Industry sponsorship and selection of comparators in randomized clinical trials


Dimitrios N. Lathyris and John P. A. Ioannidis, Clinical Trials and Evidence-Based Medicine Unit, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece. Tel.: +302651007807; fax: +302651007867; e-mail:,


Eur J Clin Invest 2010; 40 (2): 172–182


Background  Most clinical trials on medical interventions are sponsored by the industry. The choice of comparators shapes the accumulated evidence. We aimed to assess how often major companies sponsor trials that involve only their own products.

Methods  Studies were identified by searching for trials registered in 2006. We focused on randomized trials involving the 15 companies that had sponsored the largest number of registered trials in in that period.

Results  Overall, 577 randomized trials were eligible for analysis and 82% had a single industry sponsor [89% (166/187) of the placebo-control trials, 87% (91/105) of trials comparing different doses or ways of administration of the same intervention, and 78% (221/285) of other active control trials]. The compared intervention(s) belonged to a single company in 67% of the trials (89%, 81% and 47% in the three categories respectively). All 15 companies strongly preferred to run trials where they were the only industry sponsor or even the only owner of the assessed interventions. Co-sponsorship typically reflected co-ownership of the same intervention by both companies. Head-to-head comparison of different active interventions developed by different companies occurred in only 18 trials with two or more industry sponsors.

Conclusions  Each company generates a clinical research agenda that is strongly focused on its own products, while comparisons involving different interventions from different companies are uncommon. This diminishes the ability to understand the relative merits of different interventions for the same condition.


Sponsorship may affect the results and interpretation of research, including epidemiological studies [1], meta-analyses [2] and randomized trials [3,4]. Almost all high-profile trials are currently co-sponsored by the industry and the majority is funded by the industry alone without any concomitant public funding [5]. Industry-sponsored trials are not of worse quality than other trials [3], but more favourable results of industry-sponsored trials may be due to design issues, in particular the choice of comparators that are either inactive (e.g. placebo or no treatment, while effective comparators exist) or suboptimal (‘straw man comparators’) [6].

The selection of comparators in clinical trials requires further study. For many diseases, there exist many different regimens to choose as comparators against an experimental treatment. Furthermore, many new experimental treatments arise that may be tested against each other. However, these treatments often belong to different companies. It is unknown how often companies are willing to compare their products against licensed regimens belonging to other companies. Such head-to-head comparisons may be possible to perform even by one company, but ideally their facilitation requires the approval and participation of all owners. While such comparisons can yield useful evidence, companies may fear that if one treatment is shown to be inferior, its market share will be jeopardized. Conversely, some companies may sponsor together the same trial simply because they have developed together some agent and both benefit from its marketing.

The availability of a large registry such as facilitates empirical analyses of trials sponsored by the industry and one can study the antagonistic or synergistic behaviour of companies in selecting comparators. Here we analysed registered clinical trials to assess how often major companies sponsor trials that involve only their own products and how often they sponsor trials where their products are directly compared to licensed products of other major companies.


Eligible trials

We used trials registered in in 2006. We selected trials sponsored by at least one of the 15 companies with the most prolific clinical trial agenda, as determined by their number of registered trial records in that time period, regardless of study design. Records of trials sponsored by subsidiaries or branches that shared the same name(s) were merged with the mother company (e.g. Abbott Molecular, Abbott Nutrition, Abbott, Abbott Park, IL, USA) before determining the 15 most prolific companies.

The records of the clinical studies sponsored by at least one of the top-15 companies were then scrutinized to retain only randomized trials.

Data extraction

Eligible trials were downloaded from and imported into an electronic database on October 2007. From each trial we extracted: identification number, sponsors, sample size (actual for completed and no longer recruiting, anticipated for others), condition(s), intervention(s), design (parallel or crossover), phase, starting date and status. Electronically extracted information was scrutinized by two investigators (DNL, NAP). A third investigator (JPAI) arbitrated on remaining disagreements.

We assigned interventions to specific industry owners. Those with expired patent that could be manufactured as generics were not assigned to any particular company. Information was based on the trial record and was complemented by perusing the Physician’s Desk Reference (PDR, 61st Edition, 2007) and the British National Formulary (BNF, version 53, March 2007). When there were two or more company owners of the same intervention (e.g. they had co-developed and/or co-marketed the intervention), then we counted both of them as owners. When the owner of an intervention couldn’t be found in PDR and BNF and to update information on the status of the patent, we also searched Wikipedia, company websites and drug-related websites.

Financial agreements among companies are common. To avoid potential bias from studies that compared interventions which initially seemed to belong to different companies but eventually belonged to one parent company or to companies with common financial interests, we tried to identify and merge affiliates, subsidiaries, or branches that have different names. We searched on internet until 31 December 2007 – considering that 1-year period after December 2006 is a reasonable period to detect possible emerging financial agreements – the official sites of companies, sites of media specialized in financial news and Wikipedia for information on purchasing, mergers, extended cooperation agreements and commercialization of drugs.


Eligible randomized trials were grouped into three categories: trials comparing an intervention against placebo with no active treatment given in the control arm; trials comparing different doses or ways of administration of the same single intervention (regardless of whether additional treatments were given to all compared arms as common backbone); and trials that compared different active interventions in at least two arms [again, regardless of whether there were also common backbone intervention(s)].

We estimated the number of trials with single owner and those with more than one owner; a sensitivity analysis was performed excluding common backbone interventions. By default, placebo-controlled trials and dose-comparisons favour the involvement of a single sponsor and typically there would also be a single owner, unless two or more companies collaborate in developing the same intervention. Conversely, comparisons of active interventions may contain a larger proportion of trials with more than one industry sponsor or more than one industry owner, unless companies avoid co-sponsoring trials with other companies and avoid using agents owned by other companies even when they perform head-to-head comparisons. We also examined the listed concomitant non-industry sponsors, but their reporting may be less comprehensive (e.g. some trials list institutions where the trial takes place as sponsors, while others do not).

We visualized the sponsorship and ownership patterns in social networks [7]. Social networks examine the links between players in a community and find increasing applications in medicine [8]. Here the players are the 15 companies and the links are trials where these companies are sponsors or owners. Both sponsorship and ownership networks have 15 nodes, each corresponding to one of the 15 companies. For the sponsorship network, there are links between two nodes, when at least one trial was co-sponsored by the two respective companies. The thickness of each link is proportional to the number of trials co-sponsored by the two respective companies. Moreover, there are loops around each node (self-links or auto-loops) with thickness proportional to the number of trials sponsored solely by the company represented by the respective node. For the ownership networks, the thickness of the line connecting each pair of nodes is proportional to the number of trials where both of the linked companies own interventions. The thickness of the auto-loops is proportional to the number of trials where the respective company is the owner of all the interventions included in the trial. For the ownership analyses, we present networks with and without considering the common backbone interventions.

The co-existence of two or more companies in the same trial does not necessarily mean that the two companies are involved in a competitive, antagonistic head-to-head comparison. Conversely, two companies may own the same intervention together. Therefore, we also scrutinized carefully all trials with comparison of different active interventions to identify how many of these pertain to situations where there is a head-to-head comparison of interventions owned by different companies.

Finally, we examined specifically which of the analysed trials had also a listed sponsor that was a government agency and whether these trials were sponsored by a single or multiple companies and whether they were comparing interventions by a single or multiple companies.



The 15 most prolific companies were (alphabetically) Abbott, Amgen, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb (BMS), Eli Lilly, Genentech, GlaxoSmithKline, Hoffman – La Roche, Johnson & Johnson, Merck, Novartis, Pfizer, Sanofi-Aventis and Wyeth. Each sponsored between 36 and 133 clinical trials with records dated within 2006. Cumulatively, these companies had 1004 studies registered during 2006; 427 records were excluded and 577 eligible were analysed (Fig. 1), including 187 placebo-control trials, 105 trials comparing different doses or ways of administration of the same single intervention (three included also a common backbone) and 285 trials comparing different active interventions (109 included a common backbone). Tables S1 and S2 show data per company.

Figure 1.

 Flowchart of eligible trials.

Table 1 shows the characteristics per category of the eligible trials. The five most frequently studied clinical conditions were cardiovascular diseases, cancer, pulmonary diseases, psychiatric disorders and diabetes mellitus. The majority was parallel arm trials. There was almost equal representation of phase II, III and IV trials. Half of the trials were still recruiting and almost as many were completed or not recruiting, with only a minority not yet open to recruitment. On average trials comparing different active interventions were twice as large as trials with other types of comparisons.

Table 1.   Characteristics of eligible trials
 Trials comparing an intervention against placeboTrials comparing different doses or ways of administration of the same single interventionTrials comparing different active interventions in at least two arms.Total
  1. Numbers refer to numbers of studies.

  2. *Studies suspended or withdrawn prior to recruitment of patients.

Number or trials187105285577
Sample, median (IQR)126 (40–395)120 (48–360)271 (84–600) 
Clinical conditions
 Cardiovascular disorders18 (9·6%)8 (7·6%)42 (14·7%)68 (11·8%)
 Cancer10 (5·3%)8 (7·6%)42 (14·7%)60 (10·4%)
 Pulmonary diseases23 (12·3%)6 (5·7%)24 (8·4%)53 (9·2%)
 Psychiatric disorders18 (9·6%)9 (8·6%)22 (7·7%)49 (8·5%)
 Diabetes mellitus16 (8·5%)8 (7·6%)17 (5·9%)41 (7·1%)
 Others102 (54·5%)66 (62·8%)138 (48·4%)306 (53%)
 Parallel studies154 (82·3%)84 (80%)252 (88·4%)490 (84·9%)
 Cross over studies32 (17·1%)16 (15·2%)25 (8·8%)73 (12·6%)
 Factorial studies1 (0·05%)5 (4·7%)8 (2·8%)14 (2·4%)
Starting day
 Pre-200446 (24·6%)9 (8·6%)51 (17·9%)106 (18·3%)
 2004–200520 (10·7%)12 (11·4%)38 (13·3%)70 (12·1%)
 2006107 (57·2%)70 (66·7%)173 (60·7%)350 (60·6%)
 After 20067 (3·7%)5 (4·7%)7 (2·4%)19 (3·2%)
 Unknown7 (3·7%)9 (8·6%)16 (5·6%)16 (2·8%)
 Completed51 (27·3%)25 (23·8%)57 (20%)133 (23%)
 Terminated7 (3·7%)4 (3·8%)6 (2·1%)17 (2·9%)
 Recruiting83 (44·4%)49 (46·7%)150 (52·6%)282 (48·9%)
 No longer recruiting35 (18·7%)21 (20%)60 (20·9%)116 (20%)
 Not open for recruitment11 (5·9%)4 (3·8%)11 (3·8%)26 (4·5%)
 Other* 2 (1·9%)1 (0·03%)3 (0·05%)
 I or I/II10 (5·3%)24 (22·8%)13 (4·5%)47 (8·1%)
 II or II/III51 (27·3%)25 (23·8%)52 (18·2%)128 (22·1%)
 III66 (35·3%)31 (29·5%)137 (47·7%)234 (40·5%)
 IV48 (25·7%)21 (20%)60 (20·9%)129 (22·3%)
 None reported129 (6·4%)4 (3·8%)23 (8%)39 (6·7%)


There was a single industry sponsor in 82% of the trials (478/577). The proportion was 89% (166/187) of the placebo comparisons, 87% (91/105) of the trials comparing different doses or ways of administration of the same intervention, and even 78% (221/285) of the other active control trials. When two or more companies were involved, but only one intervention featured in the trial (n = 35 trials), almost always (31/35) that intervention had been co-developed and/or co-marketed by two or more companies (e.g. etanercept is marketed by both Wyeth and Amgen), with the exceptions of only two trials comparing an intervention against no treatment, and two trials comparing different doses, where such co-ownership was not readily discernible. Therefore, 88% (509/577) of the trials either had a single sponsor that evaluated only its own product(s) or evaluated a single drug that was marketed by two sponsors. At least one non-industry sponsor was listed in 44% (82/187), 38% (40/105) and 44% (126/285) of the trials in the three categories respectively (P = 0·53).


The intervention(s) involved belonged to a single company in 67% of the trials (384/577). This included 89% (166/187), 81% (85/105) and 47% (133/285) of the trials in the three categories respectively. Common backbone treatments were involved in 112 trials. Excluding common backbone interventions, the remaining intervention(s) belonged to a single company in 69% of the trials (89% (166/187), 82% (86/105) and 51% (145/285) of the trials in the three categories respectively (P < 0·001)).


The 15 selected companies were the unique industry sponsors in most trials they sponsored. The thickness of the auto-loops (trials sponsored by a single company) is gigantic in proportion to the thin links between different pairs of companies (Fig. 2). Eighty-three of the 105 possible pairs of companies had never appeared co-sponsoring even a single trial in this network and only two pairs of companies had more than three trials co-sponsored. The maximum number of co-sponsored trials was 11 (for the pair of BMS and Sanofi-Aventis), but none of these 11 trials were antagonistic comparisons; simply the same drug was marketed by both companies (irbesartan in six trials, clopidogrel in five trials).

Figure 2.

 Networks of sponsorship for the 15 companies across 577 trials. Each company is shown by a node with diameter proportional to the number of trials sponsored. Lines represent co-sponsorship between industries with the thickness proportional to the number of co-sponsored trials. Dashed lines refer to comparison with only one co-sponsored trial. The thickness of the auto-loops is proportional to the number of trials where the respective company is the unique industry sponsor. Specific numbers can be seen in the accompanying table insert.

Similar patterns were observed in the ownership networks. Thirty-seven of the 105 possible pairs of companies had never been owners within the same trial (Fig. 3a). The maximal number of trials for any pair of companies was 12 (BMS and Sanofi-Aventis, again because of co-ownership of drugs). When common backbone regimens were excluded (Fig. 3b), 47 of the 105 possible pairs of companies had never appeared together as owners within the same trial.

Figure 3.

 Networks of ownership for the 15 companies across 577 trials (a) with consideration of the common backbone interventions and (b) without consideration of the common backbone interventions. Each company is shown by a node with diameter proportional to the number of trials where a company-owned intervention is present. Lines connecting different nodes represent co-existence of ownership within a trial (one or more interventions are owned by one of the two companies and one or more interventions are owned by the other company). Dashed lines refer to comparison with only one trial with co-existence of ownership. Auto-loops in each node represent trials where the company shown by the respective node is the unique industry owner of all interventions involved in these trials. Thickness for both linking lines and auto-loops is proportional to number of trials. Respective numbers are given in the table inserts. Numbers in parentheses represent trials were different companies own interventions compared head-to-head, excluding co-ownerships.

Comparisons of active interventions with different owners and single industry sponsor

Sixty-five trials with a single industry sponsor entailed head-to-head comparisons of active interventions where the compared arms had interventions owned by different companies. Only one of the two or more owners was sponsoring each of these trials. The sponsor owned one of the compared interventions that was almost always (n = 62 trials) given alone, without concomitant interventions from other companies in the same arm. These 62 trials evaluated a total of 45 sponsor-owned drugs; those that appear is two or more trials are listed in Table 2. For these single-sponsored drugs, a comparison against placebo might have been unethical and/or there were other drugs with similar indication that were already well established in the market as being effective. The owners of the established drugs did not sponsor the trial, hence sponsorship was left to the owner of the new drug. For example, denosumab was compared against zolendronic acid or alendronate; tiotropium was compared against fluticasone or beclomethoasone; sirolimus was compared against tacrolimus; and so forth – without the company owning the comparator sponsoring the trial.

Table 2.   Sponsor-owned drugs involved in at least two trials of head-to-head comparisons of active agents where there is only a single company sponsor
Sponsor-owned drugsSponsorConditions (number of trials)TrialsComparators (number of trials)
DenosumabAmgenOsteoporosis (2), Bone metastases (3)5Zoledromic acid (3), Alendronate (2)
TiotropiumBoehringerCOPD (2), Asthma (1)3Salmuterol (3)
TelmisartanBoehringerHypertension (3)3Ramipril (2), Losartan (1)
Divalproex ERAbbottAlzheimer disease (1), Agitation (1), Bipolar disorder (1)3Quetiapine (3)
MontelukastMerckAsthma2Fluticasone (1), Beclomethasone (1)
SirolimusWyethGraft rejection (1), Carotid atherosclerosis in renal transplant (1)2Tacrolimus (2)
Paliperidone ERJohnson & JohnsonSchizophrenia (1), Mood disorders (1)2Quetiapine (2)
TigecyclineWyethBacterial infections2Ertapenem (1), Ampicillin sulbactam (1)
BifeprunoxWyethSchizophrenia2Risperidone (1), Olanzapine (1)
Ezetimibe & simvastatinMerckCardiovascular diseases (1), Hypercholesterolemia (1)2Atorvastatin (2)
SitagliptinMerckDiabetes Mellitus2Pioglitazone (1), Voglibose (1)

In the remaining three trials sponsored by a single company, the sponsor owned one intervention that was combined with interventions owned by other non-sponsoring companies in the same arm: Wyeth sponsored a small trial (n = 16) where its immunosuppressor agent sirolimus was combined with either tacrolimus or mycophenolate; Merck sponsored a 5-arm trial where its hepatitis A vaccine was used in all five arms, but four of the five arms also provided other standard vaccine; and Boehringer sponsored a trial where different doses of its nevirapine were compared against ritonavir-boosted atazanavir while all three arms received also emtricitabine and tenofovir.

Comparisons of active interventions with different owners and two or more industry sponsors

Most of the trials that were co-sponsored by two or more companies and involved a comparison of active interventions, did not actually pertain to antagonistic comparisons of different interventions owned by different companies, but simply reflected co-ownership of the same intervention(s) by two companies. There were only 18 trials that had more than one industry sponsor and examined a head-to-head comparison of different interventions owned by different companies (Table 3). As shown, even in these seemingly antagonistic comparisons, there was often room for synergism. For example, in three trials one comparator was owned by two sponsoring companies. Furthermore, in five trials there was at least one sponsoring company that owned a common backbone regimen given to all arms. Two of these trials also has a company sponsor that did not own any of the compared drugs, but still probably had to gain from the trial regardless of the outcome, since patients would use one of its products. Specifically, one trial comparing sirolimus vs. tacrolimus with mycophenolate common backbone had three sponsors: Wyeth that owns sirolimus, Roche that own the common backbone, and Genzyme that does not own any of these agents but develops treatments for bone marrow toxicity from such immunosuppressants. In another study of pioglitazone vs. rosiglitazone in untreated patients with type 2 diabetes, only one of the sponsoring companies (Takeda) owned a comparator (pioglitazone), while the other sponsoring company (Eli Lily) did not own the comparator rosiglitazone, but is well-known for manufacturing insulin.

Table 3.   Comparisons of active interventions owned by different companies and sponsored by two or more companies IDSponsoring companiesCompared interventions (Owners)Common backbone (Owners)
  1. Two studies (NCT00353834, NCT00360334) had the same sponsoring companies and comparing regimens but different primary outcomes and two studies (NCT00402428, NCT00411385) had different study populations but no difference regarding sponsoring companies, comparators, primary outcomes.

  2. Of the 18 trials above, only three also involved government or public co-sponsors (NCT00273598 Canadian Institutes of Health Research; NCT00383734 French National Agency for Research on AIDS and Viral Hepatitis; NCT00314353 National Surgical Adjuvant Breast and Bowel Project).

  3. J&J: Johnson and Johnson; Roche: Hoffman – La Roche; GSK: GlaxoSmithKline; BMS: Bristol Mayer Squibb; Sanofi: Sanofi-Aventis; Kos: Kos Pharmaceuticals; Synthes: Synthes Canada; OSI: OSI Pharmaceuticals.

NCT00273598J&J, SynthesMoss Miami Spine Instrumentation System (J&J) vs. Universal Spine Instrumentation System (Synthes) 
NCT00275535Wyeth, Roche, GenzymeSirolimus (Wyeth) vs. Tacrolimus (Astellas)Mycophenolate (Roche)
NCT00304772Pfizer, AstellasFluconazole (Pfizer) vs. Micafungin (Astellas) 
NCT00304993GSK, Kos, AbbottNiacin (Kos) vs. Rosiglitazone (GSK)Fenofibrate (Abbott)
NCT00309712Abbott, MerckFenofibrate (Abbott) vs. Simvastatin (Merck) vs. Fenofibrate and Simvastatin (Abbott and Merck) 
NCT00314353Genentech, RocheOxaliplatin (Sanofi) vs. Irinotecan (Pfizer)Bevacizumab (Genentech), Capecitabine (Roche)
NCT00327990GSK, RocheIbandronate (Roche) vs. Alendronate (Merck) 
NCT00331487Takeda, Eli LillyPioglitazone HCl (Takeda) vs. Rosiglitazole (GSK) 
NCT00353834 NCT00360334Amylin, Eli LillyExenatide (Amylin, Eli Lilly) vs. Glargine Insulin (Sanofi) 
NCT00359762Amylin, Eli LillyExenatide (Amylin, Eli Lilly) vs. Glimepiride (Sanofi) 
NCT00374036Roche, PfizerCapecitabine (Roche) vs. Irinotecan (Pfizer) 
NCT00383734Procytech, SanofiNewfill (Sanofi) vs. Eutrophill (Procytech) 
NCT00389402Roche, GileadSaquinavir (Roche) vs. Atazanavir (BMS)Ritonavir (Abbott), Tenofovir +  emtricitabine (BMS, Gilead)
NCT00392665Genentech, OSIBevacizumab (Genentech) vs. Sulindac (Merck)Erlotinib (Genentech, OSI)
NCT00402207AstraZeneca, IVAXExtrafine HFA-beclomethasone (IVAX) vs. HFA-fluticasone (GSK) 
NCT00402428 NCT00411385Human Genome Sciences, NovartisAlbumin interferon alfa-2b (Human Genome Sciences, Novartis) vs. perginterferon alfa-2a (Roche)Ribavirin (Generic)

Trials sponsored by government agencies

Government agencies sponsored 32 of the analysed trials. All of seven placebo comparisons and all of the seven trials comparing different doses or ways of administration of the same intervention with government sponsors had a single industry sponsor that was also the owner of the intervention. Of the 18 active control trials with government sponsors, only three compared interventions by different companies and had two or more industry sponsors (Table 3).


Our review and analysis of trials registered in shows that head-to-head comparisons of interventions owned by different companies are a small minority of industry-sponsored trials. The large majority of industry-sponsored randomized research is sponsored by a single company and examines a single intervention owned by this company. When two companies sponsor the same trial, usually this reflects simply co-ownership of the same agents that are marketed by both sponsors. In the few trials that compare interventions belonging to different companies, usually only one company is sponsoring them, the one trying to prove its new agent superior or non-inferior to some established comparator. The company owning the established comparator is usually not involved.

We found only 18 trials (merely 3% of the total examined in our overview) where different companies owned different interventions that were compared head-to-head and both companies sponsored that same trial. Even among these few exceptions, often the involved companies would have to gain regardless of the outcome of the trial, e.g. if they owned also a common backbone intervention that would be considered to be necessary under all circumstances. All 15 major companies strongly avoided antagonistic trials.

Almost half of the examined trials had also other listed sponsors outside industry, including academic and government institutions. For academic institutions, these data may have inaccuracies. Information on such sponsoring institutions may not be always reported, and conversely, when an academic institution is listed as sponsor it does not mean that it offers funding for the trial or some other material support other than the facilities to run the clinical research. Regardless, the agenda of comparisons seemed to fit almost ubiquitously to the needs of single companies in each trial. This was the case also when we examined more closely the 32 trials that listed government sponsors. Reporting of government sponsors is likely to be more comprehensive and accurate than the reporting of academic sponsors. We should acknowledge that we did not consider trials that are not sponsored at all by the industry and are exclusively sponsored by government or other not-for-profit institutions. However, with the exception of some government and public-sponsored structures (e.g. National Institutes of Health, Veterans Administration, Medical Research Council and World Heath Organization), other institutions are unlikely to run several large, influential trials. Even then, usually there is some involvement and sponsorship by the industry, e.g. provision of the tested interventions and placebos, and such trials would have been captured in our analysis. The industry has been involved as a sponsor in all but one of the 32 most-cited clinical trials published in 1999–2003 [5].

Although several studies [9–11] have evaluated the influence of industry sponsoring on study conclusions, there is little prior evidence on how the industry shapes the research questions to be asked. Our results suggest that even if studies are perfectly conducted and reported, the cumulative picture yields disproportionate evidence on trials that isolate the products of single companies [12,13]. The avoidance of head-to-head comparisons has been noted previously in specific drug classes, including antiepileptics, biologics and biphosphonates [12,14]. Our evaluation demonstrates the lack of head-to-head comparisons across a very diverse range.

Some limitations should be mentioned. We only evaluated trials registered in The registry has gained wide acceptance, but some trials are still not registered there or even in any registry [15]. However, it is unlikely that antagonistic comparisons between different sponsors would be preferentially not registered. It is impossible to verify if all entered data in are correct, but the information is considered reliable and data entry errors are unlikely to be important for the questions we addressed [16–19]. Second, we only analysed trials sponsored by the most prolific companies. We did not examine the behaviour of smaller companies, but these are less influential and occupy smaller shares of the market. We identified many situations where two companies co-owned and co-sponsored the same trial and this included also smaller companies among co-owners and co-sponsors. Small companies may have synergistic relationships with larger ones, and sometimes get eventually absorbed by larger ones. It is unlikely that smaller companies would be more actively pursuing antagonistic research agendas. Finally, we tried to find possible relations of companies, but some small companies may have financial arrangements with larger companies that remain undetected in internet searches. Sometimes, even seemingly not-for-profit organizations are subsidiaries of the industry, as shown in the recent controversy about the funding of the spiral computer tomography trial [20].

Acknowledging these caveats, the most impressive conclusion of our evaluation was that only a very small minority of industry-sponsored randomized trials registered in were truly ‘antagonistic’ trials and evaluated products owned by different major industries. Research, development and marketing choices may underlie this pattern. First, companies that own licensed drugs-products may not wish to risk unfavourable trial results, and their consequent financial implications, by promoting direct comparisons with products of other companies. Second, as drug development is getting more and more expensive, several companies may become specialized in circumscribed, specific medical fields where they face little competition, and may avoid expanding their activities to areas dominated by other companies. Such sharp focusing may exist even for the largest companies that were examined in our review, even though one might expect a most diversified research and development portfolio for these economic giants. The implications for medical research could be very serious. Progressively, medical research may become more and more dependant on financial and commercial criteria, and less on scientific ones; moreover research in major medical fields may become monopolized by one or two major pharmaceutical industries. These issues demonstrate some inherent major disadvantages of industry-sponsored medical research. These disadvantages may be possible to bypass only by committed independent, not-for-profit organizations that are willing to undertake antagonistic head-to-head trials.

Efforts should be made to revitalize and strengthen academic independent clinical research [21–23], e.g. by financing important trials from not-for-profit organizations and enhancing the independence of academic investigators to design trials sponsored by the industry. The industry should also be given incentives to venture more head-to-head antagonistic comparisons. One may also open to debate whether patented drugs should be protected for a shorter period and sponsoring should be controlled by independent government committees. Regulatory pressure may often force companies to perform primarily placebo-controlled trials [24–27], so it is unfair to say that the industry carries all the responsibility for the small number of antagonistic trials observed in our evaluation. The current impetus for comparative effectiveness research [28] may increase interest in pragmatic trials, comparing diverse regimens of interest and with long-term follow-up and clinically important outcomes [29]. This would require the co-ordination of industry, academia and government with a common goal of improving the yield of useful information from clinical trials.


Concept of idea was by JPAI; study design was by all authors; data extraction was by DNL and NAP, data analysis was by all authors, interpretation by DNL and JPAI, writing of the manuscript was by DNL and JPAI. All authors critically commented on the manuscript and approved the final version. JPAI is guarantor of the paper.


No funding.

Competing interests



Intensive Care Unit, General Hospital “George Papanikolaou”, Thessaloniki, Greece (D. N. Lathyris); Clinical Trials and Evidence-Based Medicine Unit, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece (D. N. Lathyris, N. A Patsopoulos, G. Salanti, J. P. A. Ioannidis); Institute for Clinical Research and Health Policy Studies, Tufts Medical Center and Department of Medicine, Tufts University School of Medicine, Boston, MA, USA (J. P. A. Ioannidis).