from competence destruction to competence access: evidence ...srockart/srf/papers... ·...
TRANSCRIPT
From Competence Destruction to Competence Access: Evidence from the Market for Anti-Cancer Drugs
M. Lourdes Sosa London Business School
Strategic and International Management Sussex Place, Regent’s Park
London NW1 4SA
Revised March 15, 2008
I thank Tom Allen, Roberto Fernandez, Tony Sinskey, Jesper Sørensen and Jim Utterback for their support and insight. I am also thankful for comments from Lisa Cohen, Michael Jacobides, Brandon Lee, Margaret Kyle, Anita McGahan, Isabel Fernández-Mateo, Woody Powell and Phanish Puranam in previous drafts of this paper. Detailed feedback from participants at the AIM Early Career Event, Cass Business School Workshop, EGOS and AoM are also gratefully acknowledged. I offer gratitude and appreciation to Kathleen Cui, Daniel Malconian, Manisha Manmohan, Lili Peng, Divya Titus and Sharon Zhang for meticulous research assistance in different phases of this project.
2
From Competence Destruction to Competence Access: Evidence from the Market for Anti-Cancer Drugs
Traditional theory on creative destruction distinguishes disruptions as competence-destroying vs. competence-enhancing to the resources and capabilities that incumbents had mastered. In the case of competence destruction, incumbents will lose their competitive advantage in the execution of in-house R&D. In the presence of organizational inertia, incumbents could even incur a disadvantage. In this study I propose analysis of technological disruptions requires attention not only to competence destruction but also to competence access. That is, the analysis of technological disruptions requires not only attention to the loss of value in incumbents’ prior capabilities. It also requires attention to the mechanisms through which incumbents might get access to the valuable, rare, and inimitable/unsubstitutable capabilities required to compete in the new technological regime of the market. I present a case study based on a mix of qualitative and quantitative data from the transition of the anti-cancer drug market into the biotechnology revolution. I compare the cases of two technological variants within the same radical, competence-destroying change. The two variants are a radical departure from the old technology, but differ in that a capability with a distinctive origin and technological trajectory is required only for the second variant. In support of my proposition, incumbents underperform in in-house R&D only in the variant in which they did not have access (whether through internal re-use or through acquisition) to the capability that represented the highest strategic value in the focal market. Indeed, it is a sub-set of diversifying entrants closely linked to the origin of this capability, that outperformed incumbents and all other entrants in this variant. I discuss implications for research on creative destruction and for emerging research on the origin and evolution of capabilities.
Key words: sustainable competitive advantage; organizational capabilities; incumbent;
technological disruption; R&D; evolutionary perspective.
3
__________________________________________________________________________________
1. Theory
Research on creative destruction has traditionally distinguished technological disruptions that are
competence-enhancing or competence-destroying to the resources and capabilities that incumbents had
mastered (Tushman and Anderson, 1986). When a technological change is competence-enhancing, the
disruption does not affect the value of the resources and capabilities already present within the research
and development (R&D) process of incumbent firms. Therefore, those resources and capabilities are
available for re-use by incumbents during the next technological regime of the market. Under the
assumption that some of these resources and capabilities exhibit an isolating mechanism (Rumelt, 1987),
such as unavailable trade in the open market (Teece, 1980; Kogut and Zander, 1992) or the presence of
time compression diseconomies in the accumulation of research know-how (Diericks and Cool, 1989),
incumbents will retain a competitive advantage above entrant firms.
In contrast, when a technological change is competence-destroying, the disruption obsoletes the value
of the resources and capabilities involved in the R&D process within the value chain of incumbents.
Incumbents will therefore lose their source of competitive advantage in R&D and be forced to start from
scratch. In the presence of organizational inertia, incumbents will further incur a disadvantage through
two main mechanisms: underinvestment and incompetence (Henderson, 1993). That is, incumbents will
tend to invest less on the new technology, and controlling for their investment, will also be less
productive in the R&D of products under the new technology, as compared to the set of firms entering the
market. Given their R&D incompetence, incumbents might retain leadership in terms of revenue only if
they retain a source of competitive advantage in the commercialization stage, that is, only if they hold
undestroyed, proprietary complementary assets (Teece, 1986; Tripsas, 1997).
As can be seen, predictions based on current theory assume the main determinant of R&D performance
for incumbents competing through a transition into a radically new technology is the loss of value in these
firms’ prior capabilities. But where do the newly required capabilities come from? If incumbents lose
their prior sources of competitive advantage but the new capabilities are easy acquired, entrants should
4
not accrue an advantage either. Under the assumption that organizational inertia would leave incumbents
behind, little attention has been paid to the mechanisms through which incumbents fail to access the
newly required capabilities during a technological transition.
In this paper, I propose that to understand the differences in performance across incumbents and
different groups of entrants during a technological discontinuity, studies of creative destruction need to
look beyond competence destruction and into competence access. That is, studies of creative destruction
should expand to investigating the differential access to the newly required capabilities that a disruption
imposes. If there is a source of competitive advantage for any group of entrants over other entering firms
and over incumbents, that source must have an isolating mechanism (Rumelt, 1987).
With this purpose in mind, I study two technological variants generated by the same competence-
destroying change in technologies, where the variants differ in the access incumbents get to one of the
newly required capabilities. Specifically, I study the anti-cancer drug market and its transition from
standard chemotherapy (i.e., cytotoxic drugs) into targeted drug development, a radical transition
generated by the biotechnology revolution. Among targeted anti-cancer drugs, I differentiate between
two variants: small-molecule and large-molecule drugs. These two variants are both radically different
from traditional cytotoxic anti-cancer drug discovery and development but differ only in one of the new
technological capabilities required. Large-molecule targeted anti-cancer drugs make use of
rDNA/antibody techniques, whereas their small-molecule counterparts do not. It is this contrast that I use
as the basis for my empirical test.
Through a combination of qualitative and quantitative methodologies, in this paper I show that
incumbents were able to adapt and sustain their competitive advantage in one variant of biotechnology:
small-molecule targeted anti-cancer drugs. In contrast, in the development of the large-molecule
counterparts, incumbents fell behind one specific group of entering firms: rDNA/antibody technique
pioneers. Although all adaptation mechanisms reported in the literature as supporting incumbents’
response to biotechnology have been applicable to both variants of targeted anti-cancer drugs (e.g., hiring
of expert personnel [Zucker and Darby, 1997]; re-use of clinical trial expertise [Rothaermel, 2001], re-use
5
of indication-specific knowledge [Sosa, 2006]), one capability, rDNA/antibody techniques, required for
large-molecule drug development was inaccessible to incumbents. As I will explain in this paper, the lack
of access to this capability on the part of incumbents stemmed from the particular origin and evolution of
the capability itself. Therefore, with this study I look to contribute not only to research in creative
destruction, but also to research on the origin and evolution of capabilities, and the role of these
characteristics in heterogeneous firm performance (Helfat and Lieberman, 2002; Ahuja and Katila, 2004;
Ethiraj, Kale, Krishnan and Singh, 2005).
2. Method
2.1 The Setting and the Technological Discontinuity
I chose as setting for this case study the market for anti-cancer drugs and its transition from cytotoxic
agents (e.g., antineoplastic antibiotics, etc.) to the radically new category termed targeted drugs (e.g.,
tyrosine kinase inhibitors, etc.), a transition brought about by the biotechnology revolution. Studies have
shown that this transition in anti-cancer drugs is a radical and competence-destroying change to the
preclinical (i.e., drug discovery) phase of R&D in this market (Rang, 2006),1 a description that is
consistent with that made of biotechnology’s impact at the industry level (e.g., Henderson, Orsenigo, and
Pisano, 1999; Rothaermel, 2001). Furthermore, as I will explain next, I collected interview material to
aid in understanding the specific industry dynamics, the operationalization of variables and the
interpretation of analyses in this market, and their responses supported this characterization. This setting
had many advantages, including its high research intensity (PhRMA, 2003) and the close connection
between product quality and profitability (Lu and Comanor, 1998). Furthermore, the biotechnology
revolution as the discontinuity of choice had advantages as well, including a wealth of data sources and
possible interviewees available.
1 Documented accounts of the development of approved targeted anti-cancer drugs can also be found in the natural
sciences literature and illustrate this point (e.g., Capdeville, et al., 2002).
6
As described in interview material in prior research (Sosa, 2006), targeted anti-cancer drug
development comprises two main variants: small-molecule and large-molecule drugs. Although both
variants of targeted anti-cancer drugs are radically different from standard chemotherapy in that targeted
drugs are developed through a “science-driven” approach, only large-molecule targeted drugs require,
among the capabilities new to incumbents, the one capability I refer to as rDNA/antibody techniques. It
is in this distinction that I could compare two variants of a technological change with the same attributes
(i.e., radical, competence-destroying, and sustaining2 to incumbents) but different access to incumbents.
The capability of interest, rDNA/antibody techniques, represents not one invention but a stream of
innovations through a particular technological trajectory (Dosi, 1982). In that sense, it represents what
strategy research refers to as a first-order dynamic capability (Teece, Pisano and Shuen, 1997; Winter,
2003). The capability is not a single set of resources and routines that are stable over time, to which firms
should pursue access at a single point in time. It is instead a series of innovations that are path-dependent
and whose definition is time-variant, that is, the relevant innovations to which an organization required
access in order to use this capability for the competitive advantage it conferred changed over time.
Whereas being able to re-engineer E.Coli bacteria to mass produce a protein represented a high level of
this capability in the mid-1970s, by the mid-1980s high levels of this capability required some dexterity in
mammalian cell culture. More importantly, although rDNA/antibody techniques as a capability coincided
with other capabilities required of incumbents in order to switch to biotechnology in that it entailed time
compression diseconomies and low levels of tradeability,3 rDNA/antibody techniques had one distinctive
feature that differentiated its origin and evolutionary path: its decreasing market specificity, that is, an
2 According to data by the American Cancer Society (2000), the market was far from satiation (i.e., far from saturation on a dimension of merit) on the levels of efficacy and safety, the main dimensions of merit considered in an anti-cancer drug, during the technological discontinuity under study. Indeed, the 5-year survival rate had changed from 50% in the years 1974-1976, to 51% in the years 1980-1982, to 59% in the years 1989-1995, a trend that proves the market in general was far from satiation, even though the difference in rates between 1974-1976 and 1989-1995 is statistically significant (p<0.05). Customer preferences were clearly the maximization of efficacy and safety, with ample room for improvement in those dimensions by any firm in competition making this a radical change that was sustaining in customer preferences (Christensen, 1997), comparable to prior cases in the literature (Henderson, 1993, 1995). Available patient accounts (e.g., Bazell, 1998) support as well this conclusion. 3 Time compression diseconomies refer to the characteristic of a capability as having some minimum time required to develop that cannot be expedited, whereas tradeability refers to not being traded in the open market, either for lack of supply or for difficulty of transfer (see Dierickx and Cool, 1989).
7
increasing number of markets to which the capability was applicable (Montgomery and Wernerfelt,
1988).
I organize the analysis in this study in two parts. Part one is the qualitative portion where I offer
evidence that rDNA/antibody techniques evolved with decreasing market specificity and therefore, that
anti-cancer drug market incumbents were unable to assess on time that this capability was also required in
cancer research, albeit only for large-molecule drugs. Part two is the quantitative portion of the study,
where I assess the impact that the delay in investment on rDNA/antibody techniques had, first, on
incumbents’ performance in this capability, and second, on incumbents’ R&D performance in the anti-
cancer drug market in particular. I present data sources and measures for both parts of the study next.
2.2 Data Sources and Measures
2.2.1 Qualitative Analysis
I collected data through 45 interviews (with four interviewees contacted repeatedly) ranging between 30
and 90 minutes each, with an evolving semi-structured interview guide. Interviewees included R&D
managers in large and small pharmaceutical firms (including R&D executives of three of the eight
incumbent firms in the sample), industry analysts, and scientists both from industry and from academia. I
complemented that data with historical material collected from Walsh’s (2003) report of large-molecule
drug development and customized searches in the PubMed database for historical background on specific
drugs. I also made use of the data in PJB Publications’ Pharmaprojects for selected information on the
introduction of drugs into clinical trials.
2.2.2 Quantitative Analysis
For part two of the study, I looked into the possible impact of the origin and technological trajectory of
rDNA/antibody techniques on the performance heterogeneity of incumbents as compared to diversifying
and de novo entrants competing in the anti-cancer drug market during the biotechnology discontinuity. In
order to test whether heterogeneity in investment in rDNA/antibody techniques led to heterogeneity in
R&D performance in the anti-cancer drug market, I first test whether heterogeneity in investment resulted
in sustainable heterogeneity in the performance of the capability itself (independent of the market
8
application). After testing for these differences, I proceed to test whether sustained heterogeneity in this
capability led to heterogeneity in R&D performance in the development of large- but not in small-
molecule targeted anti-cancer drugs.
Sample. To construct the sample of firms, I started by identifying all anti-cancer drugs in clinical trials
in the period 1989-2004 through PJB Publications’ database Pharmaprojects and then focused on the
firms responsible for them. In order to generate a sample that included firms with a clear intention to
compete in the anti-cancer drug market, I matched the firms from Pharmaprojects to the firms reported in
all available PhRMA surveys New Medicines in Development for Cancer (administered in 1988 and every
two years from 1989 to 2003). After excluding non-profit organizations, matching all cases to parent
company names only, and adjusting for mergers and acquisitions and missing data, I identified the final
sample, which comprises 165 firms (further detail of the sampling frame is given in Sosa, 2006). In
further analyses I also consider the original sample from Pharmaprojects without matching to PhRMA
surveys, which therefore expands to 704 firms.
Dependent Variables. In order to avoid the “tautology” problem, I performed separate tests on the
impact of differences in the time of investment first on the performance in the capability itself (i.e.,
rDNA/antibody techniques), and second on the performance on the R&D of anti-cancer drugs. I explain
next the two sets of dependent variables.
Competence in researching rDNA/antibody techniques (i.e., sustained advantage in the capability). I
estimated competence in this R&D capability by measuring the rate of production of patented innovations
in this area, as reported in Thomson Scientific’s Derwent World Patent Index (DWPI). This database is
constructed around innovations, and not patents. Therefore, each record represents a unique innovation
that matches to several patents, as catalogued by expert librarians. I asked expert interviewees to perform
the selection of relevant codes from the DWPI catalog. The resulting set of four specific DWPI manual
codes paired with the 165 firms in the sample generated a dataset of 1,375 patented innovations. I then,
based on these data, analyzed the rate of production of patented innovations through a Cox model with
repeated events following the design used previously in the management literature (Sørensen and Stuart,
9
2000). To do this, I used the earliest date of priority filing for the patented innovation as the time where
the event took place. I considered the start of the time at risk to be either January 1st of the first year in
the dataset (1979) or of the year of founding for the firm, whichever was latest, in the case of the firm’s
first patented innovation. All subsequent cases were set starting as the day immediately after the previous
event occurrence. I ran a second set of regressions taking into account the total number of forward
citations that the set of patents tied to each innovation generated. I incorporated this measure in Cox
regressions by using the total number of forward citations as frequency weights, that is, by duplicating
records by as many forward citations as the patents tied to the innovation had. This implies the second set
of regressions is then predicting rather the rate of production of forward citations in this area.
Competence in Targeted Anti-Cancer Drug Development (i.e., advantage in R&D performance in one
market: anti-cancer drugs). In order to test for differences in competence in the research and
development of targeted anti-cancer drugs, I constructed two separate dependent variables, which target
measures at different points of the value chain of a firm in this market (see Figure 1). The first measure
(under progress in this study) represents the outcome of drug discovery alone, whereas the second
measure represents the combined performance of the firm in drug discovery and clinical trial
management. The distinction is relevant because biotechnology is a competence-destroying change only
to the drug discovery process. The stages of clinical trial management have experienced a competence-
enhancing effect. Therefore, according to prior literature, incumbents should underperform entrants
based on the first dependent variable, with the second dependent variable being indeterminate. I describe
both dependent variables next.
Insert Figure 1
R&D Performance in Targeted Anti-Cancer Drug Discovery I identified the set of drug molecules that
each of the firms in the sample patented, regardless of whether they decided not to enter them into clinical
trials. To do this, I also used the DWPI database, this time identifying the code for anti-cancer drugs,
10
available in the period 1994-2004. Although still under construction, I present preliminary analyses of
the rate of production of patented drug molecules by the firms in the sample, by constructing a dataset
where each new patented drug molecule is a repeated event for the firm in question, alike the one used for
the rDNA/antibody techniques measurement.
R&D Performance in Targeted Anti-Cancer Drug Discovery and Development I identified the set of
drugs that each of the firms in the sample entered into clinical trials. Based on the Pharmaprojects
dataset, I identified when each drug entered and exited clinical trials in order to perform event history
analysis. I distinguished whether the drug was ultimately approved, discontinued or right-censored and
analyzed the data through Cox regression focusing on the approval event, treating discontinuations and
right-censored cases as right-censored (although alternative specifications for the discontinued records did
not alter results).
Independent Variables. The principal interest was to distinguish whether performance advantages
accrued to some categories of firms in particular, and whether there was a difference between the two
variants of the new technology that stemmed from the use of rDNA/antibody techniques. Therefore, the
following binary variables were of primary importance.
Small- vs. Large-Molecule Targeted Drugs. I classified targeted anti-cancer drugs as large- vs. small-
molecule drugs for the two different dependent variables measuring R&D performance in this market as
follows.
R&D Performance in Targeted Anti-Cancer Drug Discovery. In this dataset, I classified targeted anti-
cancer drugs through the sub-categories (e.g., interferons, interleukins) available in their classification,
according to the information that interviewees had provided.
R&D Performance in Targeted Anti-Cancer Drug Discovery and Development. In this dataset, I
classified targeted anti-cancer drugs through the information in the Pharmaprojects database. For large-
molecule drugs, this information is directly provided in the database. For small-molecule drugs, I
matched the mechanism of action reported in Pharmaprojects to the mechanisms of action described in
industry reports (e.g., Bear Sterns, 2002; Stephens Inc., 2002; UBS Warburg, 2001) as targeted (in the
11
end, mainly comprising angiogenesis and kinase inhibitors). I then generated a dummy variable “Large-
Molecule” to distinguish these two classes of drugs.
Firm Categories. I distinguished among incumbents, diversifying and de novo entrants based on
whether a firm had been present in the anti-cancer drug market prior to 1983 (the start of the
biotechnology revolution in that market) in which case it was an incumbent, and if not, whether it had
been operating in any market prior to its incursion into anti-cancer drug development, in which case it
was a diversifying entrant. The decision tree and data sources used to classify all firms in anti-cancer
drug market incumbents, diversifying or de novo entrants are presented in Figure 2. Further detail is
available in Sosa (2006).
Insert Figure 2
rDNA/Antibody Pioneers. As I will explain, the evolutionary path of rDNA/antibody techniques did
lead to differences in timing of investment across firms. Because the anti-cancer drug market is one of
the last markets to be reached by rDNA/antibody techniques and their resulting large-molecule drugs (due
to the complexity of this R&D capability), I took the date that large-molecule drugs entered clinical trials
with a clear anti-cancer application, namely 1995 (as reported in Colwell, 2002) as the cut-off date to
identify rDNA/antibody pioneers. I then used Walsh’s (2003) report to identify all large-molecule drugs
approved up to 1994 irrespective of market application, and then combined that information with
information on Pharmaprojects to pinpoint the developing firms for those drugs. I classified as an
rDNA/antibody pioneer a firm that had a large-molecule drug that (1) had been developed in-house, (2)
had been approved in 1994 or before (irrespective of market application), and (3) was the first one in its
active ingredient. The identification of these rDNA/antibody pioneering firms is shown in Table 1.
Insert Table 1
12
Control Variables. In all analyses using Cox regression, I control for the cumulative introduction (of
patented innovations or drugs in clinical trials, for the different sets of regressions). In Cox regressions
for rate of production of patented innovations weighted by forward citations, I also controlled for the
number of different patents applied for per innovation, since this number could artificially increase the
number of forward citations. In analyses of drug approval, I controlled for the novelty of the drug
through the variable “drug novelty,” defined as the natural logarithm of the inverse of the chronological
place of introduction that the drug holds on the list of drugs within the same mechanism of action (a
replica of the measure included in Guedj and Scharfstein, 2004). In analyses of drug approval, I also
controlled for the presence of an R&D alliance through a dummy variable with value 1 if the drug had an
R&D alliance associated with it reported in the cancer sub-section of the Windhover’s Pharmaceutical
Strategic Alliances collection 1989-2003. Although I could not identify a separate binary variable for
drug acquisitions, prior research (Sosa, 2006) has shown that in this market the inter-firm category
acquisition of drugs is low and more frequently present in de novo firms than in either diversifying
entrants or incumbents, hence lowering the concern on that source of variance. Likewise, although in
pharmaceuticals projects could be outsourced in their clinical trial component, the level of outsourcing in
cancer is extremely low, second-to-last after ophthalmology, with a mean outsourcing level of 10.3% in
the period 1995-1999, and even lower levels in the years preceding 1995 (Azoulay, 2004). Hence,
concern for clinical trial outsourcing as a source of variance is as well insignificant. Lastly, in
preliminary analyses of drug approval I also attempt to distinguish diversifying entrants that had prior
oncology research from those who did not in accordance to prior research in this market (Sosa, 2006).
3. Analysis and Results
3.1 The Origin and Evolution of rDNA/Antibody Techniques
As mentioned, rDNA/antibody techniques have made possible the mass-production of one variant of
biotechnology-based drugs: large-molecule drugs. For instance, interferon alpha-2, the active ingredient
in Intron A® (a recently approved, large-molecule targeted anti-cancer drug), is a cytokine naturally
produced in the human body in small quantities (Walsh, 2003). rDNA/antibody techniques made it
13
possible to produce interferon alpha-2 in therapeutically and hence commercially feasible amounts. In
fact, interviewees report innovations in rDNA/antibody techniques evolved through a series of stages, that
as I will argue in this section, decreased in market specificity (i.e., increased in the scope of markets to
which they were applicable [Montgomery and Wernerfelt, 1988]).
rDNA/antibody innovations were first developed to mass-produce proteins (i.e., large-molecule drugs)
occurring naturally in the human body. The characterization of such proteins had been performed in
academic research laboratories and was publicly available. Several of the first large-molecule drugs to
reach the market were used in the treatment of enzyme deficiencies (e.g., diabetes mellitus), diseases in
which not only the protein but also its therapeutic value (i.e., its connection to disease treatment) were
common knowledge in the scientific community. In these initial markets, firms were competing in terms
of competence in process design alone. This comprises stage I in the evolution of rDNA/antibody
techniques.
A case in point is insulin, the first product for which the radically new rDNA/antibody techniques were
commercially used. Insulin’s principal therapeutic value is the treatment of diabetes mellitus, a disease in
which patients lack natural insulin production. The enzyme received the name “insulin” in 1909, but it
was not until 1921-1922 that researchers at the University of Toronto isolated the enzyme and proved its
effect in regulating sugar metabolism (Rosenfeld, 2002). By 1976 when Genentech invested in
rDNA/antibody-based process innovations for mass-production of “artificial” insulin to be
commercialized by Eli Lilly and Co. (Christensen, 1996), the enzyme had been in commercial production
by semi-synthetic processes since 1923 (when Eli Lilly and Co. achieved successful yield and
standardization of the first mass-production method). By 1976 therefore, both the molecule for insulin
and its therapeutic connection to diabetes mellitus were public knowledge.
It was not until later, as rDNA/antibody techniques evolved, that gradually other known enzymes for
which no connection to disease treatment was known began to be researched in-depth. This is then stage
II of the evolution of rDNA/antibody techniques. A case in point is that of erythropoietin, commonly
referred to as Epo, an enzyme today commercially available as Amgen’s best-selling large-molecule drug
14
for anemia treatment, Epogen®. According to scientist J.W. Fisher’s (1998) own account of his and
others’ breakthrough research in “the quest for erythropoietin,” one of the most important academic
papers confirming the existence of Epo was published in 1950, however:
“until the gene for Epo was cloned by Lin et al. [1985] at Amgen and Jacobs et
al. [1985], Epo was [erroneously] thought to be produced in the glomerular
epithelial cells. The ability to clone made it possible [to determine Epo’s
appropriate source and therapeutic value]” (p. 10).
As the rDNA/antibody techniques developed, the therapeutic potential of large-molecule drugs grew in
relevance and ultimately a new product class emerged. This new product class comprises stage III in the
evolution of this technological competence. The pharmaceutical industry is currently in stage III, and
large-molecule drugs that enter clinical trials go beyond those naturally occurring in the human body, to
include as well laboratory-designed drugs. Clearly, the development of the latter requires investment in
terms of both manufacturing process and product design and includes markets with higher profitability
prospects (e.g., anti-cancer drugs). A case in point in stage III is Herceptin®, the new targeted anti-
cancer large-molecule drug designed by Genentech that targets Her-2 expressing aggressive breast
cancers (Bazell, 1998).
Interviewees coincided in the description of the historical progression of the R&D of large-molecule
drugs in the three stages described above: (I) a class of known proteins with known connections to disease
treatment (e.g., insulin); (II) a class of known proteins with unknown connections to disease treatment
(e.g., Epo); (III) a newly born class of engineered proteins (e.g., Herceptin®). Indeed, in a recent article
in the natural sciences, Leader, Baca and Golan (2008) have argued for the distinction of these three
classes.
15
Each disease treatment is, in rough terms, a stand-alone market.4 In fact, large-molecule drugs currently
available in the market can be classified into the three categories mentioned above. The resulting three
broad classes are shown in Table 2.
Insert Table 2
Based on this classification and the list of all large-molecule drugs approved in the USA up to 2003 as
reported in Walsh (2003), I constructed Figure 3 to illustrate the evolution of the three stages.
Insert Figure 3
What the three-stage progression of the applicability of rDNA/antibody techniques implies for our
understanding of the evolution of this capability is that the competence widened its market specificity
over time. More importantly, this temporal difference in market specificity led to delayed investment for
some firms, among them, precisely the eight firms that constituted the group of incumbents in the largest
market within pharmaceuticals: anti-cancer drugs.
Indeed, anti-cancer drug market incumbents have publicly declared their delay in investment as a result
of excess technological uncertainty, as then-president for international R&D at Hoffmann-La Roche
explained:
“At first, the pharmaceutical industry was slow to react to the challenge. Many
companies regarded [rDNA/antibody techniques] as an esoteric science with
little promise for substantial economic returns… This initial promise has since
been convincingly redeemed…” Drews (1993: S16).
In fact, expansion into markets with intense use of rDNA/antibody techniques has prompted some of
the anti-cancer drug market incumbents to resort to drug acquisitions, precisely because the lack of access
to rDNA/antibody techniques as a capability renders internal R&D unattainable. For example, whereas
4 Cancer and many therapeutic areas comply not only with the basic definition of a market (i.e., a set of products that are substitutes for one another) but also with the additional requirements discussed in the economics literature for high-tech settings (Sutton, 1998).
16
according to Pharmaprojects, the anti-cancer drug market incumbent Bristol-Myers Squibb (BMS) has
not had internally developed HIV drugs that reached clinical trials in the period 1989-2004, the firm
currently leads in sales in the HIV application. All their successful HIV drugs are acquisitions: Reyataz®
acquired from Novartis, Videx® acquired from National Institutes of Health, and Zerit® acquired in early
stages from Yale University.5 In a non-confidential interview in March 2007, an R&D executive for
BMS explained their complex pattern of decisions: whereas BMS did invest on time in the application-
specific side (Sosa, 2006) of large-molecule anti-cancer drug development (mainly through investment in
the Lewis-Y mechanism of action to target cancer treatment), they did fall behind in rDNA/antibody
techniques, that is, the technology-specific side of large-molecule drug development.
3.2 The Impact of Differences in Time of Investment on Performance in the Capability for
rDNA/Antibody Techniques
The increase over time in the number of markets for which rDNA/antibody techniques were applicable
could have generated heterogeneity in investment in this technological competence, which would then
lead to differences in the competence to research further rDNA/antibody techniques. This is therefore the
hypothesis to test in this section: whether rDNA pioneers had a sustained advantage in the competence to
research rDNA/antibody techniques. Table 3 offers descriptive statistics and Table 4 offers the Cox
model results for regressions predicting the rate of production of patented innovations in the
rDNA/antibody area in the period 1979-2004.
Insert Table 3
Insert Table 4
Notice rDNA pioneers have an advantage (model 3) and their advantage is larger than that of
incumbents (test of coefficients being equal is rejected at p < 0.03). The baseline in models 1 and 2 is
5 The firm’s compound leading their HIV pipeline as of March 2007, was as well acquired, this time from DuPont Pharma.
17
diversifying entrants (the omitted category) and in both cases de novo firms incur a disadvantage in
comparison (their hazard rates are < 1 in both models). In model 3, the baseline (the omitted category in
this case) is diversifying entrants that are not rDNA pioneers and de novo firms are still at a significant
disadvantage (their hazard rate is still < 1). This analysis implies that pioneers in the area of
rDNA/antibody techniques accrued an advantage in that R&D capability that persisted until at least 2004,
the year of end of observation.
Because not all innovations are of equal importance, I try to control for such differences by creating
frequency weights based on the forward citations of the patents tied to each innovation. Table 5 presents
descriptive statistics whereas Table 6 presents results of the Cox regressions with frequency weights.
Prior conclusions do not change: rDNA pioneers accrued a sustained advantage in the performance of
research in rDNA/antibody techniques, and incumbents followed in second place.
Insert Table 5
Insert Table 6
3.3 The Impact of Heterogeneity in the Capability for rDNA/Antibody Techniques on
Performance in the R&D of Anti-Cancer Drugs
So far, I have offered evidence supporting the result that the origin and evolution of rDNA/antibody
techniques with its gradual decrease on market specificity translated into differences in time of investment
in this technological capability, and more importantly, into a sustained advantage in the performance of
subsequent research in rDNA/antibody techniques. In this section, I look for evidence that the sustained
advantage in the research of rDNA/antibody techniques resulted in an advantage on the execution of
research and development of large-molecule targeted anti-cancer drugs, but not on that of small-molecule
targeted anti-cancer drugs. Whereas prior research found incumbents had an absolute advantage in the
18
research and development of all targeted anti-cancer drugs, in this section I hypothesize that for the sub-
set of targeted anti-cancer drugs that are large molecules, rDNA pioneers would have the largest
advantage. I perform this test using Cox regressions predicting whether and when the approval of a drug
takes place. In order to gain statistical power, I split the sample of targeted anti-cancer drugs into the sub-
sample of small-molecule and that of large-molecule drugs. Table 7 presents descriptive statistics for the
full set of drug molecules patented during drug discovery, whereas Table 8 presents preliminary analyses
(of the pooled sample as well as of small-molecule and large-molecule targeted drugs alone).
Insert Table 7
Insert Table 8
Table 8 Model 1 shows incumbents follow behind diversifying entrants in the performance of R&D of
large-molecule targeted anti-cancer drugs (see the interaction “incumbent X large-molecule” is
statistically significant and > 1). Model 2 shows this effect is driven by the set of rDNA pioneers (see the
interaction “rDNA pioneer X large-molecule” is statistically significant and > 1). In the sub-samples in
Models 3 and 4 the underlying dynamics are present: incumbents lead in small-molecule targeted anti-
cancer drug discovery but follow second in the case of large-molecule drugs (incumbents also lead, as
expected, in cytotoxic anti-cancer drug discovery, as seen in the main effect of “incumbent” in Models 1
and 2).
Moving to the analysis of both drug discovery and clinical trial management together, Table 9 presents
descriptive statistics whereas Table 10 presents results from the analyses.
Insert Table 9
Insert Table 10
19
Table 10 shows the sub-sample of large-molecule targeted drugs only, split from the pooled sample in
an attempt to gain statistical power (the sub-sample of small-molecule targeted drugs only does not
achieve statistical significance but is of the expected signs and is shown in Appendix 1). In Table 10,
Models 1 and 2 show diversifying entrants are leading in R&D performance in the large-molecule variant,
and Model 3 shows this is driven as expected by the impact of rDNA pioneers. In model 4, I expand the
sample from 165 to 704 firms (in preliminary analyses and therefore without controls) in an attempt to
gain statistical power and identify further the separate effects. Although now incumbents can be
distinguished from all entrants that are not rDNA pioneers (see the variable “incumbent” is now >1 and
statistically significant), other effects cannot be distinguished. In particular, in Model 4 I try to control for
the presence of prior oncology research, a variable shown in previous research (Sosa, 2006) to determine
R&D performance in the drug discovery of anti-cancer drugs. The subset of firms that had neither a pre-
history in rDNA/antibody techniques or in oncology research seem at a disadvantage (see the set
“Diversifying, No rDNA Pioneer, No prior oncology research” is indistinguishable from the baseline,
whereas groups with at least one of the two capabilities are >1 and significant). Nonetheless, the sample
seems too small to distinguish the effects and the test of coefficients fails in all cases.
4. Discussion
I motivated this paper by asking whether we could better understand differences in performance during
a technological discontinuity by moving beyond competence destruction and into competence access.
Using the impact of biotechnology on anti-cancer drug development, a change that has been radical
(Henderson and Clark, 1990) and competence-destroying (Tushman and Anderson, 1986) for incumbents,
I compared two variants that differed only on their use of one new technological capability among many:
rDNA/antibody techniques. Based on a combination of qualitative and quantitative methodology, I found
preliminary evidence that the market specificity of this capability decreased over time (i.e., its market
coverage increased). Such temporal difference in market specificity generated differences in time of
investment across firms interested in pharmaceuticals. This heterogeneous timing of investment then
cascaded into differences in market-level competition. In the one downstream market I measured, the
20
anti-cancer drug market, the largest market within the pharmaceutical industry, I found preliminary
evidence that the sub-set of diversifying entrants who were pioneers in rDNA/antibody techniques
accrued the largest competitive advantage in the execution of all stages of R&D. They outperformed all
other firms, including anti-cancer market incumbents, in the generation of subsequent rDNA/antibody
innovations, as well as in the overall R&D performance of the one variant of the radically new technology
that made use of rDNA/antibody techniques (namely, large-molecule targeted anti-cancer drugs).
With this case study, I hope to contribute to the literature on technological disruption, extending it
beyond our understanding of competence destruction and into the discussion of competence access. I
show how competence destruction is not in itself a predictor of underperformance for incumbent firms.
Indeed, anti-cancer market incumbents sustained their competitive advantage in the one variant of
biotech-based (i.e., targeted) anti-cancer drugs that made use of new capabilities to which they had access
from the start: small-molecule targeted anti-cancer drugs. It was only in the case where these incumbents
saw the value of their traditional competences destroyed paired with the lack of access to the new
competences required, that they underperformed behind the sub-set of entering firms that had accrued an
advantage in the technology.
Furthermore, the process that led to lack of foresight for incumbents in this case did not involve a
change in customer preferences as has been argued before in the literature (Christensen, 1997; Tripsas,
2006). In this case, the change is radical in technologies but sustaining in customer preferences and
incumbents still lacked foresight due to the particular origin and evolutionary path that one technological
capability, rDNA/antibody techniques, followed. I therefore seek to contribute with this paper to the
growing body of literature interested in the origin and evolution of capabilities as a precursor to
heterogeneity in firm performance (Ahuja and Katila, 2004; Ethiraj et al., 2005).
21
References
Ahuja, G. and R. Katila, 2004. “Where do Resources come from? The Role of Idiosyncratic Situations,” Strategic Management Journal, 25, 887-907.
Barney, J.B., 1991. “Firm Resources and Sustained Competitive Advantage,” Journal of Management, 17(1): 99-120.
Bazell, R., 1998. Her-2: The Making of Herceptin, a Revolutionary Treatment for Breast Cancer, Random House.
Bear Sterns, 2002. Biotechnology: Cancer – Concepts and Controversies, Volume 1. Colwell, R.R. 2002. “Fulfilling the Promise of Biotechnology,” Biotechnology Advances, 20: 215-228. Dierickx, I., K. Cool. 1989. Assets Accumulation and Sustainability of Competition Advantage.
Management Science. 35(12) 1504-1514. Dosi, G. 1982. Technological paradigms and technological trajectories, a suggested interpretation of the
determinants and directions of technical change. Research Policy, 11(3): 147-162. Drews, J. 1993. “Into the 21st Century: Biotechnology and the Pharmaceutical Industry in the Next Ten
Years,” Biotechnology, 11: S16-S20. Ethiraj, S.K., P. Kale, M.S. Krishnan, and J.V. Singh, 2005. “Where do Capabilities come from and How
do they matter? A Study in the Software Services Industry,” Strategic Management Journal, 26, 25-45.
Fisher, J.W., 1998. “A Quest for Erythropoietin over Nine Decades,” Ann. Rev. Pharmacol. Toxicol., 38:1-20.
Guedj, I. and D. Scharfstein. 2004. Organizational Scope and Investment: Evidence from the Drug Development Strategies and Performance of Biopharmaceutical Firms. NBER Working Paper 10933.
Helfat, C.E., and M.B. Lieberman. 2002. “The Birth of Capabilities: Market Entry and the Importance of Pre-History,” Industrial and Corporate Change, 11(4): 725-760.
Henderson, R.M. and K.B. Clark. 1990. Architectural innovation: the reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly, 35(1): 9-30.
______________, 1993. Underinvestment and Incompetence as responses to Radical Innovation – Evidence from the Photolithographic Alignment Equipment Industry. Rand Journal of Economics, 24(2): 248-270.
______________, 1995. Of Life-Cycles Real and Imaginary – The Unexpectedly Long Old-Age of Optical Lithography. Research Policy. 24(4) 631-643.
______________, L. Orsenigo, and G. Pisano, 1999. “The Pharmaceutical Industry and the Revolution in Molecular Biology: Interactions Among Scientific, Institutional, and Organizational Change,” in D.C. Mowery and R.R. Nelson (eds.) Sources of Industrial Leadership, Cambridge University Press.
Leader, B., Q.J. Baca, D.E. Golan. 2008. Protein Therapeutics: A Summary and Pharmacological Classification. Nature Reviews Drug Discovery, 7: 21-39.
Lu, Z.J. and W.S. Comanor, 1998. Strategic Pricing of New Pharmaceuticals. The Review of Economics and Statistics, 80(1): 108-118.
Montgomery, C.A., and B. Wernerfelt. 1988. “Diversification, Ricardian Rents, and Tobin’s Q,” Rand Journal of Economics, 19(4): 623-632.
Pharmaceutical Research and Manufacturers of America [PhRMA]. 2003. Pharmaceutical Industry Profile, Washington, D.C.
Rang, H.P. 2006. The Drug Discovery Process: General Principles and Some Case Histories. H.P. Rang (ed.) Drug Discovery and Development: Technology in Transition. Churchill Livingstone Elsevier.
Rosenfeld, L., 2002. “Insulin: Discovery and Controversy,” Clinical Chemistry, 48(12): 2270-2288. Rothaermel, F.T. 2001. Incumbent’s Advantage through Exploiting Complementary Assets via Interfirm
Cooperation. Strategic Management Journal. 22(6/7) 687-699.
22
Rumelt, R.P., 1987. “Theory, Strategy, and Entrepreneurship,” in D. Teece (ed.) The Competitive Advantage Challenge, Ballinger, Cambridge, MA.
Sosa, M.L., 2006. “Decoupling Market Incumbency from Organizational Experience: A Study of Biotechnology’s Impact on the Market for Anti-Cancer Drugs,” Unpublished Doctoral Dissertation, Sloan School of Management.
Stephens Inc., 2002. Cancer, Revolutionary Therapies, Innovative Investment Opportunities, 2nd ed. Sutton, J. 1998. Technology and Market Structure. MIT Press, Cambridge, MA. Teece, D.J. 1980. Economies of scope and the scope of the enterprise. Journal of Economic Behavior &
Organization. 1(3) 223-247. ______________, 1986. Profiting from Technological Innovation: Implications for Integration,
Collaboration, Licensing and Public Policy. Research Policy. 15 285-305. ______________, G. Pisano, A. Shuen. 1997. Dynamic capabilities and strategic management. Strategic
Management Journal. 18(7) 509-533. Tripsas, M. 1997. Unraveling the process of creative destruction: Complementary assets and incumbent
survival in the typesetter industry. Strategic Management Journal. 18 119-142. ______________, 2006. “Customer Preference Discontinuities: A Trigger for Radical Technological
Change,” Managerial and Decision Economics. Tushman, M. and P. Anderson. 1986. “Technological Discontinuities and Organizational
Environments,” Administrative Science Quarterly, 31(3): 439-465. UBS Warburg, 2001. The Cancer Matrix: Exploring the Future of Cancer Therapy, Walsh, G., 2003. Biopharmaceuticals: Biochemistry and Biotechnology, John Wiley & Sons, Ltd.,
Second Edition.
23
Tables Table 1
Pre-1995 USA Approvals of rDNA/Antibody-based Products and their Developing Firms
Year* Brand Name*
Active Ingredient
Indication (Market)*
Commercializing Firm*
Developing Firm**
rDNA Pioneer
1982 Humulin Insulin Diabetes mellitus Eli Lilly Genentech
1985 Protropin Human growth hormone (hGH)
hGH deficiency in children Genentech Genentech
1986 Intron A Interferon alpha 2 Cancer, genital warts, hepatitis Schering Plough Biogen
1986 Roferon A Interferon alpha 2 Hairy cell leukemia Hoffman-La Roche Genentech
1986 Recombivax Hepatitis B virus surface antigen Hepatitis B vaccine Merck
1986 Orthoclone OKT3 Muromomab CD3
Reversal of acute kidney transplant
rejection
Ortho Biotech (Johnson & Johnson)
Ortho Biotech (Johnson & Johnson)
1987 Activase Tissue
plasminogen activator (tPA)
Acute myocardial infarction Genentech Genentech
1987 Humatrope hGH hGH deficiency in children Eli Lilly Eli Lilly
1989 Epogen Epoetin alpha Anemia Amgen Amgen
1990 Procrit Epoetin alpha Cancer-related anemia Ortho Biotech (Johnson & Johnson) Amgen
1990 Actimmune Interferon gamma 1
Chronic granulomatous disease Genentech Genentech
1991 Novolin Insulin Diabetes mellitus Novo Nordisk Novo Nordisk
1991 Leukine
Granulocyte macrophage
colony-stimulating factor
(GM-CSF)
Autologous bone marrow transplantation
Amgen and Schering AG Immunex1
1991 Neupogen Filgrastim Chemotherapy-induced neutropenia Amgen Amgen
1992 Recombinate Factor VIII Hemophilia A Baxter / Wyeth Genetics Institute2
1992 Proleukin Interleukin 2 Renal cell carcinoma Chiron Chiron
1992 OncoScint CR/OV
Satumomab pendetide
Detection/staging, colorectal and ovarian
cancers Cytogen Cytogen
1993 Bioclate Factor VIII Hemophilia A Centeon Genetics Institute 1993 Kogenate Factor VIII, 2nd
generation Hemophilia A Bayer Bayer
1993 Betaseron Interferon beta 1 Relapsing multiple sclerosis
Berlex laboratories and Chiron Chiron
1993 Pulmozyme Dornase alpha Cystic fibrosis Genentech Genentech 1994 Nutropin hGH, 2nd
generation hGH deficiency in
children Genentech Genentech
1994 ReoPro Abciximab Prevention of blood clots Centocor State University,
NY
1994 Cerezyme Beta glucocerebrosidase Gaucher’s disease Genzyme Genzyme
Sources: * Walsh (2003) ** Pharmaprojects
24
Table 2
Classes of Large-Molecule Drugs that Evolved Chronologically into a New Product Class
Stage I
protein and connection to disease known
Stage II only protein known
Stage III new product class
Insulin Epo Factor VIII Interferons
Human Growth Hormone Glucocerebrosidase
Interleukins Monoclonal-Antibody-based
products
25
Table 3 Competence in rDNA/Antibody Techniques
Descriptive Statistics and Correlation Matrix for Analysis of Rate of Production of Patented Innovations
(1,452 Spells, 1,375 Events)
Count Mean Std.Dev. Min. Max. (1) Incumbent 329 (2) De Novo 252 (3) rDNA Pioneer 450 (4) De Novo, no rDNA Pioneer 245 (5) Cumulative 319 247 0 846
(1) (2) (3) (4) (5)
(1) Incumbent 1 (2) De Novo -0.25 1 (3) rDNA Pioneer -0.36 -0.28 1 (4) De Novo, no rDNA Pioneer -0.24 0.98 -0.30 1 (4) Cumulative -0.34 -0.37 0.15 -0.36 1
26
Table 4
Competence in rDNA/Antibody Techniques Cox Model Analysis of Rate of Production of Patented Innovations
(1,452 Spells, 1,375 Events) All Coefficients in Hazard Rates
Model 1 Model 2 Model 3
Incumbent 0.99 (0.05)
1.16* (0.08)
1.79*** (0.15)
De Novo 0.40*** (0.03)
0.47*** (0.04)
rDNA Pioneer 2.13*** (0.17)
De Novo, no rDNA Pioneer 0.71*** (0.07)
Cumulative 1.00*** (0.00)
1.00*** (0.00)
Log Likelihood -8,660 -8,648 -8,595 + p < 0.1, * p < .05, ** p < .01, *** p < .001 Standard errors in parentheses.
≠ coefficients p < 0.03
27
Table 5 Competence in rDNA/Antibody Techniques
Descriptive Statistics and Correlation Matrix for Weighted Rate of Production of Patented Innovations
(4,947 Spells, 4,870 Events)
Count Mean Std.Dev. Min. Max. (1) Incumbent 1,411 (2) De Novo 955 (3) rDNA Pioneer 1,323 (4) De Novo, no rDNA Pioneer 903 (5) Cumulative 764 514 0 1,710 (6) Number of Patents per Innovation 9.3 8.3 0 60
(1) (2) (3) (4) (5) (6)
(1) Incumbent 1 (2) De Novo -0.31 1 (3) rDNA Pioneer -0.38 -0.24 1 (4) De Novo, no rDNA Pioneer -0.30 0.97 -0.29 1 (5) Cumulative -0.22 -0.34 0.28 -0.33 1 (6) Number of Patents per Innovation 0.19 -0.05 0.01 -0.11 -0.38 1
28
Table 6
Competence in rDNA/Antibody Techniques Cox Model Analysis of Rate of Production of Patented Innovations
with FW Citations as Frequency Weights (Rate of FW Citation Production)
(4,947 Spells, 4,870 Events) All Coefficients in Hazard Rates
Model 1 Model 2 Model 3
Incumbent 0.90** (0.03)
1.36*** (1.01)
2.04*** (0.10)
De Novo 0.60*** (0.02)
1.00 (0.04)
rDNA Pioneer 2.21*** (0.11)
De Novo, no rDNA Pioneer 1.58*** (0.08)
Cumulative 1.00*** (0.00)
1.00*** (0.00)
Number of Patents per Innovation
0.99** (0.00)
Log Likelihood -36,684 -36,016 -35,827 + p < 0.1, * p < .05, ** p < .01, *** p < .001 Standard errors in parentheses.
≠ coefficients p < 0.07
29
Table 7 Drug Discovery
Descriptive Statistics and Correlation Matrix for Analysis of Rate of Production of Patented Molecules
(14,790 Spells, 14,717 Events)
count (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) Incumbent 5,572 1 (2) De Novo 176 -0.09 1 (3) rDNA Pioneer 2,487 -0.35 0.14 1 (4) De Novo, no rDNA Pioneer 65 -0.05 0.61 -0.03 1 (5) Large-Molecule 2,898 -0.17 0.05 0.50 -0.00 1 (6) Incumbent X Large-Molecule 595 0.26 -0.02 -0.09 -0.01 0.41 1 (7) De Novo X Large-Molecule 68 -0.05 0.62 0.12 0.16 0.14 -0.01 1 (8) rDNA Pioneer X Large-Molecule 1,593 -0.27 0.08 0.77 -0.02 0.70 -0.07 0.16 1
(9) De Novo, no rDNA Pioneer X Large-Molecule 11 -0.02 0.25 -0.01 0.41 0.06 -0.00 0.40 -0.00 1
30
Table 8
Drug Discovery Cox Model Analysis of Rate of Production
of Patented Molecules All Coefficients in Hazard Rates
All Drugs Only Large
Targeted
Only Small
Targeted
Model 1 Model 2 Model 3 Model 4
Incumbent 1.84*** (0.03)
1.91*** (0.04) 1.76***
(0.08) 1.70***
(0.06)
De Novo 0.53*** (0.04)
rDNA Pioneer 1.24*** (0.04) 1.98***
(0.08) 1.55*** (0.08)
De Novo, no rDNA Pioneer 0.40*** (0.03) 0.43***
(0.06) 0.41*** (0.05)
Large-Molecule 1.75*** (0.05)
1.10** (0.04)
Incumbent X Large-Molecule 0.63*** (0.03)
1.01 (0.05)
De Novo X Large-Molecule 0.74** (0.08)
rDNA Pioneer X Large-Molecule 1.76*** (0.10)
De Novo, no rDNA Pioneer X Large-Molecule 1.02
(0.16)
Spells 14,790 14,790 2,898 3,896 Events 14,717 14,717 2,898 3,896 Log Likelihood -127,128 -126,972 -20,513 -28,399 + p < 0.1, * p < .05, ** p < .01, *** p < .001 Standard errors in parentheses.
Test of coeff. p < 0.07
Test of coeff. p < 0.004
31
Table 9 Overall R&D Competence: Drug Discovery and Clinical Trial Management
Descriptive Statistics and Correlation Matrix Only Targeted Large-Molecule Drugs
(N = 638)
Count Mean Std.Dev. Min. Max. (1) Incumbent 47 (2) Diversifying 212 (3) rDNA Pioneer 73 (4) Diversifying, no rDNA Pioneer 152 (5) Cumulative 476 238 6 914 (6) Drug Novelty -2.55 1.71 -5.3 0 (7) R&D Alliance 14
(1) (2) (3) (4) (5) (6) (7)
(1) Incumbent 1 (2) Diversifying -0.20 1 (3) rDNA Pioneer -0.10 0.37 1 (4) Diversifying, no rDNA Pioneer -0.16 0.79 -0.20 1 (5) Cumulative -0.22 0.11 0.02 0.11 1 (6) Drug Novelty 0.10 0.12 0.09 0.08 0.14 1 (7) R&D Alliance -0.00 -0.04 -0.05 -0.01 -0.10 -0.02 1
32
Table 10
Overall R&D Competence: Drug Discovery and Clinical Trial Management Cox Model Analysis of Drug Approval
Only Targeted Large Molecules All Coefficients in Hazard Rates
Model 1 Model 2 Model 3 Model 4
Incumbent 3.05 (3.34)
2.26 (2.37)
4.37 (4.94)
7.05+ (8.09)
Diversifying 3.91* (2.22)
4.01* (2.50)
rDNA Pioneers 14.47*** (9.54)
rDNA Pioneers, With prior oncology research 11.36*
(12.58)
rDNA Pioneers, No prior oncology research 21.85***
(14.74)
Diversifying, no rDNA Pioneer 3.57 (3.02)
Diversifying, No rDNA Pioneer, With prior oncology research
8.38* (7.97)
Diversifying, No rDNA Pioneer, No prior oncology research
1.11 (1.21)
Cumulative Introduction 0.99 (0.00)
1.00 (0.00)
1.00 (0.00)
Drug Novelty 1.28+ (0.16)
1.11 (0.15)
R&D Alliance 2.69 (2.96)
4.23 (4.98)
Spells 638 638 638 1289 Events 15 15 15 15 Log Likelihood -64 -62 -57 -56.7 + p < 0.1, * p < .05, ** p < .01, *** p < .001 Standard errors in parentheses.
Test of coeff. p < 0.5
Test of coeff. p < 0.3
33
Figures
Figure 1 Typical Value Chain in the Pharmaceutical Industry
Adapted from Fig. 4.1 (p.44) in Rang (2006)
R&D Process Drug Discovery (Preclinical) Development (Clinical)
Target Selection
Target Validation
Lead Finding
Lead Optimization
Animal Studies Phase I Phase II Phase III
Commercialization
Investigational New Drug (IND)
New Drug Application
(NDA)
34
Figure 2
Did the firm have an approved cytotoxic anti-cancer drug before 1983?* Source: FDA-CDER Oncology Tools List, FDA approved drugs list, PDR collection 1947-2005.
Has the firm derived meaningful yearly revenue (as proxied by positive sales) from old-technology anti-cancer drugs in the period 1990-2002, or had a generic version introduced?** Source: Company Annual Reports, Med Ad News Top 500 Prescription Drugs Reports, Company’s Customer Service
Does the firm have a targeted anti-cancer drug in clinical trials? Source: Pharmaprojects 1989-2004
Does the firm have a targeted anti-cancer drug in clinical trials? Source: Pharmaprojects 1989-2004
Did the firm derive revenue from other market(s) before entering the market for anti-cancer drugs under the new technology (i.e., with targeted drugs)? Source: Company Websites (Corporate History section)
Yes No
Yes
No
Incumbent, investing in
the new technology
Incumbent, NOT investing
in the new technology
No
Firm is not in the market for anti-cancer drugs
Yes
Diversifying entrant
De Novo entrant
* This requirement ensures that the firm was an incumbent to the market prior to its investment in new-technology anti-cancer drugs (as opposed to just deciding to enter the market investing in both old and new technologies in parallel). The year 1983 was when the first Targeted Anti-Cancer Drug was launched on the market, and I therefore use it as a milestone. ** This requirement ensures that the firm did not leave the market and come back to it because of the new technology’s effect on lowering barriers to entry. If a firm exits a market before the transition due to the radical technological change starts, then that firm is not in the market at the time of the radical change and therefore is not an incumbent. If it stays away from the market, then it is out of the scope of relevance for this study. If it comes back after several years, investing in the new technology, then it is a diversifying entrant.
Yes No Yes No
Decision Tree to Categorize Firms
35
Figure 3
The Three Stages of Evolution of Large-Molecule Drugs in the Biotechnology Revolution
0
2
4
6
8
10
12
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
new product class
only protein known
protein and connection to disease treatment known
Year of market launch in USA
Cou
nt o
f Pro
duct
s
36
APPENDIX 1
Table 11 Overall R&D Competence: Drug Discovery and Clinical Trial Management
Descriptive Statistics and Correlation Matrix Only Targeted Small-Molecule Drugs
(N = 353)
Count Mean Std.Dev. Min. Max. (1) Incumbent 115 (2) Diversifying 142 (3) rDNA Pioneer 31 (4) Diversifying, no rDNA Pioneer 112 (5) Cumulative 467 237 5 918 (6) Drug Novelty -1.68 1.16 -4.16 0 (7) R&D Alliance 2
(1) (2) (3) (4) (5) (6) (7)
(1) Incumbent 1 (2) Diversifying -0.57 1 (3) rDNA Pioneer -0.22 0.36 1 (4) Diversifying, no rDNA Pioneer -0.47 0.83 -0.21 1 (5) Cumulative -0.47 0.25 0.17 0.16 1 (6) Drug Novelty -0.06 -0.06 -0.00 -0.07 0.56 1 (7) R&D Alliance -0.05 -0.06 -0.02 -0.05 -0.09 -0.01 1
37
Table 12
Overall R&D Competence: Drug Discovery and Clinical Trial Management
Cox Model Analysis of Drug Approval (353 Spells, 7 Events)
Only Targeted Small Molecules All Coefficients in Hazard Rates
Model 1 Model 2 Model 3
Incumbent 3.74 (4.26)
4.31 (4.68)
4.37 (4.69)
Diversifying 3.60 (4.01)
3.00 (3.87)
rDNA Pioneer 0.00 (0.00)
Diversifying, no rDNA Pioneer 3.44 (4.38)
Cumulative Introduction 1.00 (0.00)
1.00 (0.00)
Drug Novelty 1.43 (0.47)
1.41 (0.48)
R&D Alliance 0.00 (0.00)
0.00 (0.00)
Log Likelihood -28.7 -27.7 -27.3 + p < 0.1, * p < .05, ** p < .01, *** p < .001 Standard errors in parentheses.