American Home
Productsa 4 1 0 5 2 0 0 0 1 13
Bristol-Myers
Squibbb 1 5 2 2 1 0 0 0 2 13
Lilly 1 2 1 3 4 1 1 0 0 13
Abbott 3 3 0 2 2 1 0 1 0 12
SmithKline
Beecham' 1 4 2 2 2 1 0 0 0 12
Wellcorned 2 4 1 0 3 0 0 1 0 11
Bayer 0 5 0 3 0 2 0 0 0 10
Glaxod 3 3 0 1 1 0 1 1 0 10
Hoechst 0 2 2 2 1 1 0 1 1 10
Pharmacia 8
Upjohn' 0 1 3 2 1 1 0 0 1 9
Schering-Plough 2 2 2 1 1 0 0 1 0 9
Warner-Lambert 0 2 1 2 3 1 0 0 0 9
~ ~~ ~

'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. L
'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 2
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.
zg.0
 4

5.
TABLE 9
Company Diversification of United States New Chemical Entity (NCE) Approvals by $
Therapeutic Category (parent company at the time of approval)"
P
E
a
1963-1 980 1982-1 999
3
a
Number of Number of Diversification Number of Number of Diversification 2
Company Approvals Categories Index Approvals Categories Index 2.
0
a
Merck 13 6 4.83 22 7 4.17 5
Johnson & Johnson 11 5 3.90 21 9 6.78 'tr
5.
Roche 12 7 5.14 17 7 4.74
Pfizer 10 4 2.78 16 6 4.13
American Home Products 12 5 3.60 13 5 3.60 n
Lilly 17 6 2.60 13 7 5.12 E,
-.
0
Abbott 6 3 2.57 11 6 5.14 E
Bristol-Myers Squibbb 21 6 3.27 20 6 4.35 3
SmithKline Beechamb 13 7 5.45 21 8' 4.64
Wellcome 6 3 3.00 11 5 3.90 s
Bayer 2 2 2.00 10 3 2.63
Hoechst 2 2 2.00 10 7 6.25
Pharmacia & Upjohnb 15 5 4.67 21 7 5.09
Schering-Plough 10 5 3.57 9 6 5.40
Warner-Lambert 9 4 3.24 9 5 4.26
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.
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 evi-
dence 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 rank-
ing in all periods. While the average rank of
the 13 firms was 5.6 for the 1960s, the aver-
age ranks for the 1970s, 1980s, and 1990s
were 10.2, 10.6, and 15.4, respectively. Simi-
larly, 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, re-
spectively (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 indus-
try 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 technologi-
cal opportunities have not only been fol-
lowed by increasing numbers of new drug
approvals, but the evidence suggests that ad-
vances in biomedical science have fostered
deconcentration in new drug development
among existing firms and stimulated new en-
try. The number of firms with NCE approvals
increased 84% from 1970-1979 to 1990-
1999. 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 de-
concentration and new entry noted is for de-
velopment 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 market-
place. It is also the case, however, that the
biotechnology revolution has had some im-
pact 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 Phar-
maprojects Plus (CD-ROM version 2.1) and
the NDA Pipeline, to examine clinical suc-
cess rates for NCEs that had an IND first
filed during the 1980s.The Tufts CSDD data-
base provides information on all investiga-
tional NCEs taken into clinical testing by a
large sample of firms. The clinical develop-
ment 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 Pipe-
line 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
 Y
5P
s.
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) 9
00
1963-1 969 1970-1979 1980-1 989 1990-1999 3
3
0
Share of Share of Share of Share of sg.
Company Rank NCEs(%) Rank NCEs(%) Rank NCEs(%) Rank NCEs(%)

American Home Products 1 8.4 11 2.9 9 3.2 11 2.6 s

Lilly 1 8.4 4 5.8 3 5.4 17 1.8
Merck 3 7.4 5 4.4 1 5.9 2 4.4
Schering-Plough 4 5.3 11 2.9 5 4.9 23 1 .I
r,
n
Wellcome a 4 5.3 - 0.0 9 3.2 14 2.2
I
American Cyanamidb 6 4.2 13 2.2 16 2.2 32 0.7
Pfizer 6 4.2 5 3.6 7 3.8 5 4.0 E
Abbott 8 3.2 13 2.2 21 1.6 6 3.3 3
Johnson & Johnson 8 3.2 5 4.4 1 5.9 1 4.7 2
A.H. Robins' 8 3.2 20 1.5 23 1 .I - 0.0
Roche 8 3.2 3 6.6 3 5.4 6 3.3
Searled 8 3.2 30 0.7 34 0.5 23 1.1
Upjohn' 8 3.2 2 7.3 6 4.3 45 0.4
"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.
1190
60
50
50
a
40
c
30
ti
n
5 20
z
10
0
1 2
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 suc-
cess 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 diver-
sification 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 perfor-
mance 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 ap-
proved for marketing. I used the following
probit specification to model the conditional
probability of approval at a given number of
years from IND filing:
Joseph A. DiMasi
3-4 5-9 1o+
where u is the time from IND filing to either
abandonment of development or to market-
ing approval. The regression results can be
used to predict the proportion of active NCEs
that will eventually obtain approval (See Di-
Masi [3] for a discussion of a two-stage pro-
cess 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 com-
pounds in the sample analyzed here still ac-
tive, 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 clini-
cal testing [as opposed to differences in over-
all 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
Predicted
Current Final Technical Mean
Success Success Produc- Clinical Commercial
Com- Rate” Rateb tivity Com- Time Productivity
PanY (“w (“w Index’ pany (years)d Index”
A 41.9 49.3 2.07 C 4.18 2.16
B 43.9 43.9 1.76 D 3.52 1.51
C 41.7 41.7 1.75 6 4.34 1.31
D 31.7 31.7 1.42 A 6.01 1 .I2
E 18.6 23.3 1.02 H 4.00 1.08
F 21.4 21.4 0.90 E 5.52 1.05
G 20.0 22.0 0.85 M 4.86 0.96
H 15.2 16.7 0.83 F 4.64 0.91
I 17.8 19.9 0.80 L 5.46 0.86
J 15.1 16.3 0.72 G 4.32 0.77
K 13.2 13.2 0.63 J 4.36 0.76
L 10.9 13.4 0.62 I 4.25 0.69
M 11.5 13.9 0.53 N 4.40 0.61
N 11.9 11.9 0.48 K 4.69 0.60
0 8.0 8.0 0.33 0 4.01 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 suc-
cess rates is quite high.
Differences in clinical success can reflect
either heterogenous firm capabilities and per-
formance 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 dif-
ferences in the make-up of development por-
folios by source and therapeutic class, I con-
structed 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. Divid-
ing 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 vari-
ability was very high with performance rang-
ing 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 test-
ing during the 1980s. Data on pharmaceutical
1192
sales are available for 1998. The productivity
of new drug development, though, is depen-
dent 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 NCE-
year for a given firm to industry aggregate
pharmaceutical sales per NCE-year. The in-
dustry 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 ap-
proved NCEs.).
Relative to technical productivity, the re-
sults in Table 11 show somewhat less, but
still high, variability in commercial produc-
tivity. 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 com-
mercial productivity index values are close
to one another. Even with the exceptions in-
cluded, the two indices are correlated at a
statistically significant level (Pearson corre-
lation 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-
Joseph A. DiMasi
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 pro-
vided such information for the most signifi-
cant type of pharmaceutical innovation-the
development and marketing of NCEs.
Since the 1962 amendments to the Fed-
eral, 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 dis-
persed 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. Nonethe-
less, some firms have been more consistent
in maintaining relatively high rates of new
product introductions than have others. Fur-
thermore, 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 sug-
gest 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 de-
velopment performance are not random, then
the data strongly indicate a need to under-
stand 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 histo-
ries 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 nu-
merous 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 net-
works, and in research and production to the
extent needed to take advantage of econo-
mies 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 use-
ful insights into the processes by which inno-
vation 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 omis-
sions, 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 pharma-
ceutical industry. P h a m c o E c o n . 1995;7(2):152-
169.
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 in-
house scientific research: the US pharmaceutical in-
dustry in the 1980s. Research Policy. 1992;21:391-
407.
9. Lamoreaux N, Galambos L. Understanding innova-
tion in the pharmaceutical industry. Presented at Un-
derstanding Innovation, The Johns Hopkins Univer-
sity, 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. Washing-
ton, 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):435-
449.
13. Pharma Marketletter. The role of mega-mergers in
the future of the pharma industry. Pharma Markerlet-
ter. 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):363-
403.
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 Pharmaceuti-
cal 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 Manufac-
turers: Concentration Ratios in Manufacturing.
[MC-92-S-2]. Washington, DC: US Bureau of the
Census; 1996.
22. Grabowski HG, Vernon J. Structural effects of regu-
lation 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 Univer-
sity Press; 1955.
24. Cocks DL. Product innovation and the dynamic ele-
ments 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: Brook-
ings 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? Sci-
ence. 1991;252:1080-1084.
29. Henderson R, Cockburn I. Scale, scope, and spill-
overs: the determinants of research productivity in
drug discovery. RAND J Econ. 1996;Spring. 27(1):
32-59.
30. Schumpeter JA. Capitalism, Socialism, and Democ-
racy. 3rd ed. New York: Harper & Row; 1950.
31. Chandler AD. Scale and scope: The Dynamics of
Industrial Capitalism. Cambridge, MA: The Bel-
knap 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.