Statistics from Altmetric.com
Editor,—The paper by Sandramouli et alsuggested that convulsions rarely, if ever, give rise to retinal haemorrhages.1 This was based on a series of 33 patients with convulsions, none of whom would seem to have had retinal haemorrhages. The statistical analysis they undertook was based on Hanley’s rule of 3.2 Its use can be summarised by their statement ‘It is a good estimate of the worst case that is compatible with the observed data’.
Applying the rule to the series of 33 patients reported, it gives a worst case that is compatible with the observed data of 3/33 (approximately 9.1%).
The most that can be said as a result of this series, is that convulsions do not commonly give rise to retinal haemorrhages. If the word rarely was taken to mean less than 1% of cases, then the series would have had to involve 300 children and if the phrase ‘if ever’ was to mean less than 1/1000, the series would have had to include 3000 children, none of whom had retinal haemorrhages associated with convulsions.
The conclusion that convulsions rarely (if ever) give rise to retinal haemorrhages is probably a good deal more memorable than it is useful.
Mr Willshaw comments:
We thank Dr O’Donohoe for his observations on our statistical analysis of the study of 32 children (one child from the original group of 33 was excluded).
Unfortunately, the phrase ‘none of whom would seem to have had retinal haemorrhages’ seems cast some doubt on the observation. We would emphasise again that these children all receive detailed ophthalmological examination, including the use of an indirect ophthalmoscope, within 48 hours of admission. Categorically none of them had suffered retinal haemorrhages.
We would exactly concur with Dr O’Donohoe’s interpretation of Hanley’s rule of 3, but would emphasise again that this gives a 95% confidence level in this study. Within the text of the article, this is described as indicating that ‘the chance of a child having retinal haemorrhages solely on the basis of having a convulsion is unlikely’ and later ‘that retinal haemorrhages in children are rarely associated with convulsions’. We would still feel that both of these observations are entirely appropriate on the basis of this statistical analysis and Dr O’Donohoe may be interested to know that a further 32 children have now been examined, all of whom were under the age of 24 months, and still, to date, there has been no incidence of retinal haemorrhage occurring within 48 hours of the convulsion.
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.