Article Text

PDF
Failure to report and provide commentary on research ethics board approval and informed consent in medical journals
  1. K A Finlay1,
  2. C V Fernandez2
  1. 1
    Dalhousie University, Halifax, Nova Scotia, Canada
  2. 2
    Departments of Pediatrics and Bioethics, Dalhousie University and IWK Health Centre, Halifax, Nova Scotia, Canada
  1. Dr C V Fernandez, Departments of Pediatrics and Bioethics, Dalhousie University and IWK Health Centre, 5850 University Avenue PO Box 9700, Halifax, Nova Scotia, B3K 6R8 Canada; Conrad.Fernandez{at}iwk.nshealth.ca

Abstract

Background: The Declaration of Helsinki prohibits the publication of articles that do not meet defined ethical standards for reporting of research ethics board (REB) approval and informed consent. Despite this prohibition and a call to highlight the deficiency for the reader, articles with potential ethical shortcomings continue to be published.

Objective: To determine what proportion of articles in major medical journals lack statements confirming REB approval and informed consent, and whether accompanying commentary alerts readers to this deficiency.

Design: Retrospective, observational study.

Setting: Online review of five major medical journals.

Population: All clinical research articles published online between 1 January 2005 and 31 December 2006 in the BMJ, Lancet, Annals of Internal Medicine, JAMA and the New England Journal of Medicine.

Measurements: Statement of REB approval and informed consent.

Results: Of 1780 articles reviewed, 1133 (63.7%) met inclusion criteria (manuscripts reporting human subjects, human tissue or identifiable personal data research), 36 (3.2%) articles lacked a statement of REB approval, 62 (5.5%) lacked disclosure of informed consent and 15 (1.3%) articles lacked both. Articles that did not state REB approval were associated with not stating informed consent (p<0.001). There were no editorial comments to alert readers to the lack of either REB approval or informed consent statements associated with any of the deficient articles.

Conclusions: Articles that lack explicit statements of REB approval and informed consent are infrequent but continue to be published in major medical journals without editorial statements to alert the reader to this deficiency.

Statistics from Altmetric.com

The Declaration of Helsinki states that medical research involving human subjects must have approval from an independent ethics committee and, when appropriate, informed consent from study participants prior to the start of a study. Further, the Declaration stipulates that manuscripts that do not meet these principles “should not be accepted for publication” (article 27).1 Smaller studies have suggested that not all published articles explicitly report informed consent or institutional review board or research ethics board (REB) approval.26 It has been proposed that if research does not clearly follow ethical standards and is accepted for publication on scientific merit, it should be accompanied by an editorial or comment highlighting the ethical questions raised.7 8 This procedure has been followed in several instances711 and has enabled education of the reader and provided an opportunity for vigorous and public debate. Conflict of interest disclosure has been widely discussed, with most organisations now providing clear guidelines to members. No comprehensive survey of journal practice has examined the extent to which lack of reporting of essential ethical elements in manuscripts occurs.

This study investigates whether the publication practices of five major medical journals currently confirm that human participants are protected through ethics committee oversight and consent processes. Current rates of documenting conflict of interest are also examined to determine whether recent changes in guidelines to authors have resulted in improved reporting. The primary objective was to determine the frequency of reported REB review and informed consent and whether lack of reporting is accompanied by commentary revealing the deficiency.

METHODS

All clinical research articles published online between 1 January 2005 and 31 December 2006 in the BMJ (International edition), Lancet (British edition), Annals of Internal Medicine (Annals), the JAMA and the New England Journal of Medicine (NEJM) were retrospectively reviewed. The entire manuscript of all research articles involving human subjects, human tissues or identifiable data was examined for statements of informed consent and research ethics board (REB) approval. Statements clarifying that an ethics organisation had approved the study and that subjects had consented to participation, such as “written consent was obtained” or “informed consent was given,” were required for both criteria to be met. The return of a questionnaire was accepted as implied consent. For statistical purposes, articles that cited accordance with the Declaration of Helsinki or that requirements for REB approval or consent were waived were considered to have documented REB approval or consent. A second reviewer (CF) then independently searched all articles catalogued as having neither REB approval nor informed consent. In addition, the “search” function of Adobe Acrobat was utilised by both reviewers to search for the keywords “institutional”, “ethics”, “review”, “committee”, “consent” and “approval” to ensure statements were not missed. If the article referred the reader to a previous paper for complete methodology, that prior article was retrieved through PubMed Central or “Highwire” (http://highwire.stanford.edu/) and examined for the same criteria. If that study could not be accessed, a lack of a statement of REB approval or informed consent was recorded. If an article lacked statement of REB approval or informed consent, all editorials, letters and related articles linked in PubMed and Highwire were searched for a statement alerting the reader to this deficiency. Journal articles were also assessed for whether conflict of interest statements were declared. It was documented whether a specific conflict of interest statement or a general comment of financial support was published.

In all, 647 articles were excluded. Letters (7/647), literature reviews (58/647), meta-analyses (59/647), case studies or reports (23/647) and retrospective cohort studies (162/647) were excluded, as were abridged reports that did not have a complete methods section (132/647) and studies that reported data, such as census results, that were in the public domain. Ethics approval was not obtained for this study as it was a retrospective review of non-human data.

Data sources and study variables

Journal issues were retrieved via online access. Full-text articles were inspected in all cases rather than abridged versions. Information was recorded on how journals dealt with each article lacking REB approval or informed consent, and whether it was through an editorial, some other form of commentary or no action. Each journal’s written instructions to authors regarding these issues were also noted. The primary outcome measures were the overall rates of reporting informed consent and REB approval, as well as how each journal dealt with articles that did not fully meet the ethical standards set by that journal and the Declaration of Helsinki. Reporting of conflict of interest and differences in the rates of reporting between journals and types of articles in 2005 and 2006 were examined as secondary outcome measures.

Statistical analysis

Data were analysed using SPSS V.13.0. Data are reported by descriptive techniques. χ2 Analysis was used to compare the subsequently categorised primary and secondary categorical outcome variables for each journal studied. A p value of less than 0.05 was considered statistically significant.

RESULTS

A total of 1780 articles were reviewed from the five journals’ online publications for the period 1 January 2005 to 31 December 2006. Of these, 1133 (63.7%) met inclusion criteria. Randomised controlled trials (RCTs) comprised 50.6% of included articles, while prospective studies made up 25.7%. The remaining 23.7% of articles included surveys, cross-sectional observational, epidemiological and genetic screening studies. The overall proportion of RCTs published increased from 2005 to 2006 (275/917 vs 304/863; p = 0.018). The proportion of articles included in the study significantly increased between 2005 and 2006 (557/917 vs 576/863; p = 0.009).

Table 1 shows the proportion of included articles for each journal for each year of the study and the total number of articles included from each journal overall. Annals had the fewest articles included. There was a difference in the proportion of articles included from the BMJ, with fewer RCTs, prospective studies and surveys published in 2005 than in 2006 (91/202 vs 128/203; p<0.001).

Table 1 The proportion of included articles for each journal by year

Primary outcomes

All five journals had criteria for informed consent, REB approval and conflict of interest statements in their instructions to authors (accessed March 2007). These instructions were essentially unchanged for the years of the study. Annals ask authors to state in their methods whether an institution review board approved the study or if the protocol was consistent with the Declaration of Helsinki, as well as whether participants gave informed consent. The BMJ requires details of ethics approval for all original research articles but did not require explicit statement about informed consent for clinical trials. The Lancet outlines what type of study requires REB approval and informed consent and states that each should be documented in the article. JAMA states that for human experimental investigations, review by an ethics committee is required and should be documented in the methods section along with the “manner in which informed consent was obtained from the study participants”. The NEJM asks authors to provide statements in the manuscript that the protocol was approved by an ethics committee and that written informed consent was obtained.

Overall, 36 articles (3.2%) did not report REB approval and 62 (5.5%) did not document informed consent. There were 15 (1.3%) articles missing both statements. In addition, of the articles considered to have sufficient disclosure of REB approval, 28 (2.5%) required the reader to find a prior published article. Similarly, for articles considered to have obtained informed consent, 47 (4.1%) required returning to a previous article to find a statement that consent was obtained. One (0.1%) article had REB approval waived, 18 (1.6%) had requirements of consent waived and eight (0.7%) cited accordance with the Declaration of Helsinki in place of a statement of either REB approval or consent. All of these papers were considered to have documented REB approval or consent for statistical purposes. Five papers (0.4%), which we had classified as human research needing ethics oversight per the Declaration of Helsinki, reported that they did not seek REB review according to local standards, or were told by local REBs that their research did not require review. For six articles, either REB approval, informed consent, or both, could not be confirmed as the previous articles containing full methods could not be accessed through PubMed or Highwire.

There were no linked editorials, letters or related articles that we could identify in PubMed or on Highwire that discussed the lack of disclosure of REB approval or informed consent for any of the 83 articles found to lack one or both of these elements.

Secondary outcomes

Two articles (0.2%) did not have any statement of conflict of interest and both were in articles from the NEJM in 2005. However, 77 (6.8%) of the articles that were recorded to have some form of conflict of interest statement had only a statement of the funding body rather than describing explicitly whether there were any other conflicts of interest.

Reporting of REB approval improved from 2005 to 2006 (table 2). Of the articles lacking REB approval, 72.2% were published in 2005. There was no significant difference between 2005 and 2006 in the overall reporting of informed consent or conflict of interest. This did not appear to be related to a change in journal instructions to authors.

Table 2 Fraction of included studies lacking statements of REB approval, informed consent or conflict of interest for each year

The overall reporting of REB approval, informed consent and conflict of interest statements were compared for each journal (table 3). There was a significant difference between journals in reporting of REB approval, informed consent. JAMA had the highest rates of reporting both REB approval and informed consent, while Annals had the lowest rate of reporting REB approval. The BMJ had the highest rate of failing to report that adequate informed consent procedures were met. Of all articles lacking statement of informed consent, 41.2% were published in the BMJ.

Table 3 Fraction of included studies lacking statements of REB approval, informed consent or conflict of interest for each journal

Table 4 shows a comparison for each journal for each year of the proportion of studies lacking statements of REB approval or informed consent. Reporting of REB approval improved significantly and reporting of consent approached significance in the NEJM from 2005 to 2006. There was also an increase in the proportion of RCTs, prospective studies and surveys from 2005 (133/163 (81.6%)) to 2006 (160/166 (96.4%); p<0.001) in the NEJM. Reporting of REB approval showed a trend towards improvement in Annals from 2005 to 2006.

Table 4 Fraction of included studies that did not have statements of REB approval and that did not document consent for each journal for 2005 and 2006

RCTs examined in our study were less likely to be missing statements of REB approval (10/573 vs 26/560; p = 0.005) or informed consent (19/573 vs 43/560; p = 0.001) when compared with all other included study types. When compared with a combined figure of prospective studies and surveys, RCTs were more likely to have statements of REB approval (10/573 vs 21/472; p = 0.010) and consent (19/573 vs 35/472; p = 0.003). In all, 36% of articles lacking statements of REB approval and 45% of articles lacking statement of consent were prospective studies. Articles that lacked statement of REB approval were more likely to be lacking statement of informed consent (p<0.001).

DISCUSSION

Leading medical journals failed to report REB approval in 3.2% of articles and informed consent in 5.5%, and 1.3% lacked both items. Although uncommon, articles that do not report REB approval or informed consent are of concern as there is no public documentation that the trial followed broadly required ethical standards. Potentially, this both challenges public trust in research and underlines a lack of respect for the participants of such trials. The rate of reporting of REB approval and informed consent appears consistent with the detail of each of the journals’ instructions to authors. JAMA had the best rate of reporting and the most explicit requirement for documentation of REB approval and consent, including how each was obtained. The absence of any linked commentary in PubMed or Highwire for the articles missing statements of REB approval or informed consent in this study is disquieting as the reader is not alerted to these deficiencies in any way. It is also a lost opportunity to provide public debate about ethical conduct that may fall into a grey area that is difficult to classify. While it may be possible that the lack of remarks from editors or peers is due to these articles meeting ethical standards and simply not reporting them, it is suboptimal to allow the reader to assume this.

An examination by Schroter et al12 of 370 sequential articles published by the five journals in 2003 found that 27% of original research failed to report both approval and consent and that the BMJ had the lowest rate of reporting REB approval. While we also found that the BMJ had a lower rate of reporting, the overall rate of reporting in our study was much higher than found by Schroter. In contrast to our methodology, the study by Schroter et al included retrospective studies, case reports and also counted all articles which referred the reader to a prior article for REB approval and consent statements in their analysis, which may have contributed to a higher percent of non-reporters in each journal. We did not include retrospective studies or case reports because the need for research ethics committee review may vary by jurisdiction. In addition, case reports represented only a small fraction of the overall total. Schroter did not examine previous publications to which the current article referred for a statement of ethics committee review or consent and this may have significantly inflated their percentage of non-reporters. We feel our methods give a more accurate measure of what the investigators have actually done. However, we agree with Schroter et al that the practice of referring readers to a previous article to determine whether ethics review or consent occurred leads to a lack of transparency, as these articles frequently are not easily accessible. We would advocate that all human research reports describe these elements in each manuscript.

A comparison with the results of smaller sampling studies in 1997 and 2002 by Yank and Rennie respectively2 3 demonstrates that reporting continues to improve, both overall and for each journal. Of 53 articles randomly selected from consecutive issues of the same five journals in the 1997 study, only 32% recorded that both REB approval and informed consent had been obtained. By 2002, 9% of articles still had no mention of obtaining either item. The strength of the present study is that we examined all 1133 eligible articles over a two-year period for statements of REB approval and informed consent, rather than a random sampling of clinical trials. We provide therefore, a very precise estimate of the frequency of suboptimal reporting and demonstrate continuing improvement. A possible explanation for this is the increased awareness of the issue brought about by several studies in the past decade.26 Notably, the BMJ now includes a specific section within the acknowledgements of all articles for stating ethical approval, similar to specific formatting of manuscripts for statement of conflict of interest. Compared with 2002, the BMJ had the greatest improvement in reporting from a rate of approximately 74–75% in 20023 to rates of 86–96% presently.

Journals examined in this study were often inconsistent as to where statements on ethical review or informed consent were placed. While found usually in the methods section, statements of ethical conduct were sometimes embedded in the results or with the acknowledgements. It is plausible that reporting might improve if all journals adopted the current practice of the BMJ of having a specific place for REB approval, and if all journals created a place for a statement of informed consent. A secondary outcome of this study was an examination of reporting conflict of interest. The absence of a statement reporting conflict of interest was rare. However, prior to the addition of a specific competing interest statement part way through 2005 in both the NEJM and JAMA, there were 77 (6.8%) articles that had only a statement of funding agency rather than explicit statement of whether there were any conflicts of interest. We describe the excellent rate of reporting of conflict of interest as an example of how a template for authors of required elements of ethical disclosure could markedly assist in more consistent reporting.

This study has several strengths including the comprehensive nature of the examination. All articles classified as deficient for REB approval or informed consent were reviewed again by the primary reviewer and a second reviewer to minimise the risk that statements of REB approval or informed consent were missed. Both reviewers used keywords in the search function of the viewing program. In addition, a strength of our methodology was that we went on to document whether or not editorial comment regarding these omissions was present. This has not been reported to our knowledge in the past. It is also possible that articles exist that were not linked through PubMed or Highwire that discuss the cited articles lacking statements of REB approval or informed consent. However, such articles would then not be easily accessible by the reader and are of little use in alerting them to the deficiencies of the article in question. The handling of survey responses may vary among different centres. In this study, if the survey involved subjects returning a questionnaire and the study had REB approval, it was accepted that the return of the questionnaire implied consent. It is important to note that this study searched for statements of REB approval and informed consent and a lack of reporting does not mean that those articles did not meet these ethical standards. In 2007, Perkins et al found that when papers lacking evidence of REB approval were returned to authors, 48% were resubmitted with the necessary information.13 The study also revealed that International authors may be less likely to include a statement of REB approval with initial manuscript submission than US authors, citing “oversight or misunderstanding as a reason for omission.” Such errors may be prevented by the inclusion of standardised templates for statements of REB approval in all journals. The next step of this study is to contact the authors of papers lacking one or both of these statements in their manuscript to determine the proportion that were breeches of research conduct.

Articles that lack explicit statements of REB approval and informed consent continue to be published in major medical journals without editorial statements to alert the reader to this deficiency. Future work is needed to determine whether this is a deficiency in conduct or reporting. There has been a recent call for a National Research Integrity Agency to support public trust in research.14 Specific templates that encourage authors to specifically describe all ethical elements relevant to publication of human research are likely to improve the quality and consistency of reporting, and support the important aim of public trust. We also call to editors and peer reviewers to alert authors that submissions that lack statements attesting to ethics board approval, informed consent and/or disclosure of conflict of interest will not be accepted for publication until such comments are satisfactorily addressed.

REFERENCES

View Abstract

Footnotes

  • Competing interests: None.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.