Thromboprophylaxis remains a controversial issue and many disagree about the optimum method or even if it is required at all.

We present a new method of performing meta-analysis incorporating studies with both experimental and observational study designs. We have developed a model that compares study cohorts of several different methods of thromboprophylaxis with a simulated matched control group whose variance helps to adjust for bias. This allows meaningful comparisons between studies and treatments that have not been directly compared.

We performed a systematic review of the literature from 1981 to October 2004. Studies where more than one method of prophylaxis was used were excluded from analysis. For each individual method of prophylaxis, data was extracted, combined and converted to give estimates of the rates of symptomatic, proximal DVT, fatal PE and major bleeding events. We identified 1242 studies of which 203 met the inclusion criteria for further analysis. This represented the results of over fifty thousand studied patients. We expressed the results for the different prophylactic methods as odds ratios compared to no prophylaxis.

All methods showed a beneficial effect in reducing VTEs apart from stockings and aspirin which showed an increase in the number of PE events. These results are particularly interesting when viewed from the standpoint of an individual NHS hospital trust that performs around 500 hip and knee replacements per year. Over a 5 year period, the more effective methods of prophylaxis prevented between 15 and 40 symptomatic DVTs and up to 3 fatal PEs compared to no treatment. However, they cause between 8 and 40 more major bleeding events. We do not know the proportion of these major bleeding events that are fatal.