The validity of clinical examination in the diagnosis of loosening of components in total hip arthroplasty
Total hip arthroplasty (THA) is one of the most successful orthopaedic operations.12 After the reduction in the incidence of infection as a result of improved surgical techniques, laminar flow operating theatres, and the use of intravenous antibiotic prophylaxis,3 aseptic loosening remains the most severe complication. Early loosening is often painless,1,4-6 and therefore regular clinical and radiological review has been recommended in order to identify and treat it at an early stage. Chamley7 suggested that aseptic loosening was more commonly related to the acetabular component and that cups needed to be revised earlier than stems. Muller8 reported that loosening of the cup often produced no pain and was usually therefore detected late. As soon as micromovement at the implant-bone interface has begun, there is progressive destruction of bone.2.4,5,9 Because of this symptomless interval the loss of bone stock may be massive before patients seek advice. By then, the conditions for a revision procedure are unfavourable and the expected survival time for a new prosthesis is decreased.9-12 Thus regular radiological assessment of a THA is presently considered to be the most consistent method of diagnosing loosening, even in asymptomatic patients. The opposite approach is to perform clinical review based on patient demand. If clinical indicators can be found which provide the surgeon with sufficient information to decide whether a component is loose,13 regular clinical controls and radiological reviews can be minimised.
Our aim therefore was to assess the value of clinical signs in diagnosing loosening of a component in a large population.
Patients and Methods
The data were derived from the database of the Maurice E. Muller Institute for Evaluative Research in Orthopaedic Surgery. Preoperative and postoperative clinical and radiological data were prospectively documented using optically– readable code-sheets from consecutive primary THAs according to the standards of the International Documentation and Evaluation system (IDES).14
The following criteria had to be met for inclusion in the study: osteoarthritis as the main diagnosis, primary THA, age over 20 years at THA, and the availability of serial documented follow-up examinations for at least ten years after operation with a complete set of preoperative, immediately postoperative and follow-up radiographs.
The database search identified 15734 patients (50.4% male, 49.6% female) with 18 486 primary THAs. Data were collected between 1967 and 2002 from 41 hospitals in several European countries. There were 8696 left and 9790 right TRAs and 2752 patients (17.5%) had had bilateral procedures. The median age at surgery was 67.4 years with a 75% percentile of 73.8 years and a 25% percentile of 60.1 years. The median body mass index (BMI, weight (kg)/ height (m^sup 2^)) at surgery was 26.83 kg/m^sup 2^. Of the 15 734 patients, 843 had had previous surgery on the affected hip including internal fixation (95), femoral osteotomy (701), arthrodesis (16) and other procedures (31). Of the four different methods of implantation, there were 9232 classic hybrid THAs (uncemented cup, cemented stem), 4041 totally cemented THAs, 2882 totally uncemented THAs, and 2331 reverse hybrid THAs (cemented cup, uncemented stem). In the first decade after THA, 38788 follow-up examinations with radiological documentation were available. There was a mean of 2.1 documented follow-up consultations per THA.
Definition of cases (Table I). Assessment of the status of the components was performed based on standardised anteroposterior pelvic and lateral radiographs with the MEM– template for the evaluation of THA as a standardised measurement tool.15 Acetabular and femoral loosening was defined by comparing the postoperative and follow-up radiographs and measuring superior and medial migration and the tilt of the cup,16,17 radiolucencies around it,5 a broken cup 18 or broken cement,17 subsidence of the stem,19 radiolucencies at the stem-bone or cement-bone interface,18,20 a progressive tilt of the stem and cavitations and fracture of the stem.17
The following IDES variables were therefore used to define loosening of a component:
1) Continuous radiolucencies around the cup in zones 1 to 3 of DeLee and Charnley.4
2) Superior migration >= 5 mm, severe protrusion or progressive tilt of the cup. The 4 mm limits used in the literature were equated to 5 mm according to the IDES classification.
3) Fracture of the cup or the cement mantle with cemented cups. Subsidence of the stem >= 3 mm.
4) Radiolucencies > 2 mm at the cement-bone or stem-bone interface.
5) Continuous radiolucencies at the cement-bone or stembone interface.
6) Multiple small cavitations or large defects around the stem.
7) Fracture of the stem or cement.
Physical examination and patient-derived indices (Table II). The following clinical variables were included: the amount of pain in the hip with none or mild pain as the desired outcome and moderate, severe or intolerable pain as a clinical failure. Further variables were the site of pain (buttock, groin, thigh, knee), pain on testing (trochanter, axial compression, internal rotation, external rotation) and flexion.
In order to calculate sensitivity, specificity and predictive values, all clinical variables were transformed into a binary format.
Analysis of data. The data were analysed in terms of sensitivity, specificity and predictive values with the radiological definition of loose or not loose as the `gold standard. Sensitivity was defined as the proportion of symptomatic hips which were correctly diagnosed as loose. Specificity was defined as the proportion of asymptomatic hips which were correctly diagnosed as not loose. The positive predictive value (PPV) corresponds to the proportion of loose prostheses within the group of symptomatic patients, whereas the negative predictive value (NPV) refers to the proportion of fixed prostheses within the group of asymptomatic patients. All analyses were performed separately for cemented and uncemented components. The frequency of radiologically diagnosed loosening of the cup and stem was calculated as the prevalence separately for each year after operation; 95% confidence intervals for the prevalence were calculated, considering the fact that data were clustered in patients with bilateral procedures. Differences in the prevalence of loosening between cemented and uncemented components were analysed using logistic regression appropriate for clustered data.21 Adjustments for age and gender were made by including corresponding covariables into the logistic models.
The validity of the clinical indices for diagnosing loosening was assessed using a two-step procedure. In a first step, the sensitivities and specificities of each of the ten clinical variables were calculated for each year of the first decade after operation. Based on the results of this procedure and on clinical importance, the subsequent analysis was focused on four and eight years after operation. Data were analysed using STATA (Stata Corporation, East College Station, Texas). The level of statistical significance was set at 0.05. Results Figure I gives the prevalence of loosening of the cup and stem over 20 years.
Prevalence of loosening of the acetabular component. The overall prevalence of loosening of cemented and uncemented cups ranged from 0.6% to 13.9% depending on the year after operation. It was higher for cemented cups throughout the first and into the second decade after operation. Age- and gender-adjusted differences in the prevalence of loosening between cemented and uncemented cups were, however, not always significant (Table III).
Prevalence of loosening of the femoral component. The overall prevalence of loosening of cemented and uncemented stems ranged from 1.7% to 12.5% depending on the year after operation. The prevalence varied between cemented and uncemented stems throughout the first six years after operation. Between seven and more than ten years after the operation the prevalence of loosening was higher for cemented systems. Age- and gender-adjusted differences in the prevalence of loosening between cemented and uncemented stems were mostly significant (Table IV).
Sensitivity and specificity. Sensitivities ranged between 0.00 and 0.49 for uncemented and between 0.00 and 0.6 for cemented cups. Figure 2 and Tables V and VI give the mean values. A slight time-dependent increase in sensitivity was seen during the first decade after operation. The specificity of all indices was constantly between 0.89 and 1.00, regardless of the mode of fixation of the cup. Figure 3 gives the mean values. Time trends of specificity were slightly negative and, unlike sensitivity, the specificities of the various clinical indices appeared to be homogenous.
For the stems, sensitivities ranged between 0.0 and 0.57 for cemented and between 0.0 and 0.46 for uncemented components. The sensitivities of most variables showed more constant time trends compared with those of the cups. Figure 2 and Tables VII and VIII give the mean values. Most values in the uncemented group had a higher variability over time within the mentioned range. The specificity of all indices was constantly between 0.9 and 1.0 for both types of fixation. Time trends of specificity were also slightly negafive and homogenous, compared with the sensitivities. Figure 3 gives the mean values. The variability of values with time was again higher in the uncemented group.
Predictive values. With regard to loosening of the cup some types of pain were rarely diagnosed and therefore predictive values could not be calculated in all cases. PPVs increased during the first decade after operation from 0.00 to 0.66. The time-dependent variation was similar for both cemented and uncemented cups (Fig 4, Tables V and VI). NPVs decreased over time from 1.00 to 0.86. This decrease was relatively constant for uncemented cups whereas for cemented cups a relatively sharp decrease in NPV was observed at six years after operation (Fig. 5, Tables V and VI).
The calculated NPVs for loosening of the stem from one to ten years were constantly above 0.87 regardless of the year of follow-up and type of fixation of the stem. For both methods of fixation, the NPV at four years after operation was higher than at eight years (Fig. 5, Tables VI and VIII). PPVs varied considerably, especially in the uncemented group, and were rarely higher than 0.5 with a slight constant upward trend with time (Fig. 4, Tables VI and VIII).
Currently, there is some agreement that patients should be reviewed regularly after THA even if they are asymptomatic.11 This, however, involves elective and expensive procedures, which today are the subject of much discussion. Simank et a19 stated that from the economic and medical point of view, routine follow-up after THA could be abandoned and clinical review and revision procedures should be based purely on patient demand. For such a strategy, highly sensitive clinical indicators would be required to allow an accurate diagnosis of loosening without regular radiological imaging. The early detection and revision of a loose component are not only technically easier, but may improve the quality of life of a patient considerably because of the increased survival time of the revised component.10,12
Carlsson and Gentz22 found that the pain caused by a loose cup may be vague and tolerable. Wroblewski, Taylor and Siney23 and Wroblewski, Fleming and Siney24 concluded that mechanical stability could not be assessed on clinical examination alone and recommended regular radiological assessment to prevent gross loss of acetabular bone stock. Schmalzried et a125 also reported that acetabular loosening occurred earlier than femoral loosening and was often asymptomatic. Loosening of the stem is also often initially painless. The probability of loosening increases with the size of periprosthetic radiolucencies26 and often becomes clinically apparent as a sharp pain in the thigh or trochanteric area.22,27-29
The incidence of loosening of the cup and stem varies widely in the literature,2,6,9,10,30-34 mainly because of inconsistent radiological definitions of loosening and variability in the interpretation of x-ray findings. Brand et al19 showed that changing the radiological definition of loosening can double its reported incidence. Kramhoft et al35 found high intra- and interobserver variability, with kappa values ranging from 0.489 to 0.633 in the radiological assessment of the cup and from 0.737 to 0.800 for the stem. There are differences in the interpretation of both radiolucencies and migration leading to unreliable results and it is therefore not always possible to compare the outcome of different studies.36,37
In our study prevalence was used as a measure of the frequency of loosening of a component since the radiological evidence of loosening may be present for several years and may not be an indication for urgent revision. In addition, prevalence allows the direct calculation of a predictive value using the Bayes theorem.38 Therefore it is important to note that predictive values calculated in our study are a function of the observed prevalence. This has to be considered when these results are extrapolated to other patient populations. The use of multiple clinical tests for diagnosing loosening of components corresponds to a parallel testing scheme and independence of tests is a basic assumption, which is, however, not realistic in the setting of clinical follow-up. Multiple tests in parallel increase the sensitivity and the NPV for a given prevalence. Specificity and PPVs are decreased. Thus, in the clinical setting, tests interpreted in a parallel scheme will probably increase the overall sensitivity in the diagnosis of loosening at the cost of lower specificity.
The prevalences are directly related to the definition of radiological loosening of a component and taking into consideration the unreliability of the interpretation of radiographs and the adaptations which had to be made from the radiological definitions of loosening to the database classification, the limitations of such an analytical endeavour have to be clearly stated. Nevertheless, a constant increase in the prevalence of loosening of the cup with the passage of time was evident in our study. Also, the increased frequency of loosening of cemented compared with uncemented cups after the fifth year was highly significant and consistent. Assuming that the diagnostic error for both systems of implantation was the same, the increased frequency of loosening for cemented cups is an important finding. The MEM– CED data indicate a higher prevalence of loosening of the stem rather than the cup in the same population until the tenth year. This corresponds to the rates of loosening of the cemented Charnley hip arthroplasty reported by Kavanagh et al,39 which were 1.5 to 2.7 times higher for the stem than for the cup. Hozack et a140 calculated a ten-year survivorship of 91% for stems and 96% for cups. Ranawat et a12 stated that the clinical rates of failure of cemented cups were lower than those of stems. There are also studies, however, which report higher rates of failure for the cup than for the stem 41,42 depending on the design and type of the prosthetic component. This corresponds to the prevalences of component loosening after the tenth year observed in the MEM-CED database. Stauffer43 described a decreased prevalence of loosening of the stem after five years in patients with the Charnley prosthesis and a stabilisation of the prevalence of loosening of the stem after seven years is confirmed from our data (Fig. 1).
The probability of correctly diagnosing loosening of a component in asymptomatic patients without a radiological evaluation (sensitivity) was low and depended on the clinical examination which was used and on the type of fixation of the component. By contrast, specificities were very high during the first five years after operation and declined slightly with time.
The clear difference between sensitivities and specificities demonstrated the strengths and weaknesses of the clinical tests under study. The high specificities indicated that a clinical examination is very powerful in eliminating loosening of a component in an asymptomatic patient, but the test sensitivities for diagnosing loosening in a symptomatic patient were too low to replace a radiological evaluation as `the gold standard’.
An even better understanding of the validity of the clinical tests is provided by the post-test probabilities (PPV and NPV). They reflect the true probability of a loose component after a positive clinical finding and a well-fixed component after a negative clinical finding, respectively. Our analyses indicate very high NPVs in the diagnosis of loosening, especially during the first six years after operation (Fig. 5). Conversely, the PPVs were rather low depending on the clinical test (Fig. 4). The best clinical indicators for loosening of the cup were axial compression (PPV 0.21), external rotation (PPV 0.18) and hip pain (PPV 0. 18).
For femoral loosening the indicators were generally better with axial compression (PPV 0.29), external rotation (PPV 0.28), thigh pain (PPV 0.24), internal rotation (PPV 0.23), hip pain (PPV 0.23) and knee pain (PPV 0.21). These values clearly indicate that only a few patients with positive clinical findings have a loose component and therefore they all need further radiological investigation. By contrast, the very high specificities and NPVs of the clinical tests in asymptomatic patients justify an individualised follow-up strategy of THAs. Olsson, Jemberger and Tryggo44 stated 20 years ago that “radiographs were not made of all asymptomatic hips”. Our study supports their statement and we consider a radiological follow-up in order to monitor loosening of a component to be unnecessary in asymptomatic osteoarthritic patients during the first five postoperative years. Nevertheless, in the fifth to sixth postoperative year, periodic clinical and radiological reviews should be initiated even in asymptomatic patients. Patients who have spontaneous mild to severe pain, or pain on clinical testing at any postoperative point in time, require radiography. Also, patients who have new designs of implant with only short follow-up surveillance or designs with known complications such as accelerated wear, aggressive osteolysis or early loosening should be excluded from this follow-up protocol.
The authors thank E. R66sli, T. Ambrose and Dr A. Sharma for their support in the management of the data and technical questions and Professor M. E. Muller for his great endeavour in developing and applying a documentation system for THA at a time when outcome documentation was not very well received.
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
Copyright British Editorial Society of Bone & Joint Surgery Jan 2003
Provided by ProQuest Information and Learning Company. All rights Reserved.