Survival analysis of joint replacement relies on the assumption that surgical procedures in patients lost to follow-up have the same chance of failing as those in patients who continue to be assessed. Our study questions that assumption.

During the 16-year follow-up of 2268 patients who had received total hip replacements 142 (6%) were lost to follow-up. The cumulative loss at 15 years was 20%. At their last assessment, patients who subsequently failed to attend for follow-up had significantly worse pain, range of movement and opinion of their progress (p < 0.001) and significantly worse radiological features than a matched control group (p < 0.01).

Patients lost to follow-up have a worse outcome than those who continue to be assessed. Consequently, a survival analysis that does not take into account such patients is likely to give falsely optimistic results. It is therefore essential that vigorous attempts are made to minimise loss to follow-up, and that the rate of such loss is quoted. The overall loss to follow-up disguises the magnitude of the problem, which is best quantified by a cumulative rate of follow-up.

The reliability of a study can be assessed by a loss-to-follow-up quotient, calculated by the number of failures: the lower the quotient the more reliable the data. Ideally, the quotient should be less than 1.

We have assessed the relative value of various outcome measures after THR, by the analysis of follow-up data from over 2000 patients. They had been reviewed clinically and radiologically six months after operation, at one year, and then every two years, some for 16 years. At each review their pain level, stiffness and opinion of progress were scored and a radiograph taken.

We found that pain level was the most informative outcome as a predictor of revision and correlated well with the patients’ opinions.

We made a comparison between the six types of implant in the series, using survival analysis and log-rank testing with different pain levels as endpoints. This analysis revealed differences which were not detected by survival analysis using the traditional endpoint of revision.

We therefore recommend the use of different levels of pain as the main outcome measures after total hip replacement.

The recommendation that patients having a total hip replacement should receive pharmacological thromboprophylaxis is based on the belief that fatal pulmonary embolism is common, and that prophylaxis will decrease the death rate. To investigate these assumptions we performed a meta-analysis of all studies on hip replacement which included information about death or fatal pulmonary embolism. A total of 130 000 patients was included. The studies were so varied in content and quality that the results of our analysis must be interpreted with some caution.

The fatal pulmonary embolism rate was 0.1% to 0.2% even in patients who received no prophylaxis. This is an order of magnitude lower than that which is generally quoted, and therefore the potential benefit of prophylaxis is small and may not justify the risks. To balance the risks and benefits we must consider the overall death rate. This was 0.3% to 0.4%, and neither heparin nor any other prophylactic agent caused a significant decrease.

Our study demonstrates that there is not enough evidence in the literature to conclude that any form of pharmacological thromboprophylaxis decreases the death rate after total hip replacement. For this reason guidelines which recommend their routine use to prevent death after hip replacement are not justified.

We reviewed the records of the long-term outcome of 208 Charnley and 982 Stanmore total hip replacements (THR) performed by or under the supervision of one surgeon from 1973 to 1987. The Stanmore implant had a better survival rate before revision at 14 years (86% to 79%, p = 0.004), but the difference only became apparent at ten years.

The later Stanmore implants did better than the early ones (97% to 92% at ten years, p = 0.005), the improvement coinciding with the introduction of a new cementing technique using a gun. Most of the Charnley implants were done before most of the Stanmore implants so that the difference between the results may in part be explained by improved methods, but this is not the complete explanation since a difference persisted for implants carried out during the same period of time.

We conclude that improved techniques have reduced failure rates substantially. This improvement was much greater than that observed between these two designs of implant. Proof of the difference would require a very large randomised controlled trial over a ten-year period.