Statistics from Altmetric.com
The best evidence topic report (BET) by Hogg and Mahu1 raises a number of concerns, both with the article itself and the BETs process as a whole. The relative efficacy of adrenaline and vasopressin in the management of cardiac arrest is an important subject of relevance to all who work in emergency medicine. For this BET to only include those papers directly comparing vasopressin and adrenaline is to dismiss a large amount of research published in this area. A very brief search on Medline reveals a large number of articles looking at this subject, including two recently published reviews comparing adrenaline and vasopressin,2,3 not mentioned by the authors. Surely a topic such as this should be subject to a formal literature review and meta-analysis, not the “shortcut review” method advocated by the BETs methodology.
One of the stated aims of the BET methodology is to summarise the highest level of evidence available.4 In response to previous criticisms Professor Mackway-Jones was keen to emphasise that those undertaking BETs “go to great lengths to ensure that the search strategies used are highly sensitive”.5 It would appear from this article that this has not been the case on either count.
Given the above concerns could the journal editors confirm whether the BETs are subject to the same peer review process as other articles in this journal.
We read with interest the comments on our best evidence topic review on vasopressin or adrenaline in cardiac resuscitation and are happy to explain the process entailed in producing the BET.
This literature search was first conducted in March 2002. Our initial and specific question was: Is vasopressin more effective than adrenaline in achieving return of circulation and long term survival, in human cardiac arrest?
A full and sensitive search strategy was compiled. The search strategy was checked by two additional independent doctors who cross checked their own strategies to maximise the sensitivity. All titles and abstracts were appraised initially by the two authors and before publication by the third independent author. The relevant original studies and review articles were sourced in full text (18 in total). All review articles were cross referenced.
These 18 papers were reduced to four potentially relevant papers. This included the two published studies. The search strategy and all four papers were reviewed by the Manchester Royal Infirmary Emergency Medicine journal club. A consensus decision was taken to exclude from the analysis one study reporting the effects of intravenous vasopressin on coronary perfusion pressures in 10 patients1 and a second that reported the effects of intravenous vasopressin administered in refractory cardiac arrest, in eight patients.2 The first study did not use return of circulation as an outcome, and the second did not compare the effects of vasopressin and adrenaline. The decision to publish this review was taken four months before publication. At that point it was reviewed by Professor Mackway-Jones, the first author repeated the search and a third independent author checked the relevant articles and search strategy.
To directly respond to your points
The BET addresses a specific question. We did not aim to present a vague representation of all literature on vasopressin but asked the question—Is vasopressin better than adrenaline in a human cardiac arrest?
Dr Locker wonders why we did not include two recent reviews addressing this question. A less superficial appraisal of these papers would have revealed they found the same four studies. We have cross referenced all the review articles published on this subject and can find no further studies.
We do not pretend that these reviews are systematic reviews, (we used Medline only and do not search for unpublished data), but we do openly lay out our methodology. It would be possible for another doctor, anywhere in the world, to repeat this search exactly if they so required.
At present, a meta-analysis would appear wholly inappropriate to answer this question. There are only two small studies, the results of which are clearly laid out in the table. The reader is capable of drawing their own conclusions from this table. When more relevant data are available (there are at least two ongoing studies), a meta-analysis may be of use.
In conclusion, this BET is an accurate and reproducible formal review. It addresses the question posed by the authors and clearly summarises the relevant published literature.