Endorsement of reporting guidelines and compliance
A journal’s endorsement of a reporting guideline does not establish that its published reports meet the guideline’s requirements.
Endorsement and active implementation
Studies discussed in 2013 found deficiencies in Chinese manuscripts and abstracts against CONSORT and CONSORT for Abstracts, including studies published in high-impact Chinese journals that expressly endorsed CONSORT. In a trial sample discussed in 2015, 74% had at least one methodological element at unclear risk of bias because reporting was incomplete, despite widespread dissemination of reporting guidelines and an estimated 600-plus journals endorsing CONSORT.
Many editors merely mention reporting guidelines in author instructions, while more active approaches include requiring a checklist at submission or acceptance and asking peer reviewers to assess reporting against the guideline. Active implementation of CONSORT for Abstracts was associated with better abstract reporting. Even well-designed trials can be inadequately reported; clear reporting can improve the chance that their results are interpreted correctly, and adherence to CONSORT could reduce research waste.
Spread of journal policies
Three surveys of high-impact journals’ author instructions showed increasing endorsement of CONSORT, although implementation remained far from standardised. The surveys reported the following mentions of CONSORT, ICMJE and trial registration.
| Author-instruction item | 2003: 166 journals, selected using 2001 impact factors | 2007: 165 journals, selected using 2006 impact factors | 2014: 168 journals, selected using 2012 impact factors |
|---|---|---|---|
| CONSORT | 36 (22%) | 62 (38%) | 106 (63%) |
| ICMJE | 72 (43%) | 69 (42%) | 130 (77%) |
| Trial registration | Not collected | 61 (37%) | 106 (63%) |
In the 2014 survey, 42% of journals made CONSORT mandatory for trial reporting. The 1996, 2001 and 2010 CONSORT statements and elaboration documents had received more than 12,000 Scopus citations by May 2015.
A separate review on 30 November 2014 examined 60 journals: ten selected by highest impact factor in each of six medical specialties, using impact factor to identify commonly cited journals rather than to judge their quality. Of these journals, 47% required PRISMA and another 10% recommended it, most often in obstetrics and gynaecology and in general and internal medicine. Fewer journals required the checklist or flow diagram. The specialty results were as follows, with ten journals in each specialty.
| Specialty | Recommended PRISMA | Required PRISMA | Required checklist | Required flow diagram |
|---|---|---|---|---|
| Obstetrics and gynaecology | 2 (20%) | 7 (70%) | 1 (10%) | 1 (10%) |
| General and internal medicine | 2 (20%) | 6 (60%) | 3 (30%) | 4 (40%) |
| Anaesthesiology | 1 (10%) | 5 (50%) | 5 (50%) | 1 (10%) |
| Paediatrics | 1 (10%) | 3 (30%) | 3 (30%) | 2 (20%) |
| Surgery | 0 | 4 (40%) | 2 (20%) | 2 (20%) |
| Critical care medicine | 0 | 3 (30%) | 2 (20%) | 2 (20%) |
| All 60 journals | 6 (10%) | 28 (47%) | 16 (27%) | 12 (20%) |
Journals not publishing meta-analyses were replaced by the next-highest-impact-factor journal; recommendations included encouragement or descriptions of value, and requirements included “should”, “must”, requests or instructions to follow PRISMA.
Reported compliance with PRISMA
Among 44 meta-analyses published in November 2014, a median 29.6% of PRISMA items were reported inadequately. A PRISMA-like flow diagram appeared in 41 reports (93.2%).
Adequate reporting was higher where journals required or recommended PRISMA and where authors stated that they followed it. The median proportions of items reported adequately were as follows.
| Comparison | Median adequately reported |
|---|---|
| All 44 meta-analyses | 57.7% (range 16.0%–92.6%) |
| Journal required or recommended PRISMA, 18 reports | 63.0% (40.0%–92.6%) |
| Journal did not, 26 reports | 55.6% (16.0%–70.4%) |
| Authors stated PRISMA was followed, 21 reports | 59.3% (44.0%–92.6%) |
| Authors did not, 23 reports | 55.6% (16.0%–70.4%) |
The journal-policy comparison had P<.02 and the author-statement comparison P<.05 using Mann–Whitney rank-sum tests. Despite statistical significance, neither factor had a dramatic effect on reporting completeness, and the 59.3% among reports claiming adherence was lower than expected. Possible explanations included journal word limits and content requirements, authors not appreciating the importance of adequately reporting each item, and difficulty understanding how to apply PRISMA despite its checklist and explanation-and-elaboration article.