Levels of Evidence

Correspondence We welcome letters to the Editor concerning articles which have recently been published. Such letters will be subject to the usual stages of selection and editing; where appropriate the authors of the original article will be offered the opportunity to reply. Letters should normally be under 300 words in length, double-spaced throughout, signed by all authors and fully referenced. The edited version will be returned for approval before publication. Sir, We read with interest the three editorials 1-3 on levels of evidence published in the December 2005 issue. Mr Horan accurately states that surgeons provide optimal care to patients based on a combination of clinical expertise and 'best available evidence'. 1 Herein, however, lies the crux of the problem. The surgical literature very often does not agree, and this disagreement is reflected in a wide variety of treatments for the same clinical condition. Unless all the available interventions have the same risks and benefits, we (or more importantly our patients), have a problem. To resolve clinical controversies, surgeons must sort through and evaluate the conflicting evidence. Some studies are better designed and those studies provide more compelling information. For example, surgeons recognise that controlled studies, compared with uncontrolled case series, provide more convincing data; prospective is more reliable than retrospective data collection ; and randomisation is the most valid way to minimise those factors that influence the outcome of treatment other than the intervention. These universally-accepted principles are the essence of levels of evidence. Levels of evidence, introduced in the American Journal of Bone and Joint Surgery in January 2003, 4 are a simple and intuitive guide to help surgeons quickly appraise the myriad of published studies to assemble the 'best available evidence'. Levels of evidence are now also used by the Grades of recommendation, based on levels of evidence and used to summarise the quality of literature, were introduced in July 2005. 5 Although not considering all aspects of research design, levels of evidence have been shown to be reliable 6 and moreover, those journals that publish articles with higher levels of evidence have higher impact factors. 7 All levels of evidence are useful. 2 For example, John Charnley's case series fundamentally changed the practice of orthopaedics. 8 Total hip replacement is the 'parachute' of orthopaedics – a ran-domised trial was not required to demonstrate its superiority. Unfortunately most treatments in orthopaedics are not so obviously beneficial as …