Multivariate Data Analysis–In Pactice. An Introduction to multivariate data analysis and experimental design (4th edn), Kim H. Esbensen, CAMO, OSLO, 2000, ISBN 82‐9933302‐4, xviii + 600pp, US$230.00.

Abstract

The scientific quality of surgical research is generally low. Many surgical procedures carried out daily have not been validated with solid scientific evidence. For example, after more than 20 years of practice, there are only 3 prospective studies for surgery of liver metastases. Few surgical papers published are randomized trials (RT) and the vast majority present retrospective contributions. Surgical journal editorial teams encourage the publication of RTs, but often accept to publish retrospective series if the authors are internationally recognised. We ask whether the RT is the only way to improve the scientific level of surgical oncology. Are alternative methodologies possible, and are they a good idea? The RT has been designed to test new drugs and not new surgical procedures. The RT experimental paradigm is difficult to transpose to surgery for several principal reasons: 1. The lack of equipoise in surgical behaviours; 2. difficulties inherent in quality control (how to be sure that the repeat procedures in each arm are strictly identical each other); and 3. accrual in surgery is generally slow and RTs should be based on multicentre settings which are difficult to organise. Consequently, most of the RTs published in surgery are underpowered and often stratified when it is not possible to do it. Authors of these trials regularly claim that their results can change practice which is certainly questionable. Journals which encourage this practice are complicit in this denial of science. What should we do? First, we have to identify the different types of clinical research (CR) with which a surgical oncologist may be faced. Depending on whether the CR is focused on a new procedure, an existing not yet evaluated procedure, a new implantable device, an alternative strategy, or an add-on diagnostic tool, the methodological proposal should be different. Secondly, we have to propose new alternative prospective evaluation solutions. If the RT is always the best standard when it is possible to do it, alternatives do exist like prospective registries, phase 2 trials, cohort studies, etc. Prospective evaluations will always be preferable to retrospective ones. This is the only way to improve the scientific quality of surgical sciences. This task will be difficult to achieve as several 'target audiences' need to be convinced. Firstly, surgeons will be required to change their behaviour and to open their minds to different methodology alternatives. Secondly, editorial teams of surgical journals need to be convinced that CR methodology is not limited to RTs and lastly, the expert selection committees of CR tenders will also need to consider the merit of alternative experimental paradigms.