Objectives: To identify the literature on primary total hip replacement (THR) surgery that is relevant to the question of whether prostheses differ in their medium to longer term outcomes, and to synthesise this evidence. To use evidence regarding both costs and outcomes of primary THR to model how much more effective newer prostheses must be to justify higher costs. Methods: Data sources: Electronic searches of MEDLINE and EMBASE (1980- 1995). Hand-searches (1980-1995) of the 11 journals with the highest yield of relevant articles in the electronic searches. Study selection: Randomised controlled trials (RCTs) of any kind that compared prostheses for primary THR. Observational cohort studies that included concurrent controls. Observational studies of single prostheses with at least 5 years of follow- up and reporting outcomes in terms of revision rate or semi-standardised clinical assessment. Data extraction and synthesis: It was not possible to carry out meta-analysis of the evidence from RCTs because each trial compared a unique pair of prostheses. A more informal form of meta-analysis was performed in which all data (randomised and observational) were combined for any prosthesis for which at least five independent studies reporting revision surgical rates were obtained. The meta-analysis was termed 'informal' because of the impossibility of controlling for numerous biases in the data and the poor quality of reporting of much of the evidence. Revision rates for eligible prostheses were calculated, adjusted for person-years at risk. Data were also combined for meta-analysis for other outcomes (i.e. hip scores, global ratings of success, and proportion of patients pain-free). However, studies lacked evidence of patient-based outcomes, and clinicians' views of outcome required substantial modification of diverse clinical ratings to produce a standardised score. Costs and benefits of primary THR were assessed using Markov modelling, and calculation of costs per quality adjusted life- year, with sensitivity analysis of the results. Outcomes data were taken from a prospective study of a series of patients followed up for 14 years after THR. Costs were estimated from cost-generating events for THR and unit costs from a single centre (Nuffield Orthopaedic Centre, Oxford). Results: Eleven RCTs were found that compared outcomes of prostheses. The trials followed up patients for short time periods (mean, 3.9 years) and had quite small sample sizes (mean, 168 patients). A significant difference between prostheses in terms of revision rate was observed in only one RCT. When results of all reports that included a revision rate were combined, ten prostheses met the criterion set for a meta-analysis that at least five independent studies should be available for a prosthesis to be included. Adjusted THR revision rates (revision rate per 100 person-years at risk) were calculated for each of the ten prostheses to take account of different lengths of observation. The most favourable adjusted revision rates were found for the Exeter, Lubinus and Charnley prostheses. Intermediate results were found for the Muller, McKee-Farrar and Stanmore prostheses. The least favourable adjusted revision rates were observed for the Ring, Harris-Galante, PCA and Charnley- Muller prostheses. Economic modelling indicated that to be cost-effective the following improvements in THR outcome and revision rates would be needed. For a new prosthesis costing three times more than the standard Charnley (i.e. typical cost of a new cementless prosthesis): ≥ 35-44% improvement in patients aged 50-70 years; ≥ 21-27% improvement in patients aged < 50 years. For a new prosthesis costing 1.5 times more than the standard Charnley (i.e. typical cost of a new cemented prosthesis): 9-12% improvement in patients aged 50-70 years; 6-7% improvement in patients aged < 50 years. From the available evidence, the extent of the improvement required of new and more expensive prostheses is particularly implausible for older patients. However the new cheaper prostheses may be cost-effective because the improvements required are more likely to be achievable. Conclusions: There is a striking paucity of clear and relevant evidence on which to make well-informed choices about prostheses for primary THR. Although basic scientific innovation continues in relation to THR, the knowledge base to inform selection of prostheses is unlikely to improve in the foreseeable future. Of prostheses commonly used in the NHS by far the greatest volume of evidence is available for the Charnley and on the basis of that evidence the Charnley appears to perform relatively well. However, the Charnley design has changed, and it is not clear how much of the evidence is relevant to the current design. Of other prostheses currently used in the NHS, positive evidence (but no data from RCTs) was found in support of the Exeter prosthesis, and some positive evidence was found for the Stanmore (for example, evidence that it performed as well as the Charnley in an RCT). Positive evidence for the Lubinus IP (less widely used in the NHS) was also found. The quality of the evidence for other prostheses was either poor or non-existent. No substantial evidence could be found for cementless prostheses in terms of independent observation of results from five or more studies. None of the analyses used in this review, such as meta-analysis of evidence, could overcome the fundamental weaknesses of the available evidence. The poor quality of evidence overall does not provide a basis clearly and authoritatively to identify prostheses that could be - or should not be recommended for use by the NHS. However, it is clear that the more expensive the prosthesis, the more difficult it is to provide justification for its selection on the basis of the current evidence. On the basis of the economic analysis it seems that the use of the more expensive (i.e. cementless) prostheses is hard to justify on current evidence. Recommendations for future research: As a substantial proportion of the evidence on outcomes of THR comes from healthcare systems quite different from the NHS (i.e. the Swedish and Norwegian national registers) it is recommended that the case for a UK register should be evaluated. Least biased assessments would be from RCTs, but to detect the small but important differences that may exist between prostheses such trials must be more adequately designed and powered than those carried out previously, and should involve multicentre participation and long-term follow-up. Economic modelling in this review indicates that such trials might identify differences in cost- effectiveness between cemented prostheses. Patient-based outcomes provide relevant and feasible methods to conduct large multicentre studies. To obtain unbiased assessments of outcome, the focus should be on outcomes of concern to patients, particularly pain and function, and not solely on revision surgery.


Journal article


Health Technology Assessment

Publication Date