Abstract' Content and GPs' Confidence in the Conclusion

Overview

Objective: The aim of this study is to assess the impact of reporting, in abstracts of randomized controlled trials, 1) the funding sources and 2) the conflicts of interest (COIs) of the authors on General Practitioners' confidence in the abstracts' conclusions. Design: Randomized Controlled Trial in 3 arms. Participants: General Practitioners (GP). They will be recruited among residents from the Primary Care department of the Paris Descartes University and among a GP network who agreed to participate to research studies. The allocation of participants will be done by a computerized randomization list, the sequence will be created by an independent statistician with a ratio 1: 1: 1. Participants will be blind of the hypothesis and of the randomization. Allocation concealment will be provided because only the statistician will have access to the randomization list. Methods: Three interventions will be assessed corresponding to three different types of abstracts: 1) abstracts not reporting neither the funding sources neither COIs of authors, 2) abstracts reporting funding sources and 3) abstracts reporting funding sources and COIs of authors. A sample of abstracts will be selected from published randomized trials, testing superiority, assessing pharmacological treatment, in the field of Primary Care and General Practice and having a conclusion in favor of the beneficial effect of experimental treatment in terms of tolerance and / or efficacy. They will then be standardized and modified to obtain three types of abstracts: 1) with funding sources, 2) with funding sources and COIs and 3) without either source of funding or COIs. Participants will be asked to assess one abstract of their randomization arm. Outcome: The primary endpoint will be the GPs' confidence in conclusions. The secondary endpoints will be the quality perception of the study and the interpretation of the benefit of experimental treatment. Responses will be ranged by a 10-point numeric scale. Potential interests: This study takes place in the field of Primary Care. We believe that better understandings of impact of funding sources and conflicts of interests are necessary to allow more transparency in medical research and in his translation into medical practice. Sample size expected: 354 participants.

Dissemination of knowledge and research results is an essential step in the practice of Evidence-Based Medicine (EBM) (1).With the growing number of publications, a lack of accessibility, and the need for timely messages among information resources larger and larger, the General Practitioners (GP) often rely on abstracts of published trials (2). Several methodological studies have shown an association between mode of financing, including industry funding, and how the results and conclusions of published trials are reported (3-14). To improve transparency, the CONSORT recommends to report funding sources (15) in abstracts of randomized trials, which is not always followed. Efforts are also made to report conflicts of interest of authors (COIs). In its version in July 2010, the International Committee of Medical Journal Editors Uniform Disclosure Form for Potential Conflicts of Interest (ICMJE) recommends that editors ask to all authors who submit a manuscript, to complete the ICMJE statement (16,17). Despite this, understanding the complexity and severity of conflicts of interest is not yet fully taken into account by all the guidelines (18). It was further demonstrated that the interpretation of results by physicians varies with their mode of presentation (19-22). The investigators do not know to what extent the knowledge of funding sources and COIs may influence the interpretation of findings by the reader. The investigators are therefore interested in the impact of the presence of these elements in abstracts of published trials on GPs' confidence in the conclusions presented. The aim of this study is to assess the impact of reporting, in abstracts of randomized controlled trials, 1) the funding sources and 2) the conflicts of interest (COIs) of the authors on General Practitioners' confidence in the abstracts' conclusions. STUDY DESIGN The investigators plan a 3 arms randomized parallel-group, conducted among GPs residents in Primary care and planned according to the CONSORT 2010 guidelines (23). For allocation of the participants, a computer-generated list of random numbers will be used. Randomization sequence will be created by an independent statistician with R project with 1:1:1 ratio using random block sizes of 2, 4, and 6. The statistician will upload the randomization list on a secure internet system to assure allocation concealment. GPs will be invited by e-mail to participate to this study. They will log on a secure internet system that will randomize the GP in 3 different arms: 1) abstracts not reporting neither funding neither COIs, 2) abstract reporting funding source and 3) abstract with funding and COIs. Blinding participants is not possible in this study. To reduce the risk of bias, physicians will not be informed of the objectives of the study (ie study of the impact of the presentation of funding sources and COIS), or the existence of randomization. They will blind of the working hypotheses of the study. PARTICIPANTS GPs will be recruited among a network of GPs in France: this network consists of GPs who agreed to participate in research studies. It brings together members from the Collège Nationale des Généralistes Enseignants (CNGE) and GPs recruited by regional relays represented by 10 physicians members of primary Care departments. INTERVENTION Three interventions will be evaluated corresponding to 3 different types of abstracts: 1) abstracts not reporting the sources of financing or the COIs, 2) abstracts reporting the sources of funding as recommended by CONSORT for abstracts and 3) abstracts related funding sources and COIs. Abstracts identification: To allow reproducibility and generalization of our results, abstracts will be selected from randomized controlled trials (RCTs) published in field of Primary Care. The investigators chose to select abstracts of RCTs because their contents are standardized because of the CONSORT guidelines. The investigators also chose to retain only the positive trials (conclusion in favor of the beneficial effect of experimental treatment in terms of tolerance and / or efficacy) to have a uniform interpretation of the results. To identify abstracts, the investigators will search in Medline via PubMed for all indexed abstracts with using limits for randomized controlled trials, in English, published from the 1srt January 2010 to the 31 December 2010, in the Core Clinical Journal. The Core Clinical Journal was chosen to ensure a good quality of trials. The investigators restricted the search to the year 2010 to ensure that physicians do not remember the original abstract and thus preserve the blinding. The eligibility criteria will be: randomized controlled trials, testing superiority, assessing pharmaceutical intervention (drugs) in the field of Primary Care (i.e. trials assessing treatments that may be prescribed by a GP, and those assessing treatments being prescribed in pathologies supported jointly between specialist and GP.), with a positive conclusion and funded only by the pharmaceutical industry. Exclusion criteria will be: the non-pharmaceutical interventions (eg medical devices, patient education, etc..), equivalence or noninferiority trials, safety trials, trials assessing different pharmacological procedures, abstracts reporting a negative conclusion, not reporting or reporting the results in an ambiguous conclusion (successful trial but related the need for other studies), intermediate publications. Construction of the modified Abstracts: Standardization: All abstracts will be standardized: the journal, date and registration number of ClinicalTrial.gov will be deleted. Authors' names will be replaced by generic names. A list of author names will be determined by drawing 20 names among the 200 names most common in the UK and will be identical for all abstracts. The treatments will be referred to by "experimental treatment A" or "treatment B" in the abstract and in the title. The names of studies as an acronym will be deleted. Finally, abstracts will be translated into French. Modifications: The three types of abstracts will be presented as follows: 1. Abstracts not reporting funding or COIs If these items are reported in the original abstract, they will be removed. 2. Abstracts reporting funding Funding sources will be searched in abstract reporting by the editors and in the full text. If funding sources are not found, abstract won't be retained. Original funding will be reported by the name of pharmaceutical industry, at the end of the abstract, after conclusion, under the heading "FUNDING". 3. Abstracts reporting funding and COIs Funding sources will be reported as presented previously. COIs will be searched in abstract reporting by the editors and in the full text. If they are not found, abstract won't be retained. The COIs will then be ranked (1) no COIs, (2) authors employed by industry and (3) authors who have a financial relationship with industry. Financial ties regroup together all COIs other than being employed: fees, travel costs, as these are to be returned in a personal or institution with which the author works, stock options, link with a family member employed by the industry, etc.. For each abstract, the investigators will retain the number of authors having a conflict. The COIs will be given in the modified abstract by taking the initials of the authors and their type of COIs, presented in two categories: "financial or other relationship" and "being employee" under the heading "CONFLICT OF INTEREST". Note that the investigators will consider in the same way "no conflict of interest" and the authors who do not report their COIs. We won't report it in the abstract. PROCEDURES Participants will be invited to participate in a study evaluating the interpretation of abstracts of RCTs by e-mail (with 3 reminders) or by an investigator during the national annual medical congress for general practice. They will be informed that their participation will take less than 5 minutes. After accepting, physicians will complete a questionnaire to collect their characteristics: date of birth, sex, status (resident or GP). Then they will be randomized and start immediately read the abstract and questionnaire. Participants will be asked to evaluate only a abstract in his arm of randomization. SAMPLE SIZE The investigators estimate that 75 abstracts will be available to be assessed. Three main comparisons will be made: a comparison with the placebo group related funding sources, the comparison with the placebo group related funding sources and COIS and finally the comparison group of related funding sources with the related funding sources and COIS. With a significance level of 1.67% fixed for each of the three comparisons 2-2 (Bonferonni correction to maintain an overall significance level of 5%), it is necessary to have about 118 quotations in each arm to demonstrate an effect size equal to 0.5 on the numeric scale with a power of 90% for each comparison 2-2. As it is expected that each physician read a single abstract, it is necessary to include 118 physicians per arm (354 in total, each abstract is read 4.7 times on average (4 or 5 times in practice)). STATISTICAL ANALYSIS Data analysis will be done with the software R project, in intention to treat. A description of the characteristics of physicians will be made initially. Qualitative variables will be described by their numbers, percentage and missing data by mode of response and quantitative variables by their rate, mean, standard deviation. In case of quantitative variables with asymmetric behavior, they will be presented with their median and interquartile range (25th percentile-75th percentile). The primary analysis will be done on mean of GPs' confidence and the 3 possible comparisons will be made. The primary endpoint is quantitative, the test used for comparisons 2to 2 will be the non-parametric Mann Whitney 1.7% threshold. The same analysis is repeated for the secondary endpoints quantitative.