Journal of Bioethical Inquiry

, Volume 13, Issue 3, pp 439–448 | Cite as

Association Between Financial Conflicts of Interests and Supportive Opinions for Erectile Dysfunction Treatment

  • Rafael Boscolo-Berto
  • Massimo Montisci
  • Silvia Secco
  • Carolina D’Elia
  • Rosella Snenghi
  • Guido Viel
  • Santo Davide Ferrara
Original Research

Abstract

A conflict of interest (COI) is a situation in which a person has competing loyalties or interests that make it difficult to fulfil his or her duties impartially. Conflict of interest is not categorically improper in itself but requires proper management. A SCOPUS literature search was performed for publications on the efficacy/safety of Phospho-Di-Esterase Inhibitors (PDEIs) for treating erectile dysfunction. A categorization tool (CoOpCaT) was used to review and classify the publications as supportive/not-supportive for the discussed active ingredient and reporting or not reporting a COI for that specific drug or for the remaining PDEIs (i.e. competitors). Multivariable binary logistic regression was performed. In the 419 selected records the prevalence of supportive opinions was higher when a COI for the index label was declared. The CoOpCaT showed good internal consistency, discriminative validity and intra/inter-rater agreement. The strongest predictor for a supportive opinion was the total number of financial COIs for the index label. A mild protective effect of the total number of financial COIs for any competitor label was noted. Financial COIs have frequently been associated with bias, and the measures currently adopted to restrain it lack effectiveness. Some evidence for monitoring and/or compensating this bias is reported here, but the ultimate solution remains distant.

Keywords

Sponsor Conflict of interest Bias Ethics Financial disclosure 

Introduction

A conflict of interest (COI) is a situation in which a person has competing loyalties or interests (financial, personal, or professional) that make it difficult to fulfil his or her duties impartially (Lo 2012; Institute of Medicine (US) Committee on Conflict of Interest in Medical Research Education and Practice 2009; Lexchin 2012a).

Medical research investigators, as well as academic institutions, have widespread financial ties to the medical industry that frequently funds research, and it has often been postulated that these ties influence research in important ways (Lee 2008; Stossel 2005; Stossel 2008). Stossel argues that restrictions on researcher interactions with industries and commercial companies could damage research (2005, 2008), while numerous authors believe that industry sponsorship threatens the ability of the researcher to perform scientifically independent investigations (Lee 2008; American Academy of Dermatology Board of Directors 2008; Robertson, Rose, and Kesselheim 2012; Maj 2008; Saver 2012; Lundh et al. 2012; Als-Nielsen et al. 2003).

In recent years, systematic studies of the impact of COIs on research outcomes have mainly focused on randomized controlled trials, and no conclusive evidence (i.e. more positive outcomes in sponsored trials) that establishes causal links between financial ties and biased interpretation of trial results, have been found (Lo 2012; Institute of Medicine (US) Committee on Conflict of Interest in Medical Research Education and Practice 2009; Lexchin 2012a). However, a recent Cochrane review (Lundh et al. 2012) examining forty-eight papers, including cross-sectional studies, cohort studies, systematic reviews, and meta-analyses, showed that industry-sponsored drug and device studies more often had favourable efficacy results (risk ratio 1.32; 95 per cent CI 1.21–1.44), harms results (risk ratio 1.87; 95 per cent CI 1.54–2.27) and conclusions (risk ratio 1.31; 95 per cent CI 1.20–1.44) compared with non-industry-sponsored studies. Moreover, the authors found that in industry-sponsored studies there was less agreement between the results and the conclusions than in non-industry sponsored studies (risk ratio 0.84; 95 per cent CI 0.70–1.01).

Data regarding the magnitude of a potential COI effect in generating supportive opinions in non-systematic reviews, letters to the editor, and editorials are more limited than for randomized controlled trials and meta-analyses. In 2010, Wang et al. examined 202 articles commenting on the risk of myocardial infarction after intake of rosiglitazone and found a strong association between favourable recommendations on the use of rosiglitazone and financial COIs of the authors (risk ratio 3.36; 95 per cent CI 1.94–5.83). In 2011, Fugh-Berman et al. analysed the tone of narrative reviews, letters, and opinion papers regarding menopausal hormone therapy. The authors found an association between articles promoting the use of hormone therapy and a declared financial COI of the author/s, in the absence of any influence of COIs on the scientific accuracy of the paper (Fugh-Berman et al. 2011). Although promotional tone was evaluated without formal criteria, the results suggested that authors who had received payments from industry conveyed more enthusiasm about the industry’s products than non-conflicted authors (Fugh-Berman et al. 2011). Similarly, Yank et al., testing whether financial ties to one drug company were associated with favourable results or conclusions in meta-analyses on the effectiveness of the drug produced by that company found that financial COIs were significantly associated with favourable conclusions (odds ratio 4.09; 95 per cent CI 1.30–12.83) but not with favourable results (Yank, Rennie, and Bero 2007).

In light of these findings, in order to verify if and how financial COIs could influence the opinions (supportive or not-supportive) expressed in original articles (both randomized and non-randomized trials), reviews, letters to the editor, and editorials in Urology, we decided to investigate a narrow subject topic—focussing on “non-industry-sponsored” and/or “partially industry-sponsored” publications about the effectiveness and safety of phospho-di-esterase inhibitors (PDEIs) for the treatment of erectile dysfunction. We defined as “partially industry-sponsored” those publications which, in the absence of any industrial grants/funds for the overall experimental/clinical investigation or for covering any other costs related to the study/publication, included some authors with a declared personal conflict of interest with the company. We defined as “non-industry-sponsored” those publications where the company did not pay for the study/publication and all the authors declared no individual conflicts of interest with the company.

In the time frame considered (2002–2013), only three active ingredients (sildenafil, tadalafil, and vardenafil) registered by three different industrial companies were used for managing erectile dysfunction. This finite scenario facilitated proper identification and classification of the financial COIs between each author and the “index label” (i.e. the label of the active ingredient discussed in the scientific record), as well as between each author and the other two pharmaceutical companies manufacturing PDEIs (that we defined as “competitors”). In this context the number of unreported COIs was strongly limited and thus allowing us to investigate the interactions among conflicted (i.e. those expressing an opinion on a PDEI when declaring a COI for its manufacturer), non-conflicted, and counter-conflicted authors (i.e. those expressing an opinion on a PDEI when declaring a COI for a competitor).

The primary aim of our effort was to test if an association existed between a financial COI declared by at least one author and a “supportive opinion,” defined as a conclusion that supports the efficacy/safety of the “index label” for which the author declared the COI. In addition, we looked for any potential predictive factors of a “supportive opinion” such as the role/s of the conflicted author/s in the manuscript (first author, last author), the total number of COIs, the year of publication, the type of record, the number of involved affiliations, and the existence of any counter-conflict.

Materials and Methods

In August 2013, a systematic search of SCOPUS indexed medical literature was performed to identify relevant publications on the efficacy/safety of PDEIs for the treatment of erectile dysfunction. In the time frame considered (2002–2013) PDEIs comprised only three active ingredients registered by Pfizer, Eli-Lilly and Companies, Bayer and its branches. The considered active ingredients and their commercial labels include: Sildenafil (Viagra: EMEA/H/C/000202 Pfizer Limited and FDA/NDA 020895 Pfizer Ireland); Tadalafil (Cialis: EMEA/H/C/000436 Eli Lilly Nederland B.V. and FDA/NDA 021368 Lilly); Vardenafil (Levitra: EMEA/H/C/000475 Bayer Pharma AG and FDA/NDA 021400 Bayer Healthcare; Staxin: FDA/NDA 200179 Bayer Healthcare; Vivanza: EMEA/H/C/000488 Bayer Pharma AG; Nuviva). The selection process of publications involved the following five sequential steps.
  1. Step 1.

    Systematic Literature Search

     
Two authors (RBB, GV) performed the literature search using the SCOPUS online database. The search strategy was based on free-text protocols as follows: “TITLE-ABS-KEY ((vardenafil OR sildenafil OR tadalafil OR levitra OR cialis OR viagra OR vivanza OR staxin OR nuviva) AND (erectile dysfunction))” combined with “PUBYEAR > 2002, LANGUAGE (english).” The time frame (2002–2013) considered coincided with the period when there was full patent protection for all the included active ingredients.
  1. Step 2.

    Inclusion Criteria and Categorization

     
In order to standardize definitions of terms and concepts discussed in the manuscript, reference was made to the glossary developed by the International Committee of Medical Journal Editors (ICMJE available at http://www.icmje.org/). The list of publications yielded by the search was used to retrieve as many full-text publications as available from the Library System of the University of Padova, which guarantees full access to more than 43,000 e-journals. Publication selection was independently performed by three authors (RBB, GV, RS) according to the following inclusion criteria (Roseman et al. 2011):
  1. 1)

    reporting an evaluation of the therapeutic and/or adverse effects of one or more PDEI (namely Sildenafil, Tadalafil, Vardenafil);

     
  2. 2)

    expressing a supportive/not-supportive opinion on the drug/s;

     
  3. 3)

    reporting a “disclosure statement”;

     
  4. 4)

    being “not industry sponsored” or “partially industry sponsored” as retrievable from the disclosure statement (i.e. see introduction for definitions);

     
  5. 5)

    being a “multi-authored” publication (i.e. written by more than one author);

     
  6. 6)

    being written by non-employees of one of the three considered pharmaceutical companies.

     
Disagreements were settled through discussion followed by unanimous consensus.
  1. Step 3.

    Development and Validation of a Categorization Tool

     
Four authors (RBB, MM, SDF, RS) defined a set of questions based on existing literature and personal experience in order to identify and classify any existing COIs and categorize the type of opinion expressed (Supportive/Not-supportive) (Kjaergard and Als-Nielsen 2002, Stelfox et al. 1998). This process led to the construction of a categorization tool named CoOpCaT (Conflict–Opinion Categorization Tool), composed of two main domains: “Existence of any Conflict of Interest” (Domain 1) and “Expressed opinion” (Domain 2) (Supplemental material—Appendix 1. CoOpCaT). Subsequently, fifty-seven randomly selected publications were independently evaluated by the four authors using the CoOpCaT tool in order to check for internal consistency and discrimination validity (Cronbach’s alpha, Kolmogorov-Smirnov test, Mann-Whitney U-test, and Correlation statistics).
  1. Step 4.

    Data Extraction and Codification

     
Data extraction was conducted independently by three authors (RBB, SS, CD) in order to build an electronic database. For codification purposes, the following detailed indications were considered:
  • Bayer Healthcare and GlaxoSmithKline merged together in 2001. A COI for one of the branches of the Company was considered as COI for the whole company;

  • an “Index Label” was defined as the commercial label of an active ingredient for which the authors expressed a supportive/not-supportive opinion. Any other PDEI(s) discussed in that publication were considered a Competitor(s) of the Index label;

  • only financial COIs were considered in the analysis (Yank, Rennie, and Bero 2007; Bartels, Delye, and Boogaarts 2012); the types of funding considered as financial COIs are reported in Table A (Supplemental electronic material);

  • expressed opinions were classified as “supportive” or “not-supportive” following the criteria detailed in Table B (Supplemental electronic material); neutral opinions were classified as “not-supportive” (Kjaergard and Als-Nielsen 2002; Stelfox et al. 1998; Bartels, Delye and Boogaarts 2012);

  • the total number of records does not equal the total number of publications because more than one index label (with a correlated supportive/not-supportive opinion) could be discussed in a single publication.

  • Quality Check of the Codification Process

Intra- and inter-observer reproducibility of the extraction and codification processes were tested by three authors (RBB, SS, CD) on the first fifty records listed in the database and repeated two weeks later using Cohen’s and Fleiss’ kappa.
  1. Step 5.

    Statistical Analyses

     

Statistics was performed by one author (RBB) using SPSS v. 17. A preliminary analysis was conducted checking for multicollinearity by analysis of tolerance and variance inflation factor, as they can affect the parameters of the regression model (Myers 1990; Menard 2001). A multivariable binary logistic regression model was performed with the expressed opinion as outcome (“supportive opinion” as reference) and independent variables to assess the impact of a number of factors on the likelihood that the expressed opinion would have been a supportive one.

Whenever a statistical result did not reach significance with a p-value close to the significant level, a power analysis to quantify the effective probability of rejecting a false null hypothesis was performed. A two-sided p<0.05 was chosen to indicate statistical significance in all the analyses, whilst p-levels between 0.05 and 0.10 were considered tendentially significant.

Results

A total of 1995 out of 3876 publications were retrieved in full text, of which 1745 (87.5 per cent) were excluded (Fig. 1). According to the inclusion criteria, 250 multi-authored publications (12.5 per cent) expressing a supportive/not-supportive opinion on the therapeutic efficacy/safety of Sildenafil and/or Tadalafil and/or Vardenafil for treating erectile dysfunction were included in the study (Fig. 1). The in-house developed categorization tool CoOpCaT (see Appendix 1 for details) was applied on fifty-seven randomly selected publications (57/250; 22.8 per cent) in order to verify the reproducibility and consistency of the categorization process. A good internal consistency for both the “COI” and the “Expressed opinion domains” (Cronbach’s alpha 0.62 and 0.66, respectively) was evidenced (i.e. these values are acceptable for research purposes) (Hair et al. 1998) with an inter-item correlation ranging from 0.16 to 0.51 and 0.21 to 0.60, respectively and with a corrected item-total correlation value above 0.40 and 0.32, respectively. A significant difference was observed for total COI domain scores of groups with/without COI (p<0.0001) and for total “expressed opinion” domain scores of groups “supportive/not-supportive” (p<0.0001) when testing the discriminative validity. The distribution of the scores across the groups showed no evidence of “floor” or “ceiling” effects (Table 1).
Fig. 1

Flow-chart of the systematic search process

Table 1

Expressed opinion in the records extracted from the included publications

  

Not supportive opinion

Supportive opinion

Total Records

p value

Existing Conflict (Index Label)

   

<0.001

 

Absent

118 (41.0%)

170 (59.0%)

288 (100%)

 
 

Present

9 (6.9%)

122 (93.1%)

131 (100%)

 
 

Unknown

0 (0.0%)

0 (0.0%)

0 (0.0%)

 

Year of Publication

   

<0.05

 

2003-2008

69 (35.6%)

125 (64.4%)

194 (100%)

 
 

2009-2013

58 (25.8%)

167 (74.2%)

225 (100%)

 

Type of Paper

   

0.15

 

Original article

78 (33.3%)

156 (66.7%)

234 (100%)

 
 

Review

42 (25.1%)

125 (74.9%)

167 (100%)

 
 

Letter/Editorial

7 (38.9%)

11 (61.1%)

18 (100%)

 

Number of Affiliations2

   

0.18

 

1

50 (32.9%)

102 (67.1%)

152 (100%)

 
 

2

30 (27.5%)

79 (72.5%)

109 (100%)

 
 

3

26 (41.9%)

36 (58.1%)

62 (100%)

 
 

4

9 (27.3%)

24 (72.7%)

33 (100%)

 
 

5

7 (23.3%)

23 (76.7%)

30 (100%)

 
 

>5

5 (15.2%)

28 (84.8%)

33 (100%)

 

Total Number of Conflicts (Index Label)

   

<0.001

 

0

118 (41.0%)

170 (59.0%)

288 (100%)

 
 

1

9 (11.8%)

67 (88.2%)

76 (100%)

 
 

2

0 (0.0%)

31 (100%)

31 (100%)

 
 

3

0 (0.0%)

15 (100%)

15 (100%)

 
 

4

0 (0.0%)

1 (100%)

1 (100%)

 
 

5

0 (0.0%)

4 (100%)

4 (100%)

 
 

>5

0 (0.0%)

4 (100%)

4 (100%)

 

First author's Conflict

   

<0.001

 

Absent

124 (34.9%)

231 (65.1%)

355 (100%)

 
 

Present

3 (4.7%)

61 (95.3%)

64 (100%)

 

Last author's Conflict

   

<0.001

 

Absent

123 (38.3%)

198 (61.7%)

321 (100%)

 
 

Present

6 (6.1%)

92 (93.9%)

98 (100%)

 

Total Number of Conflicts (Competitor/s)3

   

<0.01

 

0

106 (35.5%)

193 (64.5%)

299 (100%)

 
 

1

4 (14.3%)

24 (85.7%)

28 (100%)

 
 

2

11 (22.0%)

39 (78.0%)

50 (100%)

 
 

3

1 (5.3%)

18 (94.7%)

19 (100%)

 
 

4

3 (30.0%)

7 (70.0%)

10 (100%)

 
 

5

0 (0.0%)

1 (100%)

1 (100%)

 
 

>5

2 (16.7%)

10 (83.3%)

12 (100%)

 

Total Records

127

292

419

 

2A maximum of one affiliation was computed per capita.

3The label for which one author expressed an opinion was considered the reference to define the “Competitor(s)” label(s). Expressed opinion and COI were considered with regard to each active ingredient discussed in the publication (named index label). As a consequence, the sum of the records exceeds the number of included publications (250).

The agreement in the codification process evaluated on the first fifty records of the database was high (Table C—Supplemental electronic material) for both the intra-rater (kappa 0.77–0.95; p<0.001) and the inter-rater agreement (kappa 0.77–0.96; p<0.001), suggesting a lack of bias (Als-Nielsen et al. 2003; Kjaergard and Als-Nielsen 2002; Bero, Glantz, and Hong 2005).

The records extracted from the 250 included publications were 419. On univariate analysis, the prevalence of supportive opinions was higher for records reporting an existing COI for the “index label” (93.1 vs 59.0 per cent; p<0.001). No differences were found between groups of “supportive vs. not supportive” opinions with regard to the type of publication (i.e. original article, review, or letter/editorial; p=0.15) and the number of different affiliations involved in writing the publication (p=0.18).

Conflicts of interest were reported in 15.3 per cent (64/419) and 23.4 per cent (98/419) of the extracted records for first and last authors, respectively (Table 1). Supportive opinions were slightly more common when a first author’s conflict was present (95.3 vs. 65.1 per cent; p<0.001), compared to a last author’s conflict (93.9 vs. 61.7 per cent; p<0.001).

There was a statistically significant association between the total number of authors declaring a COI for the index label and a supportive opinion (p<0.01) expressed for that index label. The reduced model on logistic regression containing the selected predictors for a “supportive opinion” (i.e. the role(s) of the conflicted author(s) in the publication, the number of involved affiliations, the year and type of publication, the total number of COIs, and the existence of any counter-conflicts) was statistically significant (n= 419; χ2 = 85.93, p<0.001), indicating the ability to distinguish between records reflecting “supportive vs. not-supportive” opinions (Table 2). The model as a whole explained between 18.6 per cent (Cox and Snell R square) and 26.3 per cent (Nagelkerke R squared) of the variance in the opinions expressed, correctly classifying 68.8 per cent of the records. Only the independent variables considered made a unique statistically significant contribution to the model, with the “total number of COIs” for the index label being the strongest predictor of reporting a “supportive” opinion (OR 14.25 with 95 per cent CI 5.40–37.57).
Table 2

Binary logistic regression predicting the likelihood of reporting a supportive opinion for the records extracted from the included publications4

 

B

S.E.

Wald

p value

Odds Ratio

95% C.I. for

Odds Ratio

Lower

Upper

Year of Publication

0.14

0.04

10.54

0.001

1.15

1.06

1.25

Type of Paper

0.12

0.21

0.31

0.58

1.13

0.74

1.71

Number of Affiliations

0.11

0.07

2.52

0.11

1.12

0.97

1.28

Existing Conflict (Index Label)

-14.35

2502.4

0

0.99

0

0

0

First Author’s Conflict

-0.5

0.81

0.38

0.54

0.61

0.13

2.96

Last Author’s Conflict

-15.51

3016.1

0

0.99

0

0

0

Total Number of Conflicts (Index label)

2.66

0.5

28.85

0.001

14.25

5.41

37.57

Total Number of Conflicts (Competitor/s)

-0.43

0.18

5.51

0.019

0.65

0.46

0.93

4Numbers in bold refer to variables included in the final statistical model. B = intercept in the null model; S.E. = Standard Error; C.I. = Confidence Interval.

Discussion

When an individual’s judgement related to a primary professional interest is perceived to be influenced by a secondary interest, which could be financial or non-financial in nature, a conflict of interest exists (Hampson and Montie 2012). As is well-known and agreed upon within the scientific community, the simple existence of a relationship between the researcher and the industry is not necessarily improper in itself (Lo 2012); problems may arise depending on how the researcher reacts to the COI (Lexchin 2012b; Kassirer 2009).

Previous research has shown that company funded studies are more likely to lead positive outcomes in comparison to investigations without sponsorship (Lexchin 2012b). Financial COIs, in particular, have often been linked to more positive results in randomized controlled trials (Lundh et al. 2012; Als-Nielsen et al. 2003; Kjaergard and Als-Nielsen 2002; Friedman and Richter 2004). Additionally, it has been reported that financial ties can be associated with more favourable conclusions (OR 5.11 with 95 per cent CI 1.54–16.92) (Yank, Rennie, and Bero 2007) and/or to a promotional tone of the conclusions, as recently evidenced by Fugh-Berman and colleagues (Fugh-Berman et al. 2011).

In our investigation, on univariate analysis, the prevalence of “supportive opinions” was higher in the presence of a COI (95.3 per cent vs 65.1 per cent; p<0.001). We were not able, however, to determine whether the association between COIs and “supportive opinions” was causal. Indeed, we had no data for establishing if the supportive authors formulated their opinions after having had financial interactions with the drug company or if the pharmaceutical companies recruited physicians who had already expressed favourable opinions on their products in the past.

COIs were reported in 16.0 per cent and 23.5 per cent of records for first and last authors, respectively. These data are in line with the literature, although Reichelmann reported a slightly higher prevalence for first authors’ COIs (33 per cent) (Riechelmann et al. 2007). Surprisingly, we found no clear correlation between the number of affiliations involved in writing the publication and the type of opinion expressed. This datum underscored the lack of any hypothetical protective effects of having a high number of co-authors coming from different institutions when expressing a supportive opinion in the presence of a financial COI.

The strongest predictor for a “supportive opinion” was the “total number of financial COIs” for the index label (OR 14.25 with 95 per cent CI 5.40–37.57; Table 2). This means that increasing the number of co-authors who declared a COI for the index label, the probability of having a supportive opinion on the efficacy/safety of that drug increased fourteen times.

On the other hand, the odds ratio for the total number of financial COIs for competitor label(s) (i.e. counter-conflicts) was 0.65 (95 per cent CI 0.46–0.93); this means that for each additional counter-conflict declared by a co-author the type of opinion expressed in the publication was 0.65 times less likely to be a supportive one with regard to that index label for which a counter-conflict exists. This finding, is not consistent with Stelfox et al.’s observation that authors who criticized calcium-channel antagonists were not more likely to have any financial relationships with the competing manufacturers (Stelfox et al. 1998) but is absolutely in line with more recent evidence (Pham-Kanter 2014). Pham-Kanter showed that members of the Food and Drug Administration (FDA) Advisory Committee with financial ties to both the sponsoring firm and the competitors were no more likely to vote in favour of the sponsor than those with no financial ties to any potentially affected firm (OR = 1.16, 95 per cent CI 0.77–1.76; p = 0.48). Multiple financial ties binding a single Advisory Committee Member to sponsor and competitors exerted a sort of “protective effect” with regard to the financial COI for the sponsor. In our investigation a similar “protective effect” was generated by the presence of a co-author of the manuscript having a declared financial tie to one of the three competitor manufacturers of drugs for treating erectile dysfunction.

Until now, several strategies have been proposed for addressing and managing fCOIs, with disclosure being probably the oldest and the most utilized one. However, disclosure is not a panacea and may not, on its own, resolve conflicts of interest (Kassirer 2009). The problem, indeed, is “bias” (conscious or more often subconscious), which can derive from financial ties. Studies by cognitive psychologists have shown that people are usually unaware of their biases, and that self-interest distorts judgement (Kassirer 2009). Several authors have argued that open disclosure could even increase the bias in advice, because it leads advisors to feel morally licenced and to exaggerate their advice even further (Cain, Loewenstein, and Moore 2005) and that therefore the ideal solution would be to eliminate any financial conflict of interests altogether (Kassirer 2009).

Other scholars argue that establishing a zero tolerance policy might be unwise due to the huge administrative burdens associated with policing conflicts of interest and the related potential adverse effects (Stossel 2005). Not to mention that manufacturers usually look for leading-experts in the field not only to provide support in knowledge dissemination and advertising events, but also for research purposes—aware that they are often also leading writers. Indeed, to identify an author with the necessary expertise and experience for performing a sound scientific investigation that is absent of COIs can be a difficult (although not impossible) task for editors. Having said this, however, it should be noted that Lenzer and Brownlee succeeded in compiling a list of nearly one hundred medical experts who had not received any financial support from pharmaceutical or medical device manufacturers during the past five years (Lenzer and Brownlee 2008). Furthermore, until 2002 high profile medical Journals such as the New England Journal of Medicine used specific disclosure policies for Reviews and Editorials which excluded authors who declared significant financial interests in any biomedical companies relevant to the topics and products discussed in the article.

Although the definitive solution to managing bias resulting from financial COIs seems far away, we believe that a transparent, compulsory, and standardized disclosure system, such as the one developed by Rochon et al. in 2010, should be the starting point. Rochon et al. propose a “Financial Conflicts of Interest Checklist” consisting of a glossary and four sections covering administrative, study, personal financial, and authorship information (with a total of fifteen items and related sub-items to be checked) to be completed by all the investigators at different points over the course of the study (Rochon et al. 2010), including the possibility to append updated information prior to the submission to stakeholders groups.

We believe that disclosure should be part of a coordinated policy that sets high standards of ethical conduct, clearly delineates the permissible from the unacceptable, develops institutions to monitor behaviour, and imposes meaningful sanctions to ensure compliance (Kassirer 2009). It would also be useful to place the disclosure in the abstract in order to increase the awareness of the reader on the existence of a financial COI (Kesselheim et al. 2012), to make raw data available to research participants and readers, and to strengthen the peer-review process with a validation of the data by a statistician unrelated to the industrial company.

Moreover, as reported by Pham-Kanter and shown by our results (although a validation in a wider clinical setting is needed), the promotion of multicentre studies with the inclusion of co-authors having a financial relationship with a competitor (i.e. conter-conflict) could be an interesting alternative tool to be explored for reducing potential bias deriving from financial COIs.

Limits

The results of this study need to be interpreted within the context of the following limitations.

The literature search was conducted on a single database (i.e. Scopus). We decided to include only full-text publications, aware of the fact that having a limited access to full-texts could have influenced the analysis. Whether the reported opinion was correctly based on the validity of the results of each publication was not verified in this study. Our data rely upon self-reports of any relevant potential COI by authors. As a consequence, we did not validate these data, obtaining personal financial information from anyone. The categorization of supportive/not-supportive opinions and presence/absence of COIs was not performed in blind. We focused on financial COIs, but we are aware that non-financial COIs could be equally important. Non-financial COIs are much more difficult to declare and clearly detect, so we decided to focus on a less uncertain and more easily measurable variable. Within the category of financial COIs we were not able to weigh the different types of conflict (in particular regarding the amount of money received by the authors). We were also unable to determine the temporal relation between authors’ published opinions and their COI settlement. Hence, authors may have formulated their opinions after having financial interactions with drug manufacturers, or pharmaceutical companies may have sought relationships with clinicians and researchers who had already expressed favourable opinions of their products. Moreover, our results do not identify the cause of the association between financial COIs and author opinions. Finally, we rarely found reports of COIs involving the authors’ institutions, but we cannot exclude that some kind of disclosure was collected by journal editors, even if not subsequently published.

Conclusions

Our analysis demonstrated that a positive association exists between a financial COI declared by at least one author and a “supportive opinion” defined as a conclusion that supports the efficacy/safety of the index label for which the author declared the COI. We were not able to determine whether this association was causal or not. The strongest predictor for a “supportive opinion” was the total number of authors with COIs for the index label whereas a mild “protective effect” was found when a co-author declared COIs for competitors (i.e. a counter-conflict). If this finding is confirmed in a wider clinical scenario, it could be a potential adjunctive tool for limiting bias deriving from financial ties.

Notes

Acknowledgement

The authors wish to thank Thomas Dewis who assisted in the language editing of the manuscript.

Supplementary material

11673_2016_9732_MOESM1_ESM.docx (25 kb)
ESM 1(DOCX 24 kb)

References

  1. Als-Nielsen, B., W. Chen, C. Gluud, and L.L. Kjaergard. 2003. Association of funding and conclusions in randomized drug trials: A reflection of treatment effect or adverse events? JAMA 290(7): 921–928.CrossRefPubMedGoogle Scholar
  2. American Academy of Dermatology Board of Directors. 2008. Position statement on contemporary issues: Conflict of interest. Journal of the American Academy of Dermatology 59(6): 1005–1008.CrossRefGoogle Scholar
  3. Bartels, R.H., H. Delye, and J. Boogaarts. 2012. Financial disclosures of authors involved in spine research: An underestimated source of bias. European Spine Journal 21(7): 1229–1233.CrossRefPubMedGoogle Scholar
  4. Bero, L.A., S. Glantz, and M.K. Hong. 2005. The limits of competing interest disclosures. Tobacco Control 14(2): 118–126.PubMedPubMedCentralGoogle Scholar
  5. Cain, D.M., G. Loewenstein, and D.A. Moore. 2005. The dirt on coming clean: Perverse effects of disclosing conflicts of interest. Journal of Legal Studies 34(1): 1–25.CrossRefGoogle Scholar
  6. Friedman, L.S., and E.D. Richter. 2004. Relationship between conflicts of interest and research results. Journal of General Internal Medicine 19(1): 51–56.CrossRefPubMedPubMedCentralGoogle Scholar
  7. Fugh-Berman, A., C.P. McDonald, A.M. Bell, E.C. Bethards, and A.R. Scialli. 2011. Promotional tone in reviews of menopausal hormone therapy after the Women’s Health Initiative: An analysis of published articles. PLoS Medicine 8(3): e1000425.CrossRefPubMedPubMedCentralGoogle Scholar
  8. Hair, J., R. Anderson, R. Tatham, and W. Black. 1998. Multivariate data analysis, 5th ed. New Jersey: Prentice-Hall.Google Scholar
  9. Hampson, L.A., and J.E. Montie. 2012. Conflict of interest in urology. Journal of Urology 187(6): 1971–1977.CrossRefPubMedGoogle Scholar
  10. Institute of Medicine (US) Committee on Conflict of Interest in Medical Research Education and Practice. 2009. Conflict of interest in medical research, education, and practice. Washington (DC): National Academies Press (US).Google Scholar
  11. Kassirer, J.P. 2009. Medicine’s obsession with disclosure of financial conflicts: Fixing the wrong problem. In Science and the media Delgado’s brave bulls and the ethics of scientific disclosure, ed. P.J. Snyder, L.C. Mayes, and D.D. Spencer, 79–89. London, UK: Academic press.Google Scholar
  12. Kesselheim, A.S., C.T. Robertson, J.A. Myers, et al. 2012. A randomized study of how physicians interpret research funding disclosures. New England Journal of Medicine 367(12): 1119–1127.CrossRefPubMedPubMedCentralGoogle Scholar
  13. Kjaergard, L.L., and B. Als-Nielsen. 2002. Association between competing interests and authors’ conclusions: Epidemiological study of randomised clinical trials published in the BMJ. BMJ 325(7358): 249.CrossRefPubMedPubMedCentralGoogle Scholar
  14. Lee, K. 2008. Has the hunt for conflicts of interest gone too far? No. BMJ 336(7642): 477.CrossRefPubMedPubMedCentralGoogle Scholar
  15. Lenzer, J., and S. Brownlee. 2008. Naming names: Is there an (unbiased) doctor in the house? BMJ 337: a930.CrossRefPubMedGoogle Scholar
  16. Lexchin, J. 2012. Sponsorship bias in clinical research. International Journal of Risk & Safety in Medicine 24(4): 233–242.Google Scholar
  17. ———. 2012b. Those who have the gold make the evidence: How the pharmaceutical industry biases the outcomes of clinical trials of medications. Science and Engineering Ethics 18(2): 247–261.Google Scholar
  18. Lo, B. 2012. The future of conflicts of interest: A call for professional standards. Journal of Law and Medical Ethics 40(3): 441–451.CrossRefGoogle Scholar
  19. Lundh, A., S. Sismondo, J. Lexchin, O.A. Busuioc, and L. Bero. 2012. Industry sponsorship and research outcome. Cochrane Database of Systematic Reviews 12(12): 1–61.Google Scholar
  20. Maj, M. 2008. Non-financial conflicts of interests in psychiatric research and practice. British Journal of Psychiatry 193(2): 91–92.CrossRefPubMedGoogle Scholar
  21. Menard, S. 2001. Applied logistic regression analysis. Sage university series on quantitative applications in the social sciences. Thousand Oaks, CA: Sage Publisher.Google Scholar
  22. Myers, R.H. 1990. Classical and modern regression with applications. Pacific Grove, CA: Duxbury/Thompson Learning Publisher.Google Scholar
  23. Pham-Kanter, G. 2014. Revisiting financial conflicts of interest in FDA advisory committees. Milbank Quarterly 92(3): 446–470.CrossRefPubMedPubMedCentralGoogle Scholar
  24. Riechelmann, R.P., L. Wang, A. O’Carroll, and M.K. Krzyzanowska. 2007. Disclosure of conflicts of interest by authors of clinical trials and editorials in oncology. Journal of Clinical Oncology 25(29): 4642–4647.CrossRefPubMedGoogle Scholar
  25. Robertson, C., S. Rose, and A.S. Kesselheim. 2012. Effect of financial relationships on the behaviors of health care professionals: A review of the evidence. Journal of Law and Medical Ethics 40(3): 452–466.CrossRefGoogle Scholar
  26. Rochon, P.A., J. Hoey, A.W. Chan, et al. 2010. Financial conflicts of interest checklist 2010 for clinical research studies. Open Medicine 4(1): e69–91.PubMedPubMedCentralGoogle Scholar
  27. Roseman, M., K. Milette, L.A. Bero, et al. 2011. Reporting of conflicts of interest in meta-analyses of trials of pharmacological treatments. JAMA 305(10): 1008–1017.CrossRefPubMedGoogle Scholar
  28. Saver, R.S. 2012. Is it really all about the money? Reconsidering non-financial interests in medical research. Journal of Law and Medical Ethics 40(3): 467–481.CrossRefGoogle Scholar
  29. Stelfox, H.T., G. Chua, K. O’Rourke, and A.S. Detsky. 1998. Conflict of interest in the debate over calcium-channel antagonists. New England Journal of Medicine 338(2): 101–106.CrossRefPubMedGoogle Scholar
  30. Stossel, T.P. 2005. Regulating academic-industrial research relationships: Solving problems or stifling progress? New England Journal of Medicine 353(10): 1060–1065.CrossRefPubMedGoogle Scholar
  31. ———. 2008. Has the hunt for conflicts of interest gone too far? Yes. BMJ 336(7642): 476.Google Scholar
  32. Wang, A.T., C.P. McCoy, M.H. Murad, and V.M. Montori. 2010. Association between industry affiliation and position on cardiovascular risk with rosiglitazone: Cross sectional systematic review. BMJ 340: c1344.CrossRefPubMedPubMedCentralGoogle Scholar
  33. Yank, V., D. Rennie, and L.A. Bero. 2007. Financial ties and concordance between results and conclusions in meta-analyses: Retrospective cohort study. BMJ 335(7631): 1202–1205.CrossRefPubMedPubMedCentralGoogle Scholar

Copyright information

© Journal of Bioethical Inquiry Pty Ltd. 2016

Authors and Affiliations

  • Rafael Boscolo-Berto
    • 1
  • Massimo Montisci
    • 1
  • Silvia Secco
    • 2
  • Carolina D’Elia
    • 3
  • Rosella Snenghi
    • 1
  • Guido Viel
    • 1
  • Santo Davide Ferrara
    • 1
  1. 1.Department of Cardiac, Thoracic and Vascular Sciences, Section of Legal MedicineUniversity Hospital of PadovaPadovaItaly
  2. 2.Department of UrologyOspedale Niguarda Cà GrandaMilanoItaly
  3. 3.Department of UrologyUniversity of VeronaVeronaItaly

Personalised recommendations