Aim: According to the "practice makes perfect" hypothesis, the mortality rates of many surgical procedures are inversely related to hospital and surgeon procedure volume. Currently there is a discussion regarding the implementation of high-volume centres for total hip replacement. Nevertheless, the body of evidence describing such an association is weak. Therefore the American National Cancer Policy Board came up with four evidence criteria describing a valid association between volume and outcome: 1) the association must be logical, 2) the association must be consistent in all studies, 3) the size of the measured effect must be substantial and clinically relevant and 4) the effect has to be reproducible in clinical studies.
Method: We searched MEDLINE and EMBASE (1996 until November 2006) using the following "medical subject headings" (MeSH): "Arthroplasty, Replacement, Hip" or "Hip Prosthesis" and "Volume". We searched for full articles that reported on an association between hospital and surgeon procedure volume and outcome after total hip replacement. No language restriction was applied.
Results: We found 17 relevant studies in MEDLINE and EMBASE. The outcome after primary hip prosthesis and revision hip prosthesis was better when performed in high-volume hospitals or by high-volume surgeons. Depending on the cut-off for the definition of "high volume" and "low volume" the results could be statistically significant.
Conclusion: Many of the included studies were of minor quality, the data were often not comparable between studies. Therefore, the results are not directly transferable to our medical system in Germany. Applying the criteria designed by the American National Cancer Policy Board on the studies about procedure volume and outcome after total hip replacement yields the following result: 1) the association must be logical: Yes. 2) The association must be consistent in all studies: Yes. 3) The size of the measured effect must be substantial and clinically relevant and has to be analysed using validated statistical criteria: No. 4) The effect has to be reproducible in clinical studies: No.