bridging the gap from inventions to innovations ...€¦ · bridge the – so far fairly understood...

40
Paper to be presented at DRUID15, Rome, June 15-17, 2015 (Coorganized with LUISS) Bridging the Gap from Inventions to Innovations - Increasing Success Rates in Clinical Trials Anne Assmus University of Passau Department of Economics and Business [email protected] Carolin Haeussler University of Passau Chair of Organization, Technology Management and Entreprene [email protected] Abstract Turning basic research discoveries into marketable products or technologies is a major challenge, especially in drug development. We conceptualize the necessary exploitation skills to go beyond the ??the use and development of things already known? (Levinthal & March, 1993, 108) by discussing how basic and applied research skills need to interact in order to successfully apply inventions to turn them into innovations. Using data on over 4500 clinical trials in the pharmaceutical industry, we find that the likelihood of success increases, when the investigators leading these trials have a skill set that in terms of quantity and quality balances basic and applied science. Importantly, the positive impact of such "bridging" relies on individuals bridging the basic-applied science divide. Inter-individual splitting of basic and applied skills in teams does not compensate for intra-individual bridging skills. Breadth of experience in terms of disease fields reduces success in exploitation and negatively moderates bridging skills. Jelcodes:O31,J24

Upload: others

Post on 18-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

Paper to be presented at

DRUID15, Rome, June 15-17, 2015

(Coorganized with LUISS)

Bridging the Gap from Inventions to Innovations - Increasing Success

Rates in Clinical TrialsAnne Assmus

University of PassauDepartment of Economics and Business

[email protected]

Carolin HaeusslerUniversity of Passau

Chair of Organization, Technology Management and [email protected]

AbstractTurning basic research discoveries into marketable products or technologies is a major challenge, especially in drugdevelopment. We conceptualize the necessary exploitation skills to go beyond the ??the use and development of thingsalready known? (Levinthal & March, 1993, 108) by discussing how basic and applied research skills need to interact inorder to successfully apply inventions to turn them into innovations. Using data on over 4500 clinical trials in thepharmaceutical industry, we find that the likelihood of success increases, when the investigators leading these trialshave a skill set that in terms of quantity and quality balances basic and applied science. Importantly, the positive impactof such "bridging" relies on individuals bridging the basic-applied science divide. Inter-individual splitting of basic andapplied skills in teams does not compensate for intra-individual bridging skills. Breadth of experience in terms of diseasefields reduces success in exploitation and negatively moderates bridging skills.

Jelcodes:O31,J24

Page 2: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

Bridging the Gap from Inventions to Innovations - Increasing Success Rates in Clinical Trials

ABSTRACT

Turning basic research discoveries into marketable products or technologies is a major

challenge, especially in drug development. We conceptualize the necessary exploitation skills

to go beyond the “‘the use and development of things already known” (Levinthal & March,

1993, 108) by discussing how basic and applied research skills need to interact in order to

successfully apply inventions to turn them into innovations. Using data on over 4500 clinical

trials in the pharmaceutical industry, we find that the likelihood of success increases, when

the investigators leading these trials have a skill set that in terms of quantity and quality bal-

ances basic and applied science. Importantly, the positive impact of such "bridging" relies on

individuals bridging the basic-applied science divide. Inter-individual splitting of basic and

applied skills in teams does not compensate for intra-individual bridging skills. Breadth of

experience in terms of disease fields reduces success in exploitation and negatively moderates

bridging skills.

Keywords: Bridging, Knowledge Recombination, Pharmaceutical Industry, Clinical Trials

JEL Classification: O31, O32, J24

Page 3: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

1. Introduction

"Translational gap" is a frequently used term to describe the challenge in turning basic scien-

tific discoveries into innovations. The extent to which scientific achievements have been ex-

ploited for practical applications lags far behind expectations, especially in the pharmaceutical

industry, which has spurred the interest of academia, industry and government alike in over-

coming the translational gap (Sung Ns and et al. 2003; Zerhouni 2007). Prior research has

focused on the translational gap between science discoveries and inventions (e.g, Jaffe and

Trajtenberg 1996; McMillan, Narin et al. 2000; Gittelman and Kogut 2003; Lim 2004) but

research on the gap to exploitation – development and testing stage – is lacking. This “transla-

tional gap” is particularly eminent in the pharmaceutical industry, where despite increased

efforts and investments into research and development (R&D) and major scientific advances,

the output of innovative drugs has been declining dramatically over the past years (Hu,

Schultz et al. 2007; Woolf 2008; Pammolli, Magazzini et al. 2011).

An important question is, whether a balanced skillset of individuals or teams in terms of

basic and applied science (“bridging skills”) improves success rates of exploitation projects.

There is an emerging debate in the sociological and economics of science literature on the

relationship between basic and applied science. Basic and applied science are, on one hand,

said to be in logical and normative conflict (Merton 1973; Gittelman and Kogut 2003); on the

other hand found to be complementary in certain instances (Stokes 1997) and applications

(Fleming and Sorenson 2004). Many inventions are derived from basic science discoveries

(McMillan, Narin et al. 2000), basic science skills increase the ability to absorb knowledge

from beyond firm boundaries (Cohen and Levinthal 1990) as well as support information

gathering and problem solving (e.g., (Zucker, Darby et al. 2001).

If basic science is indeed complementary to applied research, four important follow-up

questions emerge in terms of i) whether this applies also for exploitation, when turning inven-

tions into innovations, ii) whether the balance between basic science and applied science

Page 4: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

skills matter, iii) whether bridging skills interfere with breadth in experience in terms of ap-

plications and whether iv) basic and applied skills need to be intra-individually bundled or

can be split between scientists in a team.

First, we examine an area of development research that has received more limited scholar-

ly attention in comparison to research that analyzes the importance of basic science for explo-

ration output (i.e., invention). While moving from basic science to inventions is generally

seen as an explorative task of learning and creating new knowledge, moving from inventions

to marketable products represents exploitation of the already existing knowledge base (Gupta,

Smith et al. 2006; Liu 2006). Why is the "exploitative" setting interesting? The capabilities

needed for exploitative research are predicted to differ from explorative research (March

1991). Baum, Li, and Usher conceptualize exploitation to refer “to learning gained via local

search, experiential refinement, and selection and reuse of existing routines” (Baum, Li et al.

2000). Brenner and Tushman (2002, 679) highlight that “exploitative innovations involve

improvements in existing components and build on the existing technological trajectory,

whereas exploratory innovation involves as shift to a different technological trajectory”

(Benner and Tushman 2002). While the dominant view in prior literature has stressed that

experience in the relevant appliances is the dominant success factor in exploitation (Levinthal

and March 1993), increasingly voices have raised to reconsider the dominance of applied re-

search in favor of a broader base of insight (FitzGerald 2005). Particularly the term

“translational medicine” born by practitioners in the pharmaceutical industry point to the

importance of bridging skills in exploitation, though we lack empirical insights. More than in

exploration, attributing importance to and applying bridging skills might be particularly

challenging in exploitation, where appropriation by exclusive use and commercial rewards are

the guiding principles. On the other hand the “translational” aspect also illustrates that basic

research embedded in an instutional logic where recognition is the sole property right also is

an important input factor. We aim to address what enhances success under such circumstances

Page 5: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

by quantifying the relative importance of either as the input, meaning where the emphasis is

placed between the two institutional logics and their incentive systems.

Second, differentiating between having some knowledge in basic science versus having

profound knowledge or being a star basic scientist might matter. Obviously highly-

accomplished scientists are important, for example, in terms of improving the quality of new

(joiner) scientists (Agrawal, McHale et al. 2013), increasing citations when joining a depart-

ment (Azoulay, Zivin et al. 2011) or providing a quality signal to investors (Zucker and Darby

1998). However, the stars productivity in translating discoveries into inventions has been

questioned. Prior studies report that scientists engaged in highly cited academic research pro-

duce less valuable patents (Gittelman and Kogut 2003; Toole and Czarnitzki 2009) and show

a lower likelihood of patenting (Calderini, Franzoni et al. 2007). A strong orientation towards

academic science might limit its transferability or applicability in a commercial context. This

might well apply to exploiting activities where both institutional logics are present.

Third, we combine the concept of bridging the basic and applied research divide with the

level of breadth in experience on productivity in innovation research (Boh, Evaristo et al.

2014). Breadth could be seen as another form of (horizontally) bridging different fields,

which would point to a general capability of recombining knowledge domains, however, it

might overstrain the scientists’ capacity when combined with (vertical) bridging skills be-

tween basic and applied science.

Lastly, with more and more large teams are becoming the dominant form of organizing

R&D (Wuchty, Jones et al. 2007; Singh and Fleming 2010); the question arises of whether

bridging skills need to be “within” an individual capable of such activities or whether basic

and applied skills can be distributed among team members. The economics of science litera-

ture has highlighted that teams enable to assemble different pieces of knowledge from indi-

viduals (Jones 2009; Singh and Fleming 2010; Bercovitz and Feldman 2011; Jones and

Weinberg 2011) and allow division of labor between team members (Häussler and Sauermann

Page 6: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

2014). However, we conceptualize bridging skills to rely on single individuals who inhibit

both basic and applied research skills, which cannot be substituted by a split of basic and ap-

plied skills between team members.

We test our hypotheses by using a dataset of more than 4500 clinical trials, which have

been conducted between the year 2005 and 2012. During clinical trials, a patented invention is

tested in its final setting (“the human”) for the first time and in a stepwise process. Clinical

trials constitute an excellent setting for this kind of research on translational science, and rep-

resent an area of major economic relevance facing major hurdles in terms of costs, trial

lengths and high failure rates (DiMasi, Feldman et al. 2010; Pammolli, Magazzini et al.

2011). The term “translational medicine” highlights the difficult transition between the basic

(preclinical) and the clinical stages in the pharmaceutical R&D process. Academia, industry

and government jointly increased initiatives to overcome this gap. In our study, we relate the

skillset of the principal investigators (i.e., their publication activity in terms of basic and ap-

plied science as well as their experience) to the success (i.e., the successful completion of the

trial). While prior literature has mostly interpreted publication activity as basic science, we

exploit the heterogeneity of publications in terms of basic and applied science and thus can

investigate the concept of basic and applied being distinct or along a continuum.

Our findings highlight the importance of human capital bridging the basic and applied sci-

ence divide for the translation of inventions to innovations. For individuals, a balance skillset

is most optimal for taking advantage of the combination of capabilities and knowledge of ei-

ther science as well as to cope with the idiosyncrasies of the academic and commercial incen-

tives. Also within teams, such individuals confer superior performance. In general, excellence

in basic and applied science is not beneficial. Breadth of experience in terms of various dis-

eases lowers success, and negatively moderates bridging.

Our insights contribute to a growing literature on human capital and the production of in-

novation (Fleming and Sorenson 2003; Toole and Czarnitzki 2009). Extending prior literature

Page 7: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

arguing that the relationship between research and innovation is complex (Gittelman and

Kogut 2003), we add that individuals who are skilled in basic and applied science are able to

bridge the – so far fairly understood – gap from the invention to testing stage. Even in the

exploitation phase, where creativity and the linkage to basic science are proposed to be much

less prominent (Rothaermel and Deeds 2004), bridging scientists with an understanding of

both basic and applied science enhance success. However, in correspondence with (Gittelman

and Kogut 2003) who find that the ability to produce excellent science is detrimental to inno-

vative success, our findings indicate that the right dose of bridging is important as scientists

being too much drawn into the academia are sub-optimal. Hence, there is a problematic dis-

connect between excellence in basic science and in applied science which can only be over-

come by balancing the involvement in the two institutional spheres both in terms of quality

and quantity.

By disentangling intra- versus inter-individual bridging in teams, we also contribute to the

literature on team science (Wuchty, Jones et al. 2007). While prior research has shown that

knowledge in various domains can be distributed among team members (Jones 2009), bridg-

ing is only superior when it is inhibited within individuals and not split between two or more

specialized team members. This finding is novel as it speaks to the limits of specialization.

2. Conceptual Framework

Basic and Applied Research

Basic research has the goal of increasing the scientific knowledge base, which is generally not

appropriable, or "pursuing knowledge for its own sake" as often done in academia (Dasgupta

and David 1994), from which industry can source insights for innovations (Zucker, Darby et

al. 2001; Henderson, Jaffe et al. 2005). Basic research generates a public good, whose appro-

priation is difficult, which’s returns can only be expected in the long run and indirect ways,

and thus requires public investments. Still, public investments into academic and basic sci-

ence ultimately have the raison d'être that they are believed to lead different actors to take

Page 8: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

such discoveries on and apply them to generate a benefit for society (Gibbons and Johnston

1974; Pavitt 1991). In contrast, applied science is focusing on specific challenges, whose so-

lutions may represent a product, technology or service that can be commercialized and gener-

ate financial rewards.

There is a long-standing debate on the relationship between basic and applied science. One

strand of proponents argues in favor of a divide between basic and applied as polar phenome-

na that are diametrical points on a line or as two opposing ends of a continuum, (Bush 1945;

Brandl 1998). Bush (1945) states that basic and applied research follow different logics and

thus excellence in one will inhibit the other. The conflicting logic is attributed to different

epistemological communities in which scientists in basic science and commercial applications

are embedded (Dasgupta and David 1994). On the other hand, Stokes in his book titled "Pas-

teur's Quadrant" argues for a different view. He proposes the existence of a third kind of re-

search that obtains both the goal of valuable insights for application (applied science) as well

as increasing the knowledge base (basic science), besides the established purely basic and

purely applied research endeavors (Stokes 1997).

Innovation research also suggests that basic science drives much of the innovative perfor-

mance of firms, industries and nations (Rosenberg 1990; Pavitt 1991; Zucker and Darby

1995; Zucker, Darby et al. 1998). Various mechanisms have been outlined by which basic

science impacts industrial innovation. Many inventions are derived directly from basic sci-

ence discoveries, which is documented by patents citing scientific publications (McMillan,

Narin et al. 2000). For example in life-sciences, firms that publish and developed sophisticat-

ed links to academics achieve greater levels of success (Henderson and Cockburn 1994;

Owen-Smith and Powell 2004). Firms support basic research activities of their scientists in

order to stay well connected to the community and thereby understand and be able to absorb

knowledge from beyond the firm boundaries (Cohen and Levinthal 1990). Human capital well

skilled in basic research support information gathering and problem solving in firms (Zucker

Page 9: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

and Darby 1996; Zucker, Darby et al. 2001). Rosenberg ascribes basic science a mediating

role in the translation of discoveries (Rosenberg 1990). Basic science research capabilities are

essential for evaluating the outcome of much applied research and for perceiving its possible

implications and can provide valuable guidance to the directions in which there is a high

probability of payoffs to more applied research (Rosenberg 1990). Fleming and Sorenson

postulate that for coupled inventions, like in pharmaceuticals, basic science is necessary to

understand where a specific configuration of components may achieve the highest benefit or

which may be most useful and offer the highest potential (Fleming and Sorenson 2004). Fur-

ther, Cockburn and Henderson (1994, 1998) find that rewards of basic and applied research

act as complements to increase productivity of scientists when analyzing incentive systems

inside pharmaceutical firms.

While prior research has emphasized the importance of both skills in basic as well as in

applied science to turn discoveries into inventions, we examine the area of the exploitation

stage that has received more limited attention. Specifically in the setting of clinical trials, in-

vestigators with skills in basic, bench-side research, and applied research may be better at

understanding the potential, in designing experiments and trials to test hypotheses efficiently

and combine methodologies to create valuable outcomes. Being active in the basic science

paradigm should create incentives for the clinicians to report on such findings as they may be

publishable, even though clinicians generally are not incentivized to report or act on such

findings (DeMonaco, Ali et al. 2006). Thus, to successfully address the translational block

from basic science to improved health, "an interdisciplinary array of clinical investigators" is

essential (Sung Ns and et al. 2003). Hence, we:

H1: Bridging skills, i.e., skills in both basic and applied research, increase success in de-

velopment projects.

Excellence in basic and applied research

Page 10: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

Though, there are compelling reasons for the complementary view of basic and applied re-

search, scientists are still confronted with the conflicting tensions between the basic and the

applied research paradigms. Bridging both sciences may represent a challenge that needs a

very specific skill set. Being too much drawn in one of the two spheres is lowering the skills

to bridge exploitation.

Gittelman and Kogut (2003) find that patents of scientists with highly impact publications

receive fewer patent citations, but concomitantly find investigators with experience in both to

increase innovation outputs. Toole and Czarnitzki (2009) also find that there is a tradeoff be-

tween basic research productivity and a firm’s patent productivity. They differentiate scientif-

ically oriented human capital from commercial oriented human capital gained by firm bio-

scientists during their careers in academe before they joined or founded a firm. Their results

suggest that depth of scientifically oriented human capital is negatively related to a firm pa-

tent productivity but positively to the firm attracting an NIH grant after completion of proof-

of-concept studies.

Hence, we conceptualize bridging skills not only to have a skillset in basic as well as in

applied research but also to carefully balance researchers’ embeddedness in these two

spheres.

H2a: Bridging skills are most effective for the success of development projects if skills

in basic and applied are balanced. (Quantity)

Furthermore, the concept of stars has received increasingly attention, though with conflict-

ing empirical results. While some proponents claim that star scientists are much more prolific,

produce better innovations and have a large impact on the valuation of firms that they advise

or found (Zucker and Darby 1995; Darby, Liu et al. 1999; Zucker, Darby et al. 2001;

Agrawal, McHale et al. 2013), others have voiced concerns of overemphasizing the role of

starts. Rothaermel and Hess find that it is not the intellectual human capital of star scientists

that increases innovative output of a firm, but that of the non-star scientists and though aca-

Page 11: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

など

demic scientists may be high-performer in terms of publications and their impact, such scien-

tists produce less innovations (Calderini, Franzoni et al. 2007). Too strong a focus on the aca-

demic career and basic science may thus result in inferior translational capabilities due to a

different motivational or basic science-driven value focus. Bridging scientists that exhibit

excellence in both sciences, accordingly, may decrease the success of their product develop-

ment efforts as they are drawn too much towards the basic science paradigm.

H2b: Excellence in both applied and basic research tilts the balance towards too much

engagement in the academic science community and thus decreases success rates of

translational research projects. (Quality)

2.3 Breadth of Experience and Bridging

While we follow a vertical concept of bridging between basic and applied research, a hori-

zontal concept of bridging would view experience in different research areas as a capability to

linking fields. The latter is more a general capability of recombining knowledge but one,

which might interact with our (vertical) concept of bridging skills. Broad set of experience or

“breadth” is the diversity of experiences and insights that have been accumulated (Fleming,

Mingo et al. 2007). Some scholars argue that creativity requires a recombination of

knowledge, which is best achieved through individuals with broad insights into different

fields (Katila and Ahuja 2002). Too much focus may lead individuals to be trapped within

their thinking and their held beliefs about the matter at hand. Others claim that only specializ-

ing in one or few fields enables to contribute at the knowledge frontier (Sternberg and O'Hara

1999). Jones (2009) propagates the “death of the Renaissance man” as researchers are becom-

ing more and more experts in increasingly narrow knowledge domains as an answer to the

increasing “burden of knowledge”. Researchers require longer learning periods before making

contributions and therefore must specialize.

Further, a strong focus enables experts to search locally to determine what beliefs or rules

to break, whereas less depth of experience may lead to conformist behavior in terms of beliefs

Page 12: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なな

or rules (Weisberg 1999). Studying the company 3M, Boh et al. (2014) report that having in-

depth experience in one field leads to more valuable inventions, while breadth of experience

is associated with higher numbers of inventions, but less influential ones. Macher and Boerner

(2006) in their study of pharmaceutical Contract Research Organizations find that scope in

terms of different technological areas reduces the success rates of exploitation projects when

not paired with in-depth knowledge in a specific area. They recommend that firms should

prioritize development activities.

Breadth in terms of experience in different fields might be advantageous only if recombi-

nation matters and if these benefits exceed the costs of sharing attention to various areas in-

stead of concentrating on one. It is likely challenging to use the experience gained in diverse

fields to excel at exploitative research in a specific field. This might even be more severe in

projects with translational gaps. Because bridging the gap in a specific field is already highly

specific and drawing on a high level of tacit knowledge, it may even be constrained when

researchers are, additionally, active in various fields. Rosenberg reasons that much technolog-

ical knowledge in a field deals with "the specific and the particular” (Rosenberg 1978). Indi-

viduals who are expected to master the gap as well as mastering and sorting impressions and

experiences gained in very diverse applied fields might be overburdened. In translational are-

as, more vertical focused problem-solving strategies may be favored compared to horizontally

broad field knowledge. Bridging scientists are per definition knowledgeable in basic science

which is more general per nature than applied knowledge. Being also broad in terms of the

applied research/ development fields does not effectively and efficiently guide the investiga-

tors in their endeavor. The level of complexity is already too high in order to cope with shar-

ing attention with various fields and being able to make use of the benefit of bridging. The

phenomenon is seen at clinics and research hospitals: while investigators may establish their

own laboratories in which to research the molecular biology behind the disease that they try to

cure, such physicians have to have gained the accreditation for the specialization through

Page 13: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なに

years of training in the disease. Therefore,bridging skills are leveraged if the investigator ex-

perience is focused on one or few fields.

H3: Breadth of experience in applied research fields negatively moderates the effect

of bridging skills on success in exploitation projects.

2.4.Intra- versus inter-individual Bridging

When analyzing the impact of bridging the basic and applied science divide, the extent to

which bridging skills are intra-individually or can be inter-individually split between team

members is important. In general, an increasing amount of knowledge production is per-

formed by teams rather than individuals (Wuchty, Jones et al. 2007; Singh and Fleming

2010). Much of prior studies devote this to the notion that knowledge production involves the

recombination of different knowledge domains which can be achieved by division of labor

among researchers specializing in different fields and recombining their work in teams (Singh

& Fleming, 2010; (Singh and Fleming 2010; Bercovitz and Feldman 2011). On the other

hand, individuals have been shown to be more effective in integrating knowledge diversity

than teams (Taylor and Greve 2006). In contrast to this strand of research, our focus is on

bridging between scientific spheres – which often appears vertically. The translational capa-

bility is likely to have limits in being articulated and codified within teams but rather to be

“sticking” within certain individuals (DeMonaco, Ali et al. 2006). Polanyi (1958, p. 49) states

that "the aim of a skillful performance is achieved by the observance of a set of rules which

are not known as such to the person following them" (Polanyi 1958). Problem solving pro-

cesses and learning that lead to innovation result when different knowledge structures coexist

in the same mind, when at least one head can fit most of the relevant pieces of knowledge

together (Simon 1985). We thus expect individuals to be able to bridge the basic to applied

science paradigms more effectively than teams in which a part of the team members is fo-

cused on basic and the other part on applied research.

Page 14: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なぬ

H4: Intra-Individual bridging mediates the beneficial effect of bridging on the transla-

tion of inventions, even within teams, not the inter-individual bridging.

3. Clinical Research Setting

3.1. From Bench to Bedside

We test our hypotheses with a dataset on clinical trials. The development of new medical in-

terventions to improve health and lives of patients is a complex, costly and lengthy process

with high failure rates. In fact, this process is associated with substantial costs (DiMasi 1991;

DiMasi 2001; DiMasi, Hansen et al. 2003) and the majority of R&D funding of firms is allo-

cated to this process (NSB 2004). Empirical evidence suggests that most of industrial R&D is

exploitative in character with approximately 75% allocated to development activities, 20%

committed to applied research and only 5% targeting basic research (NSB 2004). Especially

in drug development, the attrition rates of novel compounds from discovery to market are

higher than in any other industry, exceeding 90% in the new chemical entity space (Kola and

Landis 2004) and 80% in the biologics space (DiMasi and Grabowski 2007); the time it takes

for a substance from discovery to market is and costs have risen over decades to over

USD800M per marketed drug and more (DiMasi, Hansen et al. 2003; DiMasi, Feldman et al.

2010; DiMasi 2014). Beside its importance in terms of economic impact, the clinical trial set-

ting enables us to analyze the exploitation of inventions or the transfer from an invention into

a product (March 1991; Nijstad and De Dreu 2002; Hoang and Rothaermel 2010). Once a

new substance has been “discovered” at the bench-side, a second set of development-oriented

activities involving bed-side activities begin with the clinical stages. The translation from

bench to bedside is particularly interesting and important with empirical ambiguity in terms of

how best to organize this translation. Early clinical trials are the first time that a treatment

drug is tested on human subjects and thus this form of research differs substantially from oth-

er forms of research. It is strictly regulated and requires approval by different authorities, be-

fore its initiation and is usually conducted at hospitals and medical centers. Developing a new

Page 15: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なね

drug or novel drug combinations to market is a well-defined stepwise process that evolved

over decades as the different stakeholders, among them pharmaceutical firms and regulators,

defined a process that would ensure safety, while maximizing output of novel therapeutics to

achieve commercial goals and benefits for patients.

Principal Investigators of Clinical Trials

Clinical research is conducted at research hospitals and led by principal investigators (PIs)

who are either employed by these hospitals, public funded institutions or research centers. A

principal investigator is responsible for ensuring that an investigation is conducted according

to the signed investigator statement, the investigational plan, and applicable regulations for

protecting the rights, safety, and welfare of subjects under the investigator's care as well as for

the control of drugs under investigation. The investigator has the close oversight over the “ac-

tual clinical research situation on the ground” and is, as part of the clinical management team,

responsible for the “scientific and ethical integrity of clinical trials” (Hoekman, Frenken et al.

2012), p.2). Recently, pharmaceutical firms are criticized for having their clinical trial per-

formed without any further input or insight from investigators, as if it were routine work

(Rasmussen 2005). But, as failures in clinical trials for pharmaceutical firms have been pub-

lished to derive from slow patient recruitment, lack of feasibility of the study protocol, too

high incidences of adverse events among other challenges (Buonansegna, Salomo et al. 2012;

Buonansegna, Salomo et al. 2014), investigators as the responsible doctors applying the drug

to patients and monitoring effects should be the closest to being able to take activities to pre-

vent failures. Investigators of clinical trials should be in a key position “to test biologic hy-

potheses in living patients” and “have the potential to change the standards of care”

(Davidoff, DeAngelis et al. 2001), p.786).

4. Dataset, Measures and Methodology

Dataset

Page 16: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なの

We use the ClinicalTrials.gov database, a comprehensive registry of clinical trials maintained

by the U.S. National Library of Medicine at the National Institutes of Health to obtain de-

tailed information on clinical studies conducted in the U.S. and over 180 other countries.

Mandatory registration of clinical trials has been extended since the establishment of the da-

tabase in the year 2000. Since 2002 sponsors or principal investigators have to register studies

addressing life-threatening diseases in addition to trials funded by the U.S. government before

the enrollment of the first subject. Today, each clinical trial of drugs or biological products

has to be registered if at least one trial site is located in the U.S., drug candidates are manufac-

tured in the U.S., or the trial is conducted under an FDA investigational new drug application.

The data is publicly available and has already been downloaded. Our final dataset covers over

4500 clinical trials conducted between 2005 and 2012 for which we have the names of inves-

tigators and with at least one investigator having published in a peer reviewed journal.1 We

merged information about the publication activity of the PIs on the trials based on a match

with last and first names from the database Scopus. Major work in data preparation has been

done on trial and publication categorization and matching of drug synonyms. Trial indications

have been categorized according the WHO ICD10 system. Each single publication of the PIs

has been classified according to the CHI system (see also (Della Malva, Leten et al. 2013)).

Measures

Dependent Variable

We follow Danzon et al. (2005) and measure success in a given phase as a binary variable

whether the drug, treatment or intervention is advanced to the next phase of testing in the

same indication or not (Danzon, Nicholson et al. 2005). A trial advances to the next phase of

testing if the data generated met the specifications in terms of scope or completeness, and if

the quantitative test results warrant the investment into the next phase of testing. Since drugs

1 Information on the full clinical trials.gov dataset and of our final sample is provided in the Appendix 1.

Page 17: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なは

are often termed in many different ways during the stages, and clinicaltrials.gov does not give

clear guidelines on how information should be provided, we first had to match drug syno-

nyms. To classify the indication of the trials, we made use of the International Classification

of Diseases (ICD) established by the WHO, which we also needed for computing independent

variables on the disease level. With the profound biology background of one co-author, we

were able to design a drug synonym matching and disease and disease fields classification

scheme, which we also advanced with the help of 5 biology and medical experts. Appendix 1

describes the classification and matching scheme as well as the extensive manual coding. This

procedure allows us to identify all trials of a specific drug and their specific disease and dis-

ease field. This information enables us to code trials which show a follow on test phase as

success. We are only able to code a trial as successful, if a follow-on trial is feasible. Hence, we ex-

cluded all phase 3 and phase 4 trials which are most often the final phases of drug testing.2

Independent Variables

Basic, applied and bridging PIs

Publications of investigators are categorized into basic, applied or bridging researchers ac-

cording to their publication record. Our ability to differentiate publications based on the re-

search focus of the journal(s) they publish, whether basic or applied, allows us to control for

different functionality of the underlying research result. While our category "basic research"

represents publications which tend to be "general" in nature (Pavitt 1991) and not appropria-

ble (Rosenberg 1990), publications within the "applied" journal category may be either of

applied research nature, specific to a certain setting, or even intended to position an innova-

tion in the market place (Polidoro and Theeke 2012).

We use the CHI Journal classification system to categorize the PI’s publications

2 The coding details are provided in the Appendix A2.

Page 18: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なば

(Della Malva, Leten et al. 2013). CHI categorizes journals into 5 levels ranging from 0

(very applied) to 4 (very basic). We code publications in journals from level 0 to 2 as applied

and from level 3 and 4 as basic. The cut off points were selected after we presented a sample

of 10 journals to 3 researchers and asked them to attribute them to the basic or applied catego-

ry. For example, Nature as a very strong molecular and mostly basic science journal is cate-

gorized to level 4, as well as Biological Chemistry, whereas the New England Journal of

Medicine belongs to category 2, which our experts denote as an applied research journal. Fur-

ther applied journals are for example Current Opinion in Cardiology or American Surgeon.

Citation rates are considerably higher for level 4 journal publications, with the other levels not

deviating much from each other, as seen in the Appendix 2. We use two approaches to code

the PIs’ scientific orientation, the first highlighting the distinct nature of basic and applied

science and the second following a continuum view.

First, we code three binary variables: the variable d_bridging is one, if the PIs have at least

one publication in the basic applied and in the applied journal category. The binary variable

d_basicpubs_only is 1 if the PIs have only basic research publications and respectively

d_appliedpubs_only is 1 for PIs with only applied research publications.

Second, as an alternative coding, we viewed the basic and applied orientation along a con-

tinuum. In doing so, we coded the binary variable d_balancebridging as being 1, if the PIs

had between 25% and 75% of publications in the basic science journals, 0 otherwise. Hence,

whereas a PI with 1 out of 10 publications would be viewed as bridging researcher in terms of

d_bridging being 1, the same PI would not be classified as “bridging” in terms of the continu-

um view (i.e., d_balancebridging is 0). In addition, we compute the fraction of basic science

publications (share_basic and its square share_basic_squ) to further analyze the quantitative

relationship between bridging and success.

Further, we take excellence in bridging into account and code the binary variable

d_topcitbridging as being 1 if PIs have publications among the top decile cited publications in

Page 19: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なぱ

basic and in an applied journal. We classified a PI or investigator team as in the top decile

according to the 5 year citation rate to their publications. In correspondence, we coded the

binary variable d_topcit_basic for PIs who are among the top 10% of cited PIs for their basic

publications.

Inter- vs intra-individual bridging

While the former defined variables relate to the publications of an individual PI, we also

take into account that trials list a team of PIs. For the hypothesis testing on whether bridging

between basic and applied research needs to be intra-individually versus inter-individually,

i.e., split between members of a team of PIs, we also coded team based variables. We code the

binary variable d_basicPIs_only as being 1 for teams with only PIs listed who publish in basic

scientific journals. Correspondingly, d_appliedPIs_only is coded 1 for teams with only ap-

plied investigators. Further binary variables denote whether they had at least one bridging

investigator on a trial (d_mixofPIsinbridging), whether the team only had bridging investiga-

tors (d_bridgingPIs_only) or whether bridging is achieved through the team by including

basic science investigators and applied science investigator on the team

(d_bridgingthroughteam) but without having a bridging PI.

Breadth of experience

We measure the level of breadth of experience by counting the number of different disease

fields (according to the ICD chapter, e.g., “cancer”) in which the investigator has been listed

in the 5 years prior to the focal field. The variable is denoted ln_fielddivexp. Given the skew-

ness of the experience measures, we use the logarithmic form (1+log).

Control Variables

Several variables are included which are known or expected to influence the success and du-

ration of clinical trials and will allow us to isolate the bridging effect on trial success.

We control for the experience of the investigators in the disease of the focal trial. The vari-

able ln_disexp is measured as the logarithmic form (1+log) of the accumulated number of

Page 20: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

なひ

trials all investigators conducted over a time span of five years prior to the trial in the given

disease field such as cancer (according to the ICD chapter).

We include the variable no_investigators to measure the number of PIs on a trial as well as

the accumulated number of publications in the 5 years prior to the trial of all investigators on

a trial (no_pubs).

We also control for institutional differences by including the binary variable d_industry for

the type of institution to differentiate from university or non-profit sponsors of the trials.

Since big pharma firms might act differently, we code the variable d_bigpharma being 1, if a

sponsor belongs to the top 25 pharmaceutical companies according to their sales in 2012. We

control for the experience of the sponsor in clinical trials by sponsor_trialsexp which depicts

the number of trials the sponsor is listed on the 5 years prior to the focal trial and by including

sponsor_disexp to control for the experience in a given disease 5 years prior. We also include

how many trials the drug has been tested in in the same phase priordrugtrials_samephase.

Our notion is that it is likely that if a sponsor has conducted already several trials in the same

phase, it is more experiences than if it is the first trial in a specific phase. As already outlined

above, trial success and duration might differ in regards to the phase of testing. Hence, we

include binary variables for the phase (d_phase_0, d_phase_1, d_phase2, d_phase_12,

d_phase23). The reference class is d_phase_1. As a control for size, the variable enrollment

counts the number of patients enrolled. We code the number of different sites which are re-

cruiting for a given trial location_count. We also control whether the site is one of the 10% of

most prominent sites in the overall trials database (top10site) and whether the trial is conduct-

ed in the USA versus non-USA. Furthermore, prior studies report that success rates differ sig-

nificantly between disease fields. We use the ICD 10 coding system of the WHO to code bi-

nary variables for the most prominent disease fields such as cancer, endocrinedisease, men-

taldisease, circulatorydisease, infectiousdisease. All other disease areas are coded as

otherdisease. Endocrinedisease is used as the reference group in our econometric models.

Page 21: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にど

Descriptive Statistics

Descriptive statistics of the dependent, independent and control variables are reported in

Table 1.3 We first show the descriptive statistics for all observations in Column 2 and then

separate for trials with a single PI (Column 3) and trials with a team of investigators (Column

4). 51% of our trials are led by a single PI. In the following, we briefly comment for the full

sample and for the sake of brevity only mention specifics for the sample split.

Our dependent variable success shows a mean of 43% for the full sample which corre-

sponds with prior findings reporting a success rate of 70% for phase 1 and 30% for phase 2

trials, whereby phase 2 trials account for the majority of trials in our sample (Kola and Landis

2004). The success rate of single led trials is with 45% slightly higher than that of teams led

trials being 40%.

52% of the trials in the overall sample are led by (single or a team of) investigators which

show bridging skills (d_bridging), i.e., have published in a basic and applied journal, com-

pared to 30% being led by pure basic scientists (d_basicpubs_only). The remaining 18% work

in the applied field. 27% are led by well-balanced engaged investigators, i.e., who publish

between 25 and 75% of their output in basic journals (d_balancedbridging). The average share

of basic research articles is 61% (share_basic). Though we have a large number of bridging

investigators, only 7% achieve excellence in both fields, i.e., they are among the 10% of most

cited investigators in terms of their publications in basic as well as applied journals

(d_top_citbridging). Interestingly, 11% of teams with 2 or more PI’s show a top bridging in-

vestigators compared to a share of 7% for trials led by one investigator. Top bridging scien-

tists might prefer to work in a team when they participate as investigator in a clinical trial

Among the sample of team led trials, the majority of teams consist of all members being

bridging investigators (57%, d_bridgingPIs_only), 39% have some bridging PIs among the

3 Correlation tables for all dependent, independent and control variables are provided in the Appendix.

Page 22: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にな

team members (d_mixofPIsinbridging), only 1% of teams consists of pure basic as well as

pure applied investigators (d_bridgingthroughteam), and the remaining 3% consist of only

pure basic or only pure applied investigators.

We include two types of measures for the experience of PIs: The average clinical trial has

been led by a single or team of investigators which had already conducted 0.69 trials in the

same disease in the previous five years (disexp). Not surprisingly, this number is highly

skewed and higher for team led trials. Similarly, our measure of breadth suggests that the av-

erage trial is led by PIs who have an experience in 1.22 different diseases – with this variable

ranging from 0 experience up to experience in 56 different disease fields driven by team trials.

The average trial lists 2.51 investigators, whereby the average team lead trial has 4.09

members. The average trial is conducted by PIs who have published 11.6 publications in peer-

reviewed journals in the 5 years prior to the trial.

36% of trials are sponsored by industry and 5% by a Big Pharma firm. The average spon-

sor has initiated 326 trials and 58.35 trials in the same disease field in the 5 years prior. The

average trial has enrolled 89 patients. The trials are conducted on average at 4.3 sites, which

is about the same as within the team and the individual-led trials and shows that it is rarely

possible to recruit sufficient patient number at one site. 60% of trials have at least one site in

the US, which is the most important market for innovative pharmaceuticals. 9% of trials take

place at a top site. Most trials are in test phase 2 (46%) followed by phase 1 (33%). On aver-

age the drug has been tested in the same phase (but mostly in different indication) in 15.6 tri-

als before. Most drugs are tested for several different indications (diseases) and if successful,

tested in many different specific disease settings and under different formulations. Interesting-

ly, the team led trials show a lower number of previous trials with the same drugs (13 times)

than individual led trials (18 times). Note that we did not include phase 3 or 4 since they usu-

ally do not have a follow on test phase. The largest share of tests is conducted in cancer

(43%), followed by infectious diseases (7.9%), endocrines (7.5%), mental (6.5%).

Page 23: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にに

*** Table 1 about here ***

Econometric Model: Instrumental Variable approach for Selection of investigators

Examining the causal implications of PI characteristics on the success of clinical trials re-

quires a methodology that takes potential selection of specific PIs on a trial into account.

There are two potential sources of selection: (1) the sponsor trying to select the best PIs for a

critical trial and (2) the PI trying to participate in the most interesting trials or the one with a

high probability of success. We consulted prior literature and engaged into in-depth discus-

sions with experts of clinical trials to further our understanding of potential selection. We did

this with the intention to comprehensively control for the types of influences in our models.

Azoulay (2004) suggests that the capabilities of the investigators are affecting success in

trials when hypothesis generation is important (Azoulay 2004). Such hypothesis generation

and testing is especially important in the early phases of the clinical research up to phase 2.

Trials in phase 3 and later focus on statistically proving the effectiveness of the treatment ver-

sus placebo or the then-current standard of care, based on its results in phase 2 trials. In our

study we focus on trials in phase 0 up to phase 2/3 and control for the specific phases. Huck-

man and Zinner (2005, p. 180) argue that the experience of the investigator is believed by

sponsors to be a “significant predictor of enrollment” of patients in a clinical trial. We control

for the number of enrolled patients (Huckman and Zinner 2008).

Our own interviewees expressed that trials are very disease specific. A PI or team of PIs is

preferred with experience in the same disease – but the pool of potential investigators with

disease specific experience strongly determines their possibility to select between investiga-

tors. From these we followed that our variable disexp depicting PIs experience in a specific

disease might be endogenous. In addition, we learned that selection possibility depends on the

pool of potential investigators. We build on these insights and employ an instrumental varia-

ble approach. We run 2 stage least square models and instrument disease experience by the

number of investigators which did run a trial in the same disease the five years prior to a focal

Page 24: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にぬ

trial. Since the increasing number of trials over the years impacts the pool of investigators

with disease specific experience, we normalize the number by dividing it through the number

of total trials. We believe the number of available investigators in a field to be random and

independent of the field and successes of therapeutics in the field. Career choices are made on

a long-term view, also in terms of specialization. Attractiveness of therapies and therapeutic

areas through scientific advancement or developments in therapeutics, though, cannot be pre-

dicted and can change a field of practice dramatically within a few years. The case of meta-

static melanoma is a prominent example, in which no therapeutics achieved better survival

over supportive care until 2011, when the first treatment from the field of immuno-oncology

was approved which showed not only improvement in overall survival, but also >10 year

complete responses, which could mean cure of the disease (Maio 2012). This revolutionary

development not only changed the face of metastatic melanoma, but also of the immuno-

oncology field and has made it much more attractive. Thus, we argue that the number of in-

vestigator in a certain disease area to be independent of the attractiveness of a field and as-

sume that the possibility of selecting a highly experienced investigator is correlated with the

number of investigators that are working in the disease. Previous research and our interviews

did not reveal that PIs with bridging skills are selected consciously so far.

5. Results

5.1. Bridging and Breadth of Experience and the Success

Table 2 reports our results on how bridging skills impact the success of trials for the full

sample of 4,816 trials. We run each model first as a two-stage least square regression with

experience being instrumented and second as an ordinary probit model for which we report

the marginal effects.

We will briefly comment on the appropriateness of the instrumental variable approach be-

fore we turn to the interpretation of the results. The endogeneity test indicates that ln_disexp

is indeed endogenous. The first stage regression of the two-stage least square suggest that the

Page 25: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にね

pool of PIs is a strong predictor of selecting an investigator with prior disease experience (1%

significance). For example, the F-statistic of 144 in Model 2a strongly suggests that our IV is

not weak (Stock-Yogo and Cragg-Donald Wald test).

Model 1a and 1b include the control variables. Our measure for experience in disease

(ln_disexp) is strongly significant in both models. The probit result of Model 1b suggests that

an increase in experience by 1 % increases success by nearly 7%. In the following models we

use various approaches to test our hypotheses on bridging skills of investigators. In Model 2a

and b we add the dummy variables d_bridging and d_basicpubs_only. Results suggest that

trials with bridging investigators are more successfully compared to the reference group

which consists of teams with investigators having solely applied skills. Hence, our results

support our hypothesis 1 and show that bridging is not only relevant for turning discoveries

into inventions (Gittelman and Kogut 2003)but also for turning inventions into innovations. It

is not that the exploitation stage solely requires the use of existing applied knowledge, pro-

cesses and methods (Baum, Li et al. 2000), instead coupling this with knowledge in basic

supports the successful translation.

Models 3 and 4 measure the bridging effect in a quantitative way and thus build on a con-

tinuum view of basic and applied science – the models suggest that a well- balanced skillset is

important. In Models 3a and b, we refine our bridging measure of Models 2a and b to count

only those observations as bridging, where the share of basic publications is between 25 and

75% of all publications. The significant coefficients of d_balancedbridging suggest that the

investigators which have a share of basic publications between 25% and 75% are particularly

successful. Correspondingly, Models 4a and b suggest that the share of basic publications

shows an inverted U-shape relationship with success. Basic research skills are important but

with diminishing return. Hence, there is an optimum in the share of basic science. Insofar our

results build on Toole and Czarnitzki’s (2009) findings that commercial productivity decreas-

Page 26: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にの

es with depth of scientifically oriented human capital but up to a certain level it spurs the suc-

cess rate (Toole and Czarnitzki 2009).

We add also a qualitative aspect to the analysis in Models 5a and b by including a measure

for excellence, which is measured as a dummy variable for the investigators that are the top

10% of cited scientists over a 5 year period in either their basic or applied science, or both.

Models 5a and b provide support for our hypothesis 2b that being excellent in both realms is

too much and has a negative effect on the success rate of a trial. Being among the most cited

individuals in both academic and commercially oriented research may divert the focus or the

investigator from the trial, or may place focus on other aspects that will increase his or her

competitiveness in both worlds of science. This is in line with the findings of Toole and

Czarnitzki (2009) that productivity stems not from the stars, but from the non-star scientists as

academic human capital, at least to the extent that we consider these to be the stars within the

basic research paradigm. Interestingly, being top in applied is positively related to success but

being top in basic is negatively related. Excellence in applied research by itself, though, will

actually increase success. The negative effect of excellence for bridging scientists on success

may result from a kind of “pull effect” towards academe that tilts the balance shown to be

important in previous models, and not on excellence per se, which would need to be further

investigated

*** Table 2 about here ***

Table 3 investigates to what extent breadth of experience in disease fields of the investiga-

tors impacts success. We only display the 2sls-models with experience in diseases instru-

mented since the results are quite robust to the probit model. Model 1 includes our measure

for breadth in field experience with the various forms of bridging variables along with the

interaction with breadth are included in Models 2 to 4. The negative coefficients of

ln_fielddivexp suggest that the broader investigators’ experience in different fields lowers the

success rate of trials. This correspondents with the findings of Macher and Boerner (2006)

Page 27: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

には

reporting that companies face diseconomies of scope when working in different therapeutic

areas (Macher and Boerner 2006).

When we add the bridging variables along with the interactions, we find strong support for

H4 claiming that breadth of experience negatively moderates the impact of bridging on suc-

cess. Models 2 to 4 with breadth in terms of field experience reveal that H4 holds for all types

of interactions with bridging and breadth of experience in fields. Hence, if breadth of experi-

ence is coupled with bridging we find a negative impact on success. Presumably, the level of

complexity in coping with sharing attention horizontally in terms of different fields as well as

vertically in terms of basic and applied overstrains the capacity of investigators.

*** Table 3 about here ***

5.2. Trials with one investigator versus trials with a team of investigators

In the following, we split our sample into trials with just one investigator versus trials with a

team of investigators. Our objective is to provide insights whether and how the mechanism of

bridging may differ between individual investigators and teams of investigators.

Trials with one PI and Bridging

In the Appendix 3, we show models for the sample with one investigator. We restrict the table

to report the 2sls models with the instrumental variable approach for disease experience

(ln_disexp) but for reader’s interest re-run Model 1 as a probit. The results with the probit

model are quite similar to the 2sls, though sometimes shows larger significance of coeffi-

cients. Comparing the 2sls-models of the reduced sample in Appendix 3 to the 2sls-models

with the full sample in Table 2, we find that the results are quite robust.

When we include our measure for breadth of experience in field along with the interac-

tions, we do not find any significant result neither for ln_fielddivexp nor for the interactions.

Hence, breadth of experience seems not to matter for trial success. In terms of brevity we only

report but do not show a table with results.

Teams of Investigators and Bridging

Page 28: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にば

Teams of investigators are more likely to horizontally and vertically tap into different areas by

combining specialists in either area within one team. The importance of bridging the basic

and applied science divide may thus derive from very different configurations of individuals

within teams.

Figure 1 depicts the share of basic, applied and bridging investigators across different sizes

of investigator teams. It illustrates that the share of bridging investigators decreases while the

share of applied investigators, i.e., investigators who only publish in applied journals, increas-

es. The larger the team, the higher the fraction of specialized members, which could be a sign

that the diversity of knowledge required for the project requires it to be combined through the

members of the team as indicated by (Jones 2009).

Figure 1 Share of basic, applied and bridging investigators and teamsize

Table 4 reports the econometric models including only the observations with two or more

investigators on the trial. We briefly comment on our “usual” bridging variables but empha-

size Model 5 which splits the bridging measures into inter- and intra-individual bridging.

Models 1 to 4 suggest that the bridging concepts are too a lower extent present in teams.

While all coefficients of the bridging variables go in the same direction as in the model with

the full sample (and as hypothesized), the coefficients for d_bridging in Model 1a and b and

for testing an inverted U-shaped relation for share_basic and the coefficient of top citing in-

vestigators in basic and applied do not show significance. Corresponding to former models,

Page 29: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にぱ

balanced bridging shows a positive and significant impact on success. Comparing the insig-

nificance of Model 1b with the significance of Model 2 suggests that bridging through a min-

imal engagement in both basic and applied research does not influence success of the trial.

The partly insignificance of these measures may be due to the fact that for teams they capture

too many diverse bridging constellations, such as the combination of a basic and an applied

investigator within the team, teams with only bridging investigators or a combination of all

before mentioned.

Our main attention is devoted to Model 5, which splits all team trials into ones with only

basic investigators on the trials (d_basic_only), with only bridging scientists

(d_bridgingPIs_only), teams with a mix of bridging and other PIs (d_mixofPIsinbridging) and

the ones which only achieve bridging by combining PIs focused on either basic or applied

research (d_bridgingthroughteam). The reference category is teams with only applied investi-

gators. Our results suggest that teams with solely bridging scientist or teams with a mix of

bridging and other scientists are more successful than teams with only applied researchers.

This implies that only intra-individual bridging matters. Bridging skills need to be united in

one individual. The bridging concept cannot be achieved by combining pure applied with pure

basic scientists. This is a novel finding that corresponds with the notion of "generalists" who

are found to be driving highly innovative output in teams (Melero and Palomeras 2014). It

suggests that knowledge and skills are best combined, when present within one person

(Simon 1985).

However, teams might have advantages in assembling diverse field experience compared

to single PIs. Models 6 to 10 include breadth of field experience as well as its interactions

with our bridging variables. Breadth of field experience shows a negative significant effect on

success trial suggesting that though teams might be able to acquire diverse field knowledge it

is not paying off. Furthermore Models 8 to 10 suggest that breadth of experience even nega-

Page 30: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

にひ

tively moderates the bridging variables. Only Model 10 which splits teams into intra- and

inter-individual bridging does not show any significant interaction effects.

6. Conclusion

We argue in this paper that individuals who possess a balanced set of skills between basic

and applied science can bridge the translational gap between inventions and innovations and

improve success rates in product development, by themselves or as members of a team. This

supports the notion that basic science and applied science are complementary. Our results

from the analysis of the influence of investigator characteristics on clinical trials show that the

value of basic science insight and skills reaches not only into applied research stages but even

into product development. Basic science is not only a source for discoveries that may become

commercial products, its insights, methodologies and problem solving skills of basic science

are important tools even in such "late" activities as final testing and development of the prod-

uct (Pavitt 1991; Zucker, Darby et al. 1998; Zucker, Darby et al. 2001). The translational step

performed in drug development is marked by two different institutional logics and incentive

systems: It incorporates the basic science that enables an understanding of the underlying

functionality of the product or technology, but requires the commercial mindset to optimally

exploit its potential in the market. Investigators exhibit superior performance, if their scien-

tific career and engagement has been marked by both institutional paradigms. Working in the

development stage of the value chain, a main focus is the exploitation of existing knowledge

and the commercialization of the product under testing. Still, the methodologies and problem

solving skills that are gained through insights into the exploration of basic science (Rosenberg

1990) enable both a faster and more successful interrogation of a potential product and its

commercial opportunity.

Essential to the success of bridging is the relative intensity with which both basic and ap-

plied science is conducted by the investigator(s) and serves as input, and where the emphasis

Page 31: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬど

is placed between the two systems. Our findings suggest that both activities are complemen-

tary up to a certain degree. Excellence in both will actually decrease success. The values and

reward systems of the highly cited bridging scientists may still lie too much with the scientific

community so as to increase their successes at translating research into product development

(Gittelman and Kogut 2003). As is the case for the balance between exploration and exploita-

tion of (Gupta, Smith et al. 2006), excellence in both basic and applied science causes a trade-

off that inhibits success in exploitation projects, potentially as excellence in either is mutually

exclusive.

The interactive effect of bridging and breadth of experience suggests both benefits and

costs to the increasing diversity of activities. Penalties exist for who attempts to master both

the vertical translational abilities and attempts broadly apply their expertise across fields, sup-

porting findings by Macher and Boehrner (2006). The more general ability of bridging di-

verse fields of exploitation, horizontal bridging, showed negative effects in its interaction

with vertical, or basic to applied bridging. The costs of sharing attention to various areas in-

stead of concentrating on one are higher than the benefit of recombining knowledge from di-

verse fields.

Furthermore, we find intra-individual bridging to be of importance, not inter-individual

bridging within teams. While a lot of research has found specialization (Jones 2009), team

research (Wuchty, Jones et al. 2007) and team size (Milojević 2014) to increase, we find the

generalists in terms of basic and applied science within the teams to moderate the positive

impact of bridging. This finding corresponds with prior research that generalists within teams

are essential for the impact of innovations (Melero and Palomeras 2013; Melero and

Palomeras 2014). Our finding that intra-individual bridging of the basic/applied science

spheres enhances success of turning invention into innovations confirms other findings that

individuals are better at integrating diverse knowledge than teams of specialized members

(Gupta, Smith et al. 2006; Taylor and Greve 2006). Also for teams, we find that breadth of

Page 32: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬな

experience negatively moderates bridging. This supports the notion that, though teams should

be able to combine more diverse knowledge, they may face challenges in communication and

coordination, and thus do not perform better at knowledge recombination than individuals

(Williams and O'Reilly 1998).

For the industry, this work adds insights that may aid to alleviate the economic burden of

product development by increasing chances of success through correct selection of key indi-

viduals to lead projects and team members within such projects. It adds supporting evidence

to a broader education of clinical trial investigators in terms of research methodologies

(FitzGerald 2005).Furthermore, the advent of outsourcing of clinical trials to so called clinical

research organizations, which causes an even stronger divide between the scientists and la-

boratories that originally discovered and patented the product and the clinical researchers, that

may not only inhibit learning by the originator company (Smed and Getz 2013), but also de-

crease the likelihood of conducting a successful trial.

Page 33: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬに

References Agrawal┸A┻┸J┻Mc(ale┸etal┻ゅにどなぬょ┻Collaboration┸Stars┸andtheChangingOrganizationofScience┺EvidencefromEvolutionaryBiology┻TheChangingFrontier┺RethinkingScienceand)nnovationPolicy┸UniversityofChicagoPress┻Agrawal┸A┻┸J┻Mc(ale┸etal┻ゅにどなぬょ┻gWhystarsmatter┻gUniversityofTorontomimeo┻Azoulay┸P┻ゅにどどねょ┻gCapturingknowledgewithinandacrossfirmboundaries┺evidencefromclinicaldevelopment┻gTheAmericanEconomicReview94ゅのょ┺なのひな┽なはなに┻Azoulay┸P┻┸J┻S┻G┻Zivin┸etal┻ゅにどななょ┻TheDiffusionofScientificKnowledgeAcrossTimeandSpace┺EvidencefromProfessionalTransitionsfortheSuperstarsofMedicine┸NationalBureauofEconomicResearch┻Baum┸J┻A┻┸S┻X┻Li┸etal┻ゅにどどどょ┻gMakingthenextmove┺(owexperientialandvicariouslearningshapethelocationsofchainsfacquisitions┻gAdministrativeScienceQuarterly45ゅねょ┺ばはは┽ぱどな┻Benner┸M┻J┻andM┻Tushmanゅにどどにょ┻gProcessmanagementandtechnologicalinnovation┺Alongitudinalstudyofthephotographyandpaintindustries┻gAdministrativeScienceQuarterly47ゅねょ┺はばは┽ばどば┻Bercovitz┸J┻andM┻Feldmanゅにどななょ┻gThemechanismsofcollaborationininventiveteams┺Composition┸socialnetworks┸andgeography┻gResearchPolicy40ゅなょ┺ぱな┽ひぬ┻Boh┸W┻F┻┸R┻Evaristo┸etal┻ゅにどなねょ┻gBalancingbreadthanddepthofexpertiseforinnovation┺AぬMstory┻gResearchPolicy43ゅにょ┺ぬねひ┽ぬはは┻Brandl┸J┻E┻ゅなひひぱょ┻PasteurfsQuadrant┺Basicscienceandtechnologicalinnovation┸WileyOnlineLibrary┻Buonansegna┸E┻┸S┻Salomo┸etal┻ゅにどなにょ┻Towardsaframeworkofsuccessfactorsforclinicaltrials┻なひth)nternationalProductdevelopmentManagementConference┻Buonansegna┸E┻┸S┻Salomo┸etal┻ゅにどなねょ┻gPharmaceuticalnewproductdevelopment┺whydoclinicaltrialsfail╂gR┃DManagement44ゅにょ┺なぱひ┽にどに┻Bush┸V┻ゅなひねのょ┻gScience┺Theendlessfrontier┻gTransactionsoftheKansasAcademyofScienceゅなひどぬょ┺にぬな┽にはね┻Calderini┸M┻┸C┻Franzoni┸etal┻ゅにどどばょ┻g)fstarscientistsdonotpatent┺Theeffectofproductivity┸basicnessandimpactonthedecisiontopatentintheacademicworld┻gResearchPolicy36ゅぬょ┺ぬどぬ┽ぬなひ┻Cohen┸W┻M┻andD┻A┻Levinthalゅなひひどょ┻gAbsorptivecapacity┺anewperspectiveonlearningandinnovation┻gAdministrativesciencequarterly┺なにぱ┽なのに┻Danzon┸P┻M┻┸S┻Nicholson┸etal┻ゅにどどのょ┻gProductivityinpharmaceutical┽biotechnologyR┃D┻Theroleofexperienceandalliances┻gJournalofhealtheconomics24ゅにょ┺ぬなば┽ぬぬひ┻Darby┸M┻R┻┸Q┻Liu┸etal┻ゅなひひひょ┻Stakesandstars┻Theeffectsofintellectualhumancapitalonthelevelandvariabilityofhigh┽techfirmsfmarketvalues┻Cambridge┸Mass┻┺NationalBureauofEconomicResearch┻Dasgupta┸P.andP┻A┻Davidゅなひひねょ┻gTowardaneweconomicsofscience┻gResearchpolicy23ゅのょ┺ねぱば┽のにな┻Davidoff┸F┻┸C┻D┻DeAngelis┸etal┻ゅにどどなょ┻gSponsorship┸authorship┸andaccountability┻gJAMA┺thejournaloftheAmericanMedicalAssociation286ゅなどょ┺なにぬに┻DellaMalva┸A┻┸B┻Leten┸etal┻ゅにどなぬょ┻gBasicScienceasaPrescriptionforTechnologicalBreakthroughsinthePharmaceutical)ndustry┻gDeMonaco┸(┻J┻┸A┻Ali┸etal┻ゅにどどはょ┻gTheMajorRoleofCliniciansintheDiscoveryofOff̺LabelDrugTherapies┻gPharmacotherapy┺TheJournalof(umanPharmacologyandDrugTherapy26ゅぬょ┺ぬにぬ┽ぬぬに┻

Page 34: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬぬ

DiMasi┸J┻ゅにどどなょ┻gRisksinnewdrugdevelopment┺Approvalsuccessratesforinvestigationaldrugs┻gClinicalPharmacology┃Therapeutics69ゅのょ┺にひば┽ぬどば┻DiMasi┸J┻A┻ゅなひひなょ┻gCostofinnovationinthepharmaceuticalindustry┻gJournalofhealtheconomics10ゅにょ┺などば┽なねに┻DiMasi┸J┻A┻ゅにどなねょ┻gPharmaceuticalR┃DPerformancebyFirmSize┺ApprovalSuccessRatesandEconomicReturns┻gAmericanjournaloftherapeutics21ゅなょ┺には┽ぬね┻DiMasi┸J┻A┻┸L┻Feldman┸etal┻ゅにどなどょ┻gTrendsinrisksassociatedwithnewdrugdevelopment┺successratesforinvestigationaldrugs┻gClinicalPharmacology┃Therapeutics87ゅぬょ┺にばに┽にばば┻DiMasi┸J┻A┻and(┻G┻Grabowskiゅにどどばょ┻gThecostofbiopharmaceuticalR┃D┻)sbiotechdiffent╂gManagerialanddecisioneconomics28ゅね【のょ┺ねはひ┽ねばひ┻DiMasi┸J┻A┻┸R┻W┻(ansen┸etal┻ゅにどどぬょ┻gThepriceofinnovation┻Newestimatesofdrugdevelopmentcosts┻gJournalofhealtheconomics22ゅにょ┺なのな┽なぱの┻FitzGerald┸G┻A┻ゅにどどのょ┻gAnticipatingchangeindrugdevelopment┺theemergingeraoftranslationalmedicineandtherapeutics┻gNatureReviewsDrugDiscovery4ゅなどょ┺ぱなの┽ぱなぱ┻Fleming┸L┻┸S┻Mingo┸etal┻ゅにどどばょ┻gCollaborativebrokerage┸generativecreativity┸andcreativesuccess┻gAdministrativeScienceQuarterly52ゅぬょ┺ねねぬ┽ねばの┻Fleming┸L┻andO┻Sorensonゅにどどぬょ┻gNavigatingthetechnologylandscapeofinnovation┻gM)TSloanManagementReview┻Fleming┸L┻andO┻Sorensonゅにどどねょ┻gScienceasamapintechnologicalsearch┻gStrategicManagementJournal25ゅぱひょ┺ひどひ┽ひにぱ┻Gibbons┸M┻andR┻Johnstonゅなひばねょ┻gTherolesofscienceintechnologicalinnovation┻gResearchPolicy3ゅぬょ┺ににど┽にねに┻Gittelman┸M┻andB┻Kogutゅにどどぬょ┻gDoesgoodscienceleadtovaluableknowledge╂Biotechnologyfirmsandtheevolutionarylogicofcitationpatterns┻gManagementScience49ゅねょ┺ぬはは┽ぬぱに┻Gupta┸A┻K┻┸K┻G┻Smith┸etal┻ゅにどどはょ┻gTheinterplaybetweenexplorationandexploitation┻gAcademyofManagementJournal49ゅねょ┺はひぬ┽ばどは┻(äussler┸C┻and(┻Sauermannゅにどなねょ┻gTheAnatomyofTeams┺DivisionofLaborandAllocationofCreditinCollaborativeKnowledgeProduction┻gAvailableatSSRN┻(enderson┸R┻and)┻Cockburnゅなひひねょ┻gMeasuringcompetence╂Exploringfirmeffectsinpharmaceuticalresearch┻gStrategicmanagementjournal15ゅSなょ┺はぬ┽ぱね┻(enderson┸R┻┸A┻Jaffe┸etal┻ゅにどどのょ┻gPatentcitationsandthegeographyofknowledgespillovers┻Areassessment┺comment┻gTheAmericaneconomicreview95ゅなょ┺ねはな┽ねはね┻(oang┸(┻andF┻T┻Rothaermelゅにどなどょ┻gLeveraginginternalandexternalexperience┺exploration┸exploitation┸andR┃Dprojectperformance┻gStrategicManagementJournal31ゅばょ┺ばぬね┽ばのぱ┻(oekman┸J┻┸K┻Frenken┸etal┻ゅにどなにょ┻gThegeographicaldistributionofleadershipinglobalizedclinicaltrials┻gPloSone7ゅなどょ┺eねのひぱね┻(u┸M┻┸K┻Schultz┸etal┻ゅにどどばょ┻gTheinnovationgapinpharmaceuticaldrugdiscovery┃newmodelsforR┃Dsuccess┻g(uckman┸R┻S┻andD┻E┻Zinnerゅにどどぱょ┻gDoesfocusimproveoperationalperformance╂Lessonsfromthemanagementofclinicaltrials┻gStrategicManagementJournal29ゅにょ┺なばぬ┽なひぬ┻Jones┸B┻F┻ゅにどどひょ┻gTheburdenofknowledgeandthe╉deathoftheRenaissanceman╊┺isinnovationgettingharder╂gTheReviewofEconomicStudies76ゅなょ┺にぱぬ┽ぬなば┻Jones┸B┻F┻andB┻A┻Weinbergゅにどななょ┻gAgedynamicsinscientificcreativity┻gProceedingsoftheNationalAcademyofSciences108ゅねばょ┺なぱひなど┽なぱひなね┻

Page 35: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬね

Katila┸R┻andG┻Ahujaゅにどどにょ┻gSomethingold┸somethingnew┺Alongitudinalstudyofsearchbehaviorandnewproductintroduction┻gAcademyofmanagementjournal45ゅはょ┺ななぱぬ┽ななひね┻Kola┸)┻andJ┻Landisゅにどどねょ┻gCanthepharmaceuticalindustryreduceattritionrates╂gNatRevDrugDiscov3ゅぱょ┺ばなな┽ばなの┻Levinthal┸D┻A┻andJ┻G┻Marchゅなひひぬょ┻gThemyopiaoflearning┻gStrategicmanagementjournal14ゅSにょ┺ひの┽ななに┻Liu┸W┻ゅにどどはょ┻gKnowledgeexploitation┸knowledgeexploration┸andcompetencytrap┻gKnowledgeandProcessManagement13ゅぬょ┺なねね┽なはな┻Macher┸J┻T┻andC┻S┻Boernerゅにどどはょ┻gExperienceandscaleandscopeeconomies┺trade┽offsandperformanceindevelopment┻gStrategicManagementJournal27ゅひょ┺ぱねの┽ぱはの┻Maio┸M┻ゅにどなにょ┻gMelanomaasamodeltumourforimmuno┽oncology┻gAnnalsofoncology23ゅsupplぱょ┺viiiなど┽viiiなね┻March┸J┻G┻ゅなひひなょ┻gExplorationandexploitationinorganizationallearning┻gOrganizationscience2ゅなょ┺ばな┽ぱば┻McMillan┸G┻S┻┸F┻Narin┸etal┻ゅにどどどょ┻gAnanalysisofthecriticalroleofpublicscienceininnovation┺thecaseofbiotechnology┻gResearchpolicy29ゅなょ┺な┽ぱ┻Melero┸E┻andN┻Palomerasゅにどなぬょ┻TheRenaissanceoftheRenaissanceMan╂TheRoleofBroad)ndividualKnowledgeinTeamsof)nventors┻ぬのthDRU)DCelebrationConference┸Barcelona┸Spain┻Melero┸E┻andN┻Palomerasゅにどなねょ┻gTheRenaissanceManisnotdead┿Theroleofgeneralistsinteamsofinventors┻gResearchPolicy┻Merton┸R┻K┻ゅなひばぬょ┻Thesociologyofscience┺Theoreticalandempiricalinvestigations┸UniversityofChicagopress┻Milojević┸S┻ゅにどなねょ┻gPrinciplesofscientificresearchteamformationandevolution┻gProceedingsoftheNationalAcademyofSciences111ゅななょ┺ぬひぱね┽ぬひぱひ┻Nijstad┸B┻A┻andC┻K┻DeDreuゅにどどにょ┻gCreativityandgroupinnovation┻gAppliedPsychology51ゅぬょ┺ねどど┽ねどは┻NSB┸N┻S┻B┻ゅにどどねょ┻gScienceandEngineering)ndicatorsにどどぬ┻gU┻S┻GovernmentPrintingOffice┺Washington┸D┻C┻┻Owen┽Smith┸J┻andW┻W┻Powellゅにどどねょ┻gKnowledgenetworksaschannelsandconduits┺TheeffectsofspilloversintheBostonbiotechnologycommunity┻gOrganizationscience15ゅなょ┺の┽にな┻Pammolli┸F┻┸L┻Magazzini┸etal┻ゅにどななょ┻gTheproductivitycrisisinpharmaceuticalR┃D┻gNaturereviewsDrugdiscovery10ゅはょ┺ねにぱ┽ねぬぱ┻Pavitt┸K┻ゅなひひなょ┻gWhatmakesbasicresearcheconomicallyuseful╂gResearchPolicy20ゅにょ┺などひ┽ななひ┻Polanyi┸M┻ゅなひのぱょ┻PersonalKnowledge┻Towardsapost┽criticalphilosophy┻┻Chicago┸TheUniversityofChicagoPress┻Polidoro┸F┻andM┻Theekeゅにどなにょ┻gGettingCompetitionDowntoaScience┺TheEffectsofTechnologicalCompetitiononFirmsfScientificPublications┻gOrganizationScience23ゅねょ┺ななぬの┽ななのぬ┻Rasmussen┸N┻ゅにどどのょ┻gThedrugindustryandclinicalresearchininterwarAmerica┺threetypesofphysiciancollaborator┻gBulletinofthe(istoryofMedicine79ゅなょ┺のど┽ぱど┻Rosenberg┸N┻ゅなひばぱょ┻gThediffusionoftechnology┺Aneconomichistorian╆sview┻gTheDiffusionof)nnovations┺AnAssessment┸Centerfor)nterdisciplinaryStudyofScienceandTechnology┸Evanstonゅ)ll┻ょ┺NorthwesternUniversity┻

Page 36: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬの

Rosenberg┸N┻ゅなひひどょ┻gWhydofirmsdobasicresearchゅwiththeirownmoneyょ╂gResearchpolicy19ゅにょ┺なはの┽なばね┻Rothaermel┸F┻T┻andD┻L┻Deedsゅにどどねょ┻gExplorationandexploitationalliancesinbiotechnology┺Asystemofnewproductdevelopment┻gStrategicmanagementjournal25ゅぬょ┺にどな┽ににな┻Simon┸(┻A┻ゅなひぱのょ┻gWhatweknowaboutthecreativeprocess┻gFrontiersincreativeandinnovativemanagement4┺ぬ┽にに┻Singh┸J┻andL┻Flemingゅにどなどょ┻gLoneinventorsassourcesofbreakthroughs┺Mythorreality╂gManagementScience56ゅなょ┺ねな┽のは┻Smed┸M┻andK┻A┻Getzゅにどなぬょ┻gUnfulfilledtranslationopportunitiesinindustrysponsoredclinicaltrials┻gContemporaryclinicaltrials35ゅなょ┺ぱど┽ぱは┻Sternberg┸R┻J┻andL┻A┻Of(araゅなひひひょ┻gなぬCreativityand)ntelligence┻g(andbookofcreativity┺にのな┻Stokes┸D┻E┻ゅなひひばょ┻PasteurfsQuadrant┽BasicScienceandTechnological)nnovation┸Brookings)nstitutionPressWashington┻SungNs┸C┻J┻W┻F┻G┻M┻andetal┻ゅにどどぬょ┻gCentralchallengesfacingthenationalclinicalresearchenterprise┻gJAMA289ゅなどょ┺なにばぱ┽なにぱば┻Taylor┸A┻and(┻R┻Greveゅにどどはょ┻gSupermanorthefantasticfour╂Knowledgecombinationandexperienceininnovativeteams┻gAcademyofManagementJournal49ゅねょ┺ばにぬ┽ばねど┻Toole┸A┻A┻andD┻Czarnitzkiゅにどどひょ┻gExploringtherelationshipbetweenscientisthumancapitalandfirmperformance┻ThecaseofbiomedicalacademicentrepreneursintheSB)Rprogram┻gManagementScience55ゅなょ┺などな┽ななね┻Weisberg┸R┻W┻ゅなひひひょ┻gCreativityandKnowledge┺AChallengetoTheories┻g(andbookofcreativity┺にには┻Williams┸K┻Y┻andC┻A┻OfReillyゅなひひぱょ┻gDemographyanddiversityinorganizations┺Areviewofねどyearsofresearch┻gResearchinorganizationalbehavior20┺ばば┽なねど┻Woolf┸S┻(┻ゅにどどぱょ┻gThemeaningoftranslationalresearchandwhyitmatters┻gJama299ゅにょ┺になな┽になぬ┻Wuchty┸S┻┸B┻F┻Jones┸etal┻ゅにどどばょ┻gTheincreasingdominanceofteamsinproductionofknowledge┻gScience316ゅのぱにばょ┺などぬは┽などぬひ┻Zerhouni┸E┻ゅにどどばょ┻gTranslationalresearch┺movingdiscoverytopractice┻gClinicalPharmacology┃Therapeutics81ゅなょ┺なには┽なにぱ┻Zucker┸L┻G┻andM┻R┻Darbyゅなひひのょ┻Virtuouscirclesofproductivity┻Starbioscientistsandtheinstitutionaltransformationofindustry┻Cambridge┸Mass┻┸NationalBureauofEconomicResearch┻Zucker┸L┻G┻andM┻R┻Darbyゅなひひはょ┻gStarscientistsandinstitutionaltransformation┺Patternsofinventionandinnovationintheformationofthebiotechnologyindustry┻gProceedingsoftheNationalAcademyofSciences93ゅにぬょ┺なにばどひ┽なにばなは┻Zucker┸L┻G┻┸M┻R┻Darby┸etal┻ゅにどどなょ┻gCommercializingKnowledge┺UniversityScience┸KnowledgeCapture┸andFirmPerformanceinBiotechnology┻gNationalBureauofEconomicResearchWorkingPaperSeriesNo.8499┻Zucker┸L┻G┻┸M┻R┻Darby┸etal┻ゅなひひぱょ┻g)ntellectualhumancapitalandthebirthofUSbiotechnologyenterprises┻gTheAmericaneconomicreview88ゅなょ┺にひど┽ぬどは┻

Page 37: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬは

Table 1: Summary Statistics

Variable Mean Std. Dev. Min Max Mean Std. Dev. Min Max Mean Std. Dev. Min Max

Dependent variablesuccess 0.43 0.49 1 0.45 0 1 0.40 0 1

Independent variablesd_bridging 0.52 0.50 1 0.44 0 1 0.61 0 1

d_basicpubs_only 0.30 0.46 1 0.33 0 1 0.26 0 1

d_appliedpubs_only 0.18 0.39 1 0.23 0 1 0.13 0 1

d_balancedbridging 0.27 0.44 1 0.24 0 1 0.31 0 1

share_basic 61.56 37.70 0 100 59.12 39.87 0 100 64.40 34.80 0 100

d_top_citbridging 0.07 0.26 1 0.04 0 1 0.11 0 1

d_top_citbasic 0.10 0.30 1 0.08 0 1 0.13 0 1

d_top_citapplied 0.11 0.31 1 0.10 0 1 0.12 0 1

d_basic_only 0.03 0.18 1 0.05 0 1 0.02 0 1

d_applied_only 0.01 0.11 1 0.01 0 1 0.01 0 1

d_bridgingPIs_only 0.57 0.50 1 n/a 0.57 0 1

d_mixofPIsbridging 0.39 0.49 1 n/a 0.39 0 1

d_bridgingthroughteam 0.01 0.10 1 n/a 0.01 0 1

fielddivexp 0.98 1.63 0 31 0.59 0.82 0 6 1.43 2.14 0 31

disexp 0.69 1.54 0 27 0.59 1.30 0 11 0.80 1.78 0 27

Control variablesno_investig 2.09 1.65 1 8 n/a 3.35 1.71 2 8

no_pubs 11.61 20.89 1 265 7.75 13.84 1 260 16.10 26.15 1 265

d_industry 0.36 0 1 0.40 0 1 0.31 0 1

d_bigpharma 0.05 0 1 0.05 0 1 0.06 0 1

sponsor_trialexp 326.82 492.31 1 1979 327.61 482.90 1 1979 325.91 503.14 1 1979

sponsor_disexp 59.33 150.40 0 1021 62.95 153.24 0 977 55.12 146.95 0 1021

priordrugtrials_samephase 15.64 34.53 0 252 17.95 38.21 0 252 12.96 29.47 0 252

enrollment 89.49 231.98 1 6600 81.65 223.25 1 6600 98.59 241.47 1 5029

location_count 4.30 8.46 1 50 4.20 8.36 1 50 4.41 8.58 1 50

top10site 0.09 0 1 0.10 0 1 0.08 0 1

USA 0.60 0 1 0.64 0 1 0.56 0 1

d_phase0 0.04 0 1 0.02 0 1 0.05 0 1

d_phase1 0.33 0 1 0.32 0 1 0.35 0 1

d_phase12 0.11 0 1 0.11 0 1 0.11 0 1

d_phase2 0.46 0 1 0.50 0 1 0.41 0 1

d_phase23 0.06 0 1 0.05 0 1 0.07 0 1

endocrinedisease 0.07 0 1 0.08 0 1 0.07 0 1

cancer 0.43 0 1 0.44 0 1 0.42 0 1

circulatorydisease 0.05 0 1 0.04 0 1 0.06 0 1

mentaldisease 0.07 0 1 0.07 0 1 0.06 0 1

infectiousdisease 0.08 0 1 0.08 0 1 0.08 0 1

otherdisease 0.30 0 1 0.29 0 1 0.31 0 1

Full Sample (N=4816) Sample with only one PI (N=2589 Sample with a team of Pis (N=2227)

Page 38: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬば

Table 2: Bridging Skills and the Impact on Trial Success

(1a) (1b) (2a) (2b) (3a) (3b) (4a) (4b) (5a) (5b)

VARIABLES success success success success success success success success success success

IV Probit IV Probit IV Probit IV Probit IV Probit

d_basicpubs_only 0.001 -0.021

(0.023) (0.023)

d_bridging 0.045* 0.046*

(0.024) (0.025)

d_balancedbridging 0.065*** 0.077***

(0.017) (0.017)

share_basic 0.002*** 0.003***

(0.001) (0.001)

share_basic_squ -0.000*** -0.000***

(0.000) (0.000)

d_top_citbasic -0.019 -0.054*

(0.030) (0.029)

d_top_citapplied 0.059** 0.083***

(0.028) (0.028)

d_top_citbridging -0.080** -0.081**

(0.040) (0.038)

ln_disexp 0.516*** 0.070*** 0.519*** 0.069*** 0.518*** 0.070*** 0.516*** 0.068*** 0.503*** 0.068***

(0.087) (0.014) (0.087) (0.014) (0.087) (0.014) (0.088) (0.014) (0.087) (0.014)

priordrugtrials_samephas 0.004*** 0.007*** 0.004*** 0.007*** 0.004*** 0.007*** 0.004*** 0.007*** 0.004*** 0.007***

(0.000) (0.001) (0.000) (0.001) (0.000) (0.001) (0.000) (0.001) (0.000) (0.001)

no_investig -0.017*** -0.002 -0.017*** -0.002 -0.018*** -0.003 -0.018*** -0.003 -0.016*** -0.001

(0.006) (0.005) (0.006) (0.005) (0.006) (0.005) (0.006) (0.005) (0.006) (0.005)

ln_no_pubs -0.025*** -0.008 -0.037*** -0.024** -0.029*** -0.014* -0.033*** -0.017* -0.016 0.003

(0.008) (0.008) (0.010) (0.010) (0.008) (0.008) (0.009) (0.009) (0.012) (0.011)

d_industry 0.032* 0.034** 0.032* 0.033* 0.032* 0.034** 0.032* 0.033* 0.031* 0.034*

(0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017)

d_bigpharma -0.186*** -0.068 -0.185*** -0.066 -0.183*** -0.064 -0.184*** -0.066 -0.180*** -0.066

(0.050) (0.043) (0.050) (0.043) (0.050) (0.043) (0.050) (0.043) (0.049) (0.043)

sponsor_trialexp 0.000* 0.000 0.000* 0.000 0.000* 0.000 0.000* 0.000 0.000* 0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

sponsor_disexp -0.000 0.000 -0.000 0.000 -0.000 0.000 -0.000 0.000 -0.000 0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

enrollment -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

location_count 0.002* 0.003*** 0.002* 0.003*** 0.002* 0.003*** 0.002* 0.003*** 0.002* 0.003***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

top10site -0.010 -0.008 -0.010 -0.008 -0.011 -0.010 -0.011 -0.009 -0.009 -0.005

(0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030)

USA -0.078*** -0.051*** -0.078*** -0.051*** -0.077*** -0.050*** -0.077*** -0.050*** -0.076*** -0.049***

(0.018) (0.017) (0.018) (0.017) (0.018) (0.017) (0.018) (0.017) (0.018) (0.017)

d_phase0 -0.019 -0.057 -0.018 -0.055 -0.016 -0.054 -0.015 -0.051 -0.019 -0.056

(0.043) (0.041) (0.043) (0.041) (0.043) (0.041) (0.043) (0.042) (0.043) (0.041)

d_phase12 0.068*** 0.059** 0.069*** 0.061** 0.068*** 0.059** 0.070*** 0.063** 0.064** 0.054**

(0.026) (0.025) (0.026) (0.025) (0.026) (0.025) (0.026) (0.025) (0.026) (0.025)

d_phase2 -0.055*** -0.080*** -0.056*** -0.082*** -0.058*** -0.083*** -0.057*** -0.083*** -0.058*** -0.085***

(0.018) (0.018) (0.019) (0.018) (0.019) (0.018) (0.018) (0.018) (0.018) (0.018)

d_phase23 0.015 -0.023 0.015 -0.025 0.016 -0.022 0.015 -0.024 0.013 -0.025

(0.035) (0.033) (0.035) (0.033) (0.035) (0.034) (0.035) (0.034) (0.035) (0.034)

cancer 0.069** 0.041 0.066** 0.037 0.067** 0.039 0.065** 0.035 0.070** 0.042

(0.031) (0.030) (0.031) (0.030) (0.031) (0.030) (0.031) (0.031) (0.030) (0.031)

circulatorydisease -0.039 -0.121*** -0.040 -0.123*** -0.041 -0.123*** -0.041 -0.123*** -0.040 -0.120***

(0.046) (0.040) (0.046) (0.040) (0.046) (0.040) (0.046) (0.041) (0.045) (0.041)

mentaldisease 0.052 -0.002 0.050 -0.008 0.047 -0.010 0.046 -0.014 0.049 -0.006

(0.042) (0.039) (0.042) (0.039) (0.042) (0.039) (0.042) (0.039) (0.042) (0.040)

infectiousdisease 0.053 0.026 0.050 0.020 0.043 0.015 0.042 0.008 0.046 0.012

(0.039) (0.038) (0.039) (0.038) (0.039) (0.038) (0.040) (0.038) (0.040) (0.039)

otherdisease 0.046 -0.056* 0.046 -0.057* 0.046 -0.057* 0.046 -0.057* 0.046 -0.055*

(0.036) (0.030) (0.036) (0.030) (0.036) (0.030) (0.036) (0.030) (0.036) (0.030)

Constant 0.272*** 0.275*** 0.267*** 0.280*** 0.259***

(0.042) (0.045) (0.042) (0.044) (0.043)

Observations 4,816 4,816 4,816 4,816 4,816 4,816 4,816 4,816 4,816 4,816

ll ど3661 ど2909 ど3665 ど2903 ど3659 ど2899 ど3655 ど2897 ど3625 ど2896F 29.22 27.02 28.69 27.60 26.78

r2_p 0.115 0.117 0.118 0.119 0.119

Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1

Page 39: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬぱ

Table 3: Breadth of Experience of Fields and Interactions with Bridging Variables

(1) (2) (3) (4)

success success success success

VARIABLES IV IV IV IV

d_basicpubs_only 0.013

(0.034)

d_bridging 0.264***

(0.069)

d_balancedbridging 0.234***

(0.046)

d_top_citbasic -0.019

(0.039)

d_top_citapplied 0.087**

(0.035)

d_top_citbridging -0.011

(0.065)

ln_fielddivexp -0.551*** -0.499*** -0.540*** -0.535***

(0.123) (0.125) (0.125) (0.120)

int_lnfielddivexp_bridging -0.442***

(0.116)

int_lnfielddivexp_balancedbrid -0.305***

(0.074)

int_lnfielddivexp_topcitbridging -0.106*

(0.061)

ln_disexp 1.109*** 1.432*** 1.230*** 1.088***

(0.240) (0.352) (0.276) (0.238)

ln_no_pubs -0.017* -0.034** -0.024** -0.010

(0.010) (0.014) (0.011) (0.015)

no_investig 0.020** 0.027*** 0.022*** 0.021***

(0.008) (0.010) (0.009) (0.008)

priordrugtrials_samephase 0.004*** 0.004*** 0.004*** 0.004***

(0.000) (0.000) (0.000) (0.000)

d_industry 0.033 0.036 0.038* 0.031

(0.022) (0.025) (0.023) (0.022)

d_bigpharma -0.223*** -0.235*** -0.223*** -0.215***

(0.067) (0.078) (0.071) (0.066)

sponsor_disexp 0.000 0.000 0.000 0.000

(0.000) (0.000) (0.000) (0.000)

sponsor_trialexp 0.000* 0.000 0.000 0.000*

(0.000) (0.000) (0.000) (0.000)

enrollment -0.000 -0.000 -0.000* -0.000

(0.000) (0.000) (0.000) (0.000)

location_count 0.002* 0.003** 0.003** 0.002*

(0.001) (0.001) (0.001) (0.001)

top10site -0.019 -0.048 -0.035 -0.019

(0.037) (0.044) (0.039) (0.037)

USA -0.081*** -0.075*** -0.078*** -0.077***

(0.023) (0.025) (0.024) (0.023)

Constant 0.266*** 0.104 0.193*** 0.249***

(0.0536) (0.0953) (0.0654) (0.0551)

Observations 4,816 4,816 4,816 4,816

F 17.21 11.88 14.73 15.41

Note: Phase and Field Fixed Effects included.

Page 40: Bridging the Gap from Inventions to Innovations ...€¦ · bridge the – so far fairly understood – gap from the invention to testing stage. Even in the exploitation phase, where

ぬひ

Table 4: Models with Trials with a Team of Investigators (>1)

(1a) (1b) (2) (3) (4) (5) (6) (7) (8) (9) (10)

success success success success success success success success success success success

VARIABLES Probit IV IV IV IV IV IV IV IV IV IV

d_basicpubs_only -0.030 -0.005 -0.012

(0.039) (0.038) (0.048)

d_bridging 0.025 0.034 0.239***

(0.041) (0.039) (0.077)

d_balancedbridging 0.059** 0.227***

(0.024) (0.051)

share_basic 0.002

(0.001)

share_basic_squ -0.000

(0.000)

d_top_citbasic -0.020 -0.022

(0.041) (0.048)

d_top_citapplied 0.066* 0.108**

(0.039) (0.046)

d_top_citbridging -0.047 0.052

(0.052) (0.079)

d_basic_only 0.202 0.213

(0.130) (0.198)

d_bridgingPIs_only 0.267** 0.402**

(0.104) (0.166)

d_mixofPIsinbridging 0.277*** 0.387**

(0.104) (0.166)

d_bridgingthroughteam 0.220 0.188

(0.148) (0.217)

ln_fielddivexp -0.401*** -0.341*** -0.378*** -0.391*** -0.265

(0.097) (0.090) (0.094) (0.096) (0.251)

int_lnfielddivexp_bridging -0.294***

(0.080)

int_lnfielddivexp_balancedbrid -0.231***

(0.063)

int_lnfielddivexp_topcitbridging -0.121*

(0.065)

int_lnfielddivexp_basic_only -0.109

(0.310)

int_lnfielddivexp_bridging_only -0.182

(0.239)

int_lnfielddivexp_mix -0.150

(0.239)

int_lnfielddivexp_bridging_team 0.177

(0.292)

ln_disexp 0.058*** 0.522*** 0.518*** 0.521*** 0.511*** 0.530*** 0.893*** 1.088*** 0.965*** 0.892*** 0.943***

(0.020) (0.103) (0.102) (0.102) (0.102) (0.103) (0.206) (0.270) (0.227) (0.206) (0.216)

priordrugtrials_samephase 0.012*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005*** 0.005***

(0.002) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.001) (0.000) (0.001)

no_investig 0.012* -0.008 -0.008 -0.008 -0.007 -0.009 0.013* 0.012 0.013 0.014* 0.013

(0.007) (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) (0.008) (0.009)

ln_no_pubs -0.031** -0.043*** -0.037*** -0.038*** -0.027* -0.039*** -0.024* -0.046*** -0.029** -0.020 -0.031**

(0.014) (0.013) (0.011) (0.012) (0.016) (0.011) (0.012) (0.017) (0.013) (0.019) (0.013)

d_industry -0.031 -0.029 -0.029 -0.029 -0.028 -0.028 -0.035 -0.030 -0.030 -0.038 -0.035

(0.028) (0.027) (0.027) (0.027) (0.026) (0.027) (0.031) (0.034) (0.032) (0.031) (0.032)

d_bigpharma 0.095 -0.091 -0.084 -0.090 -0.087 -0.099 -0.064 -0.055 -0.053 -0.056 -0.079

(0.079) (0.079) (0.079) (0.079) (0.078) (0.079) (0.089) (0.096) (0.091) (0.089) (0.092)

sponsor_trialexp -0.000 0.000 0.000 0.000 0.000 0.000 -0.000 -0.000 -0.000 -0.000 -0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

sponsor_disexp 0.000 0.000 0.000 0.000 0.000 0.000 0.000** 0.000*** 0.000** 0.000* 0.000**

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

enrollment 0.000 0.000 0.000 0.000 0.000 0.000 -0.000 -0.000 -0.000 -0.000 -0.000

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

location_count 0.002 0.001 0.001 0.001 0.001 0.001 0.001 0.000 0.001 0.001 0.000

(0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (0.002) (0.002) (0.002)

top10site 0.000 0.001 -0.002 0.000 -0.000 -0.002 0.000 -0.022 -0.015 0.001 -0.001

(0.047) (0.044) (0.044) (0.044) (0.044) (0.044) (0.051) (0.056) (0.053) (0.051) (0.053)

USA -0.048* -0.099*** -0.097*** -0.099*** -0.098*** -0.098*** -0.091*** -0.084** -0.088*** -0.089*** -0.091***

(0.026) (0.027) (0.027) (0.027) (0.027) (0.027) (0.031) (0.033) (0.032) (0.031) (0.032)

Constant 0.159** 0.146** 0.164*** 0.147** -0.099 0.174*** 0.055 0.100 0.160** -0.192

(0.064) (0.059) (0.063) (0.061) (0.116) (0.067) (0.094) (0.075) (0.070) (0.175)

Observations 2,227 2,227 2,227 2,227 2,227 2,227 2,227 2,227 2,227 2,227 2,227

r2_p 0.180

F 16.98 18.18 17.29 16.77 15.82 13.09 9.855 11.64 11.41 9.395

Note: Phase and Field Fixed Effects included.