Anecdotes as evidenceBMJ 2003; 326 doi: http://dx.doi.org/10.1136/bmj.326.7403.1346 (Published 19 June 2003) Cite this as: BMJ 2003;326:1346
- Jeffrey K Aronson, consultant clinical pharmacologist ()
“The short story” wrote William Trevor, a master of the genre, “should be an explosion of the truth.” Being interested in the clinical short stories that are anecdotal reports of adverse drug reactions and interactions, I am concerned that they should explode the whole truth.
There are many such anecdotes. Of the 3252 citations in the 24th volume of the Side Effects of Drugs Annual,1 in which the world literature on adverse drug reactions and interactions for 2000 was critically reviewed, about a third (1075 citations) were anecdotes; in contrast, there were only 45 systematic reviews.2
Now the hierarchy of clinical evidence emphasises large randomised controlled trials and systematic reviews; in this scheme anecdotal reports are ill favoured.3 Nevertheless, anecdotal reports of adverse reactions should be published, for they have different functions to randomised controlled trials (table), a fact that is not emphasised by the evidence hierarchy.
A short story sometimes illuminates an aspect of life that a roman à thèse cannot. Likewise, trials typically show the sizes of benefits or adverse effects, but anecdotes call attention to potential adverse reactions or interactions, mechanisms, diagnostic techniques, or methods of management; anecdotes can generate or test hypotheses, and remind or educate; and, like trials, they can be subjected to systematic review, but of a different kind.4
Pharmacovigilance demands at least 19 essential pieces of information about the patient and the adverse reaction, listed on bmj.com. They include 14 items that are mentioned or implied on the Committee on Safety of Medicines' yellow card for reporting suspected adverse drug reactions in the United Kingdom, and 14 that are mentioned or implied on the MedWatch adverse event forms in the United States. Several of these are important in computing the likelihood that the adverse event was an adverse reaction (that is, an event that is caused by the drug) and in elucidating the mechanism.14
However, anecdotal reports of suspected adverse drug reactions and interactions often do not contain all the information that they should. For example, I have reviewed 35 reports about 48 patients that appeared in the BMJ from January 2000 to October 2002 under “Drug points.” The median numbers of items mentioned in those reports were 9 (range 5-12) of all the 19 essential items mentioned above, 9 (4-12) of the yellow card items, and 8 (3-11) of the 14 MedWatch items. And this analysis omits other desirable features of anecdotal reports, such as formal assessment of the likelihood that the event was an adverse drug reaction, possible mechanisms, and review of previous cases.
Of course, it can be difficult to obtain high quality data when an event occurs, but often a more assiduous approach would help. We need formal guidance on what is required.
The CONSORT group has developed standardised guidelines on reporting randomised controlled trials.15 However, no similar guidelines exist for anecdotal reports. I have therefore outlined a proposed set of guidelines (see bmj.com), which would encourage clinicians to report cases uniformly (facilitating direct comparisons of individual reports) and obtain the information needed to assess whether an adverse event is actually an adverse drug reaction and to establish diagnostic techniques, mechanisms, and guidelines for management. Uniform presentation would also facilitate systematic review of suspected reactions.
“The short story tells only one thing,” wrote VS Pritchett, another master of the genre, “and that intensely.” The clinical anecdote should do likewise.
Competing interests None declared.
Extra tables and a suggested protocol for publishing anecdotal reports appear on bmj.com