Scope and Impact of Financial Conflicts of Interest in Biomedical Research:  A Systematic Review FREE

Industry support of biomedical research in the United States increased dramatically in the last 2 decades. Industry's share of total investment in biomedical research and development grew from approximately 32% in 1980 to 62% in 2000, while the federal government's share fell.^1- 2 During this period, the relationship between academic institutions and industry flourished, spawning medical advances, creating new biotechnology markets, and providing needed support for further discovery. However, an entanglement of relationships among industry, investigators, and academic institutions also emerged.

Conflicts of interest are "a set of conditions in which professional judgment concerning a primary interest (such as a patient's welfare or the validity of research) tends to be unduly influenced by a secondary interest (such as financial gain)."³ Financial interests are not the only, or necessarily the most powerful, secondary interests faced by investigators and academic institutions. For investigators, other pressures, including the desire for professional recognition and the need to compete successfully for research funding, are intrinsic to the research process.⁴ Institutions also confront myriad pressures arising from balancing the needs of diverse departments and constituencies. However, financial interests related to biomedical research are nonobligatory and often unrecognized unless disclosed. They are the focus of current national discussion because they induce public anxiety about the influence of money on the research process.^3,5- 6

Several reports suggest that financial ties pose a threat to scientific integrity.^7- 11 In 1999, the tragic death of a participant in an industry-sponsored clinical trial in which investigators and the academic institution held significant financial stakes focused national attention on financial conflicts of interest and safeguards for human participants.¹² The nation's medical leadership warned that public trust in research might falter if action were not taken.^13- 14 However, attempts to develop new conflict of interest policies have encountered substantial controversy. For example, one member of the Association of American Medical Colleges task force charged with developing new guidelines on conflicts of interest believed that the proposed guidelines were too limited.¹⁵ Another member, representing industry, refused to support the guidelines, suggesting that they would only serve to impede innovation.¹⁶

Consensus for reform may only arise from a full understanding of the nature and influence of financial conflicts of interest. In the context of disagreement, a synthesis of evidence may strengthen the bond between impassioned debate and optimal policy.^17- 18 Meta-analyses, which combine the results of several studies, can also derive more definitive conclusions than primary studies alone.¹⁹

We therefore performed a systematic review to answer the following 3 questions: (1) How common are financial relationships among industry, scientific investigators, and academic institutions; (2) What is the impact of these financial relationships; and (3) How are these financial relationships managed? We aimed to develop an objective framework within which more informed policy decision making can occur.¹⁷

Data sources included studies that contained original, quantitative data addressing financial relationships among industry, investigators, and academic institutions. The MEDLINE database was searched from January 1980 through October 2002, using the free text "conflict of interest" and the exploded Medical Subject Heading terms conflict of interest or conflict (psychology) and research support with clinical trials, commerce, or industry, and schools in all possible combinations. Three experts (David Blumenthal, MD, Jerome Kassirer, MD, and David Korn, MD) identified additional studies. Further articles were identified from the reference sections of relevant studies, letters, editorials, comments, and books. This strategy was supplemented by using the Web of Science database to generate a list of articles that cited identified original quantitative studies.

A study was included if it met the following criteria: (1) its stated primary or secondary purpose was to assess the extent, impact, or management of financial relationships among industry, investigators, or academic institutions; (2) it contained a section describing study methods; (3) it was written in English; and (4) it was published following the passage of the Bayh-Dole Act of 1980,²⁰ which encouraged academic institutions and scientific investigators to seek industry partnerships. A study was excluded if it did not contain data pertaining to any of the 3 objectives, or if it was a commentary, letter, editorial, abstract, dissertation, or case study. Two investigators (J.E.B. and C.P.G.) reviewed 1664 citations and selected appropriate studies.

After the initial MEDLINE search, we retrieved a total of 144 articles identified as potentially containing original quantitative data on financial relationships among industry, investigators, and academic institutions. Of these, 21 studies met our inclusion criteria. An additional 16 studies were identified by other sources (8 from other articles' references, 5 from Web of Science, and 3 from experts). Hence, a total of 37 studies were included.

One investigator (J.E.B.) extracted the following unambiguous data for each study: design, data source, type of financial interaction studied (ie, industry sponsorship, consultantship, employment, technology transfer, new venture formation, gifts, or personal funds), and results. For studies assessing the extent of financial relationships, the main outcome was the prevalence of specific types of industry relationships (eg, industry sponsorship, new venture formation, etc). For studies assessing the impact of financial relationships, the main outcomes were the proportion of industry-sponsored and nonindustry-sponsored studies with a certain outcome (ie, pro-industry conclusion) or characteristic (ie, study design or study quality).

Criteria used to appraise methodological quality varied according to study design. To appraise the methodological quality of cross-sectional surveys, we extracted sample size and response rate. For systematic reviews, we extracted how industry sponsorship was defined, whether the outcome or characteristic was defined, and blinding (ie, whether the outcome or characteristic was assessed independently of study sponsorship). For systematic reviews on study quality, we also extracted how the quality assessment instrument was defined or validated. For content analyses and secondary data analyses, we extracted whether the assessment instruments were predesigned, pilot tested, or validated.

Evidence was integrated into tabular displays, drawing from the balance sheet model for integrating and interpreting multiple types of evidence.^21- 22 We presented original data from studies assessing the relation between industry sponsorship and study outcome in a common format: the proportion of industry-sponsored vs nonindustry-sponsored studies that yielded a pro-industry conclusion. We used the authors' published data to calculate these proportions when available. We contacted authors directly if data were not reported.

A pro-industry conclusion was defined as a study outcome that was favorable to industry sponsors. For example, if an industry-sponsored study concluded that a new therapy produced by the sponsoring entity was superior, this was considered a pro-industry conclusion. If a study examined whether an industry product was harmful, we considered a negative result to be a pro-industry conclusion. Neutral studies were classified as neither positive nor negative, and were not considered pro-industry.

We applied meta-analytic techniques to the subset of studies addressing the association between industry sponsorship and outcome in original research. Before pooling the results of these different studies, we determined that the data were not significantly different by testing for data homogeneity.^23- 24 The homogeneity test evaluates whether different odds ratios (ORs) across studies can be explained by the random variation of a common underlining OR.^23- 24 The homogeneity test showed that the ORs for these studies were not significantly different (P = .75) and could therefore be pooled. We then used Mantel-Haenszel techniques to calculate a summary OR.^25- 26 Analyses were performed by using STATA version 6.0 (Stata Corp, College Station, Tex) and P<.05 was considered statistically significant.

Of the 37 studies eligible for inclusion, 10 addressed the extent of financial relationships, 23 addressed the impact of financial relationships, and 8 addressed the management of financial relationships. Four studies addressed more than 1 objective.

Ten studies^27- 36 documented the extent of financial relationships among industry, scientific investigators, and academic institutions. Eight of these studies were cross-sectional surveys and reported response rates of about 60% or more, which approximates the mean response rate among surveys in medical journals (Table 1).³⁷

Studies suggest that 23% to 28% of academic investigators in biomedical research receive research funding from industry.^27,33 A 1998 survey found that 43% of investigators also receive research-related gifts, including biomaterials and discretionary funds.³⁵ Approximately one third of investigators at academic institutions have personal financial ties with industry sponsors. Earlier studies have shown that 37% of investigators in the National Academy of Sciences had "dual affiliations" with both universities and companies.²⁸ A 1992 analysis of 789 articles from major medical journals found that 34% were written by lead authors with relevant personal financial interests in their research (ie, company patents, equity, or advisory board, or director positions).²⁹ An analysis of disclosure forms at a single institution found that 7.6% of investigators had personal financial ties with sponsors of their research, including paid speaking engagements (34%), consulting arrangements (33%), positions on advisory boards (32%), or equity (14%).³⁶ A range of financial interactions clearly exists.

It also appears that life science companies are increasingly involved with academia. A 1986 survey³⁰ reported that 46% of firms supported academic research, while a 1996 survey³¹ found that 92% supported academic research.

Only 1 study investigated the prevalence of industry involvement (defined as equity ownership) among academic institutions. The Association of University Technology Managers reported in 1999 that 124 of its 183 member institutions (68%) in the United States and Canada held equity in businesses that sponsor research performed at the same institutions.³² Some institutions were heavily involved; 27 institutions reportedly held equity in 10 or more start-ups.³²

Relation Between Industry Sponsorship and Study Conclusion. Eleven studies concluded that industry-sponsored research tends to yield pro-industry conclusions (Table 2).^38- 48 The quality of these studies was relatively strong, as all 11 explicitly defined study outcome a priori, although only 7 used a blinded review.^{38,40- 41,45- 48}

Figure 1 demonstrates the OR and 95% confidence interval (CI) for each of the 8 articles that compared the outcome of industry-sponsored vs nonindustry-sponsored original research studies.^38- 44,48 The summary OR for all 8 articles, which together evaluated 1140 original studies, was 3.60 (95% CI, 2.63-4.91). When the studies were stratified into groups involving only randomized controlled trials (RCTs)^38- 40,48 (OR, 4.14; 95% CI, 2.72-6.32) and other study types^41- 44 (OR, 3.00; 95% CI, 1.89-4.77), the findings did not differ significantly (P = .31).

Another study⁴⁵ analyzed 61 nonsteroidal anti-inflammatory drug (NSAID) RCTs, all of which were industry-sponsored, and found that the investigational therapy was comparable with (71.4% of studies) or superior to (28.6% of studies) the comparison therapy. No studies found that the comparison therapy was superior.⁴⁵

Two studies evaluated the relation between industry sponsorship and authors' published positions. A 1998 study⁴⁶ compared authors' financial relationships with industry with their published positions about the safety of calcium channel blockers (Table 2). Authors who had financial relationships with pharmaceutical companies were significantly more likely to reach supportive conclusions than authors without such industry affiliations (51% vs 0%; P<.001).⁴⁶ Similarly, a 1998 analysis⁴⁷ of 106 review articles on the health effects of second-hand smoking showed that industry-sponsored reviews were significantly more likely to yield pro-industry conclusions than nonindustry-sponsored studies (94% vs 13%; P<.001).

Four studies investigated the relation between sponsorship and study design (Table 2). In an analysis of multiple myeloma RCTs, industry-sponsored studies were substantially more likely to use inactive controls (ie, placebo or no-therapy controls) than were nonindustry-sponsored studies (60% vs 21%; P<.001).⁴⁰ The authors also found that the use of inactive controls increased the likelihood of positive study results.⁴⁰ An analysis of 159 RCTs also reported that trials funded by for-profit organizations were more likely to use an inactive control.⁴⁸ In the analysis of NSAID RCTs, the dose of the industry-associated drug was higher than that of the comparison drug in 27 trials (48%), although the dosing was comparable in 23 trials (41%).⁴⁵ Another study⁴⁹ found that industry-sponsored RCTs of oral fluconazole for systemic fungal infections tended to use poorly absorbed oral drugs as comparison agents, thus favoring the success of fluconazole, which is well absorbed from the gastrointestinal tract.

Relation Between Industry Sponsorship and Study Quality. Five analyses reported that industry-sponsored studies were of comparable quality to nonindustry-sponsored studies^{40- 41,50- 52} (Table 3). Four of these studies^{40- 41,50,52} used validated quality-assessment tools and 3 studies^40- 41,52 used a blinded review. Two other studies^53- 54 found that RCTs published in industry-sponsored supplements were generally of lower quality compared with RCTs published in parent journals. These findings were ascribed to a difference in peer-review process between industry-sponsored supplements and parent journals.

Relation Between Industry Sponsorship and Restrictions on Investigator Behavior. Seven studies^{27,31,33,55- 58} investigated whether industry ties prohibit open collaboration or delay publication of results (Table 4). All of these studies were cross-sectional surveys and reported response rates more than 60%. In a 1994 survey³¹ of 210 life science companies, 58% indicated that they typically require investigators to keep information confidential for more than 6 months in order for industry to file a patent. Another analysis⁵⁵ found that industry-sponsored faculty were more likely than other faculty to report delays in the publication of their research results.

Other surveys have suggested that 12% to 34% of academic researchers have requested and been denied access to research results.^55- 56 Controlling for other variables, 2 studies^55,57 found that participation in commercial activities (ie, patenting or start-up companies) was significantly associated with data withholding, although industry sponsorship alone was not.

Industry sponsorship may be associated with a shift in research emphasis from basic research to clinical research.^27,33,58 Faculty members with industry relationships are more than twice as likely as those without such funding to take commercial considerations into account when choosing research topics (35% vs 14%; P<.001).²⁷ In addition, 50% of respondents to a 1995 survey of recombinant DNA researchers believed that industry sponsorship shifts the emphasis of research programs.⁵⁸

Eight studies^36,59- 65 addressed the management of financial relationships among industry, scientific investigators, and academic institutions (Table 5). A 2000 analysis of 17 federal agencies sponsoring human participant research reported that only 4 had policies explicitly governing extramural researchers.⁵⁹ Similarly, 4 surveys of major US academic institutions found substantial variability of policies governing conflicts of interest. A survey of 250 institutions found that management of conflicts and penalties for nondisclosure were almost universally discretionary.⁵⁹ A survey of 10 research-oriented medical schools reported that only 1 institution prohibited investigators from having equity, consulting agreements, or decision-making positions in a company sponsoring their research.⁶¹ Another survey found that only 19% of institutions had specific limits on their faculties' research-related financial interests.⁶⁰ A survey of 122 medical schools reported poor compliance with recently revised guidelines for trial design, access to data, and control over publication in contractual agreements with industry sponsors of clinical research.⁶⁵

Although peer-reviewed journals have taken a role in managing conflicts of interest, journal policies also vary considerably. An analysis of 47 high-impact biomedical journals published in 2000 found that 43% had policies requiring disclosure of conflicts of interest.⁵⁹ A more extensive analysis in 1997 found that 157 of 474 medical journals (33%) and 24 of 922 science journals (3%) had conflict of interest policies in effect.⁶² However, even among journals with stated disclosure policies, few articles contained financial conflict disclosures.^62,64

This comprehensive review of the literature confirms that financial relationships among industry, scientific investigators, and academic institutions are pervasive. About one fourth of biomedical investigators at academic institutions receive research funding from industry.^27,33 One study²⁹ reported that lead authors in 1 of every 3 articles published hold relevant financial interests, while another³² reported that approximately two thirds of academic institutions hold equity in "start-up" businesses that sponsor research performed by their faculty.

Despite the prevalence of these relationships and the broad concerns they have generated, a relative paucity of data has been published describing the impact of financial ties on biomedical research. Although only 37 articles met inclusion criteria, evidence suggests that the financial ties that intertwine industry, investigators, and academic institutions can influence the research process. Strong and consistent evidence shows that industry-sponsored research tends to draw pro-industry conclusions. By combining data from articles examining 1140 studies, we found that industry-sponsored studies were significantly more likely to reach conclusions that were favorable to the sponsor than were nonindustry studies.

There are several possible reasons for this finding. It is possible that, given limited resources, industry only funds potentially winning therapies. However, we found 4 studies that empirically demonstrated that industry preferentially supports trial designs that favor positive results, such as the use of placebo as the comparison therapy in controlled trials.⁴⁰ Comparisons of new therapies to placebo may be appropriate in some cases, although such comparisons are likely to favor the new therapy.^66- 67 The frequent use of placebo controls in clinical trials is often attributed to Food and Drug Administration regulations; however, the Food and Drug Administration does not require the use of placebo and is supportive of trials that incorporate active controls.^68- 70 Use of active controls does not eliminate the potential for bias; indeed, evidence from NSAID and fluconazole RCTs has revealed that inappropriate administration and dosing disparities decrease the effectiveness of active controls.^45,49

Publication bias, or the phenomenon of positive results being published more frequently than negative results, may also play a role in the relation between industry sponsorship and study outcome.⁶⁶ The review of 61 published industry-sponsored RCTs involving NSAIDs, none of which reported a negative conclusion, is consistent with this hypothesis.⁴⁵ This finding is inconsistent with the ethical principle of equipoise, which holds that, over time, there should be no difference in the number of results that favor investigational therapies vs comparison therapies.^40,71 In addition, some have suggested that industry sponsorship may be associated with multiple reporting of studies with positive outcomes, further compounding publication bias and potentially swaying review articles toward more positive results.^49,72

Several studies found that industry-sponsored research appears to be of similar quality to other research.^{40- 41,50- 52} However, studies addressing the quality of industry-sponsored clinical trials used assessment instruments based on selected methodological criteria, such as randomization and blinding. These criteria are important components of high-quality clinical trials, but fall short of determining a study's overall quality.⁷³ Other important considerations should include the relevance of the question being asked and the use of appropriate control therapies.^66,74

Consistent evidence also demonstrated that industry ties are associated with both publication delays and data withholding. These restrictions, often contractual in nature, serve to compound bias in biomedical research. Anecdotal reports suggest that industry may alter, obstruct, or even stop publication of negative studies.^7- 8 Such restrictions seem counterproductive to the arguments in favor of academic-industry collaboration, namely encouraging knowledge and technology transfer. Evidence shows, however, that industry sponsorship alone is not associated with data withholding.⁵⁵ Rather, such behavior appears to arise when investigators are involved in the process of bringing their research results to market.⁵⁵

The extensive equity holdings of academic institutions are particularly concerning. In many ways, these equity arrangements are simply extensions of the increase in university patent licensing activity encouraged by the Bayh-Dole Act.³² However, institutional ownership of equity is different from accrual of patent royalties because ownership carries the responsibility of business stewardship. Equity ownership has created a new revenue model for academic institutions and has induced a dramatic increase in institutional medical entrepreneurialism, further blurring the lines between academic and commercial values. A shift in institutional priorities could potentially affect the distribution of scarce resources.⁷⁵ More research is required to elucidate the extent of these institutional equity holdings and their precise role in realizing the promise of academic research or fostering a shift in the academic mission.

This review identified uneven adherence to methodologic standards. Cross-sectional surveys were almost universally successful in reporting high response rates, and systematic reviews defined outcome measures a priori. However, substantial heterogeneity was found in the use of blinding in systematic reviews. In addition, the potential hazards of financial conflicts of interest should be assessed in light of the potential benefits of academic-industry collaboration. These include significant advances in scientific knowledge and public health, wellness, and productivity.⁷⁶ Future studies should be performed to better understand how collaboration and technology transfer contribute to these benefits.^76- 77

Current management of financial conflicts of interest is in a state of flux. Several studies in this review reported substantial variability among academic institutions and peer-reviewed journals in their policies governing financial conflicts.^59- 62 Efforts to respond to these shortcomings by professional societies and journals have also differed substantially, reflecting the controversy underlying the proposals for reform. Some policies call for the prohibition of certain financial relationships,^78- 80 while others suggest only strict disclosure and monetary limits.^16,81- 85 Journals also have made an attempt to ensure that investigators retain control of and full access to their study data.^86- 87 Despite these efforts, overall compliance of academic institutions and peer-reviewed journals with these guidelines appears poor.^62,64- 65

An effective policy approach to financial conflicts of interest in biomedical research must tread a delicate path. The safety of human participants must remain the paramount concern, bias in the research process must be minimized, and appropriate incentives for research innovation must be preserved. Policy must also take into account the industrialization of clinical research.⁸⁸ Academic institutions are no longer central to research involving human participants. For-profit contract research organizations now consume more than 60% of clinical research funding from industry, leveraging their ability to complete trials more rapidly and less expensively than academic institutions.⁸ In addition, management of financial interests at the institutional level is particularly challenging, as it is questionable whether institutions that stand to gain substantial benefits from research commercialization can still police themselves.^75,89- 90

The variety of proposals for reform likely stems from lack of consensus about the gravity of the problem and the optimal approach for a solution. This review shows that financial relationships are pervasive and problematic. A convergence of pressures, including increasing industry financing of biomedical research and development, encouragement of technology transfer by the federal government, and erosion of academic medical center revenue, will likely lead to increased reliance on industry financing in the future. Lasting and balanced reform may emerge when all stakeholders in the research process build consensus around a system of checks and balances to promote medical innovation while improving oversight and transparency. As a first step in this process, all investigators and sponsors undertaking human participant research should not only fully disclose the nature and extent of their relationships but also make available all research results from completed clinical trials in a comprehensive, publicly accessible registry.^{66,72,89,91- 92} To preserve lasting benefits and enable future advances, close scrutiny will be required to understand and monitor the unintended consequences of academic-industry collaboration.