Authorship Ethics Requirements in Medical Journals William Gardner, PhD Center for Innovation in Pediatric Practice Columbus Children’s Research Institute.

Authorship Ethics Requirements in Medical Journals

William Gardner, PhD

Center for Innovation in Pediatric PracticeCenter for Innovation in Pediatric PracticeColumbus Children’s Research InstituteColumbus Children’s Research Institute

Center for Innovation in Pediatric PracticeCenter for Innovation in Pediatric PracticeColumbus Children’s Research InstituteColumbus Children’s Research Institute

Departments of Pediatrics,Departments of Pediatrics,Psychology, & PsychiatryPsychology, & PsychiatryThe Ohio State UniversityThe Ohio State University

Thanks to Kendra Heck, MPH, and Ling Fan, MD. Supported by NIH/ORI: RO1 NS49591

Journals’ Statements of Ethical Requirements for Authors

• Requirements for publications of clinical trials cover human subjects protection, disclosure of conflicts of interest, and authorship ethics.

• Journals publish statements detailing norms.

• Authors affirm their compliance with the norms.

• The content of these statements are evidence of professional ethical expectations for authors.

Prior Literature on Requirements for Authors

• Previous studies focused on human subjects protection. • They used convenience samples and did not explore

variation across journals.• In the population of journals publishing RCTs, what are

expectations for human subjects protection, conflict of interest, and authorship ethics?

• What explains variation across journals in requirements?

Journals Publishing Clinical Trials (CTs) in 2005

Median = 2.5 articles.

We retrieved 13,184 references to CTs, published in 2,056 journals.

We used Haynes et al.’s ‘conservative’ strategy to retrieve CTs from PubMed.

Average journal published 6.4 articles.

Top 25% of journals publish 74% of articles.

Journals by CT Count: Log-Log Plot

80/20 laws: Web site hits, crime, income, journal citations.

Matthew effect: “Unto him that hath, shall be given.”

The distribution of journal CT counts falls somewhere between a Power Law (dashed line) and Log-Normal (curved line).

Sampling Journals & the Distribution of CTs

Looking just at high impact / high prestige journals would just give us data on a very select set of journals.

So, we sampled all journals that published 50 or more CTs in 2005, and randomly chose journals from the rest.

This meant that journals that published < 50 CTs were undersampled. So we weighted them in the analyses to correct for the undersampling.

However… simple random sampling of journals would give us very few of the journals that matter to the most readers and that publish a disproportionate share of articles.

Scientific & Ethical Quality Control

• We retrieved instructions to authors from the web or sent a request to the journal editor.

• We coded the instructions for the following:• Journal Self-Interest: e.g, no previous

publication.• Human Subjects: Informed Consent + IRB

review.• CT Standards: CONSORT + Registration

Components of Authorship Ethics (AE) Score

• Disclosures: Source of funding, COIs.

• Authorship: Author Contributions, Components of Authorship as defined by Vancouver.

• Affirmations of Responsibility: For scientific work and data integrity.

• 10 elements in total, Cronbach’s = .87.

Journal Requirements

13%

24%

43%

79%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Journal SelfInterest

Human Subjects Authorship CT Standards

Ethics Requirement

Elements Present

N = 289 journals, weighted analysis.

Estimated Population Density: Authorship Ethics Requirements

N = 289 journals, weighted analysis.

0%

5%

10%

15%

20%

25%

30%

0 25 50 75 100

Authorship Ethics Score

% of Journal Population

What Journal Factors Affect Authorship Ethics?

• Negative binomial regression predicting number of AE elements present (with weights).

• Model includes journal age, CT count, US or European publication, journal ‘discipline’, and whether the journal was electronic.

• Important methods problem: Journals are not independent observations.

What Journal Factors Are Associated with AE?

Variable b z p

Journal Age .002 1.77 .077

Pediatrics .26 2.47 .014

Pharmacology -1.04 -6.54 <.0001

Surgery .22 2.55 .011

Electronic Journal 1.17 8.26 <.0001

CTs Published .016 7.29 <.0001

Where the journal is published does not matter. Only statistically significant ‘discipline’ factors are shown.

Journal age is associated with AE (older have higher AE), but not when you control for CTs Published.

Authorship Ethics Score & Number of CTs Published in 2005

N = 289 journals, weighted analysis.

0

10

20

30

40

50

60

70

80

0 30 60 90 120 150 180

CTs Published in 2005

Authorship Elements

Conclusions

• CTs are published in a ‘long tail’ of journals: Many appear in lesser known journals that previous studies have ignored.

• AE requirements vary across journals.• Journals publishing many CTs have stiffer

requirements.• AE requirements may vary by discipline; pure

electronic journals had high AE requirements.

Implications for Ethics Research on Scientific Journals

• Studies of journal population should use epidemiological rather than convenience designs.

• Studies should look for journal factors associated with variation in ethical requirements.

• Don’t just look at the high volume journals. There is also a ‘long tail’ of small journals

• How will E-publishing affect AE policies?