‘Safety’ is at the centre of surgical practice with the aim of minimising the risks of complications and adverse events. Much evidence, based on either retrospective case series or prospective cohorts, concerns the frequency of adverse events. There may be a temptation to describe a procedure as ‘safe’ if no – or few – serious adverse events (the numerator) have occurred out of a number of procedures performed (the denominator). In 1983, Hanley and Lippman-Hand described a simple algorithm to calculate the 95% upper Confidence Interval for data sets in which the numerator is zero (ie series in which there no adverse events). Paediatric orthopaedics suffers from small datasets which may make its researchers especially prone to the erroneous attribution of procedures being ‘safe’. The aim of the current study was to formally assess the evidence on which paediatric orthopaedic surgical procedures are described as ‘safe’. In particular, the objective was to ascertain the proportion of studies describing a procedure as ‘safe’ which achieved a 95% upper limit Confidence Interval of risk of 5% for major adverse events.

We examined all papers published by the Journal of Paediatric Orthopaedics in the previous 5 years searching for the single term ‘safe’. 84 papers were returned and 71 were considered appropriate for analysis. Of these 60 papers positively identified their intervention as ‘safe’. These papers were read in full and the number of interventions was recorded along with the rate of complication. 66 data sets were created and the 95% upper confidence interval was calculated for complication rates. Only 16 out of 66 data sets could safely predict a major complication rate of under 5%.

Our work would tend to suggest that a failure to apply proper statistical tools is leading to procedures being erroneously classified as safe in the published literature.