- Email: [email protected]
Accurate diagnosis of hip prosthesis loosening using a vibrational technique
Clinical Biomechanics 16 (2001) 315±323
www.elsevier.com/locate/clinbiomech
Accurate diagnosis of hip prosthesis loosening using a vibrational technique A.P. Georgiou a, J.L. Cunningham b,* a
Department of Orthopaedic Surgery University of Bristol, Bristol Royal In®rmary, Bristol BS2 8HW, UK b Department of Mechanical Engineering, University of Bath, Bath BA2 7AY, UK Received 6 October 2000; accepted 10 October 2001
Abstract Objective. To examine the potential role of vibration testing as a non-invasive method of diagnosing loosening of total hip replacements in the clinical setting. Design. Single blind cohort study in two hospitals. Background. Diagnosing loosening of total hip replacements is heavily dependent on investigative techniques that are unreliable. Previous studies into the use of vibration testing have produced con¯icting results. Methods. Comparison of vibration testing and radiographs in patients with a total hip replacement experiencing hip pain symptomatic of loosening, with patients showing evidence of a secure prosthesis. Results. Vibration testing has a sensitivity of 80% and a speci®city of 89%. The positive predictive value was 92% and the negative predictive value was 73%; it was unable to produce a de®nitive diagnosis in 8% of patients. When compared with radiographs from the same patients, vibration testing was shown to be 20% more sensitive and able to diagnose 13% more patients. Conclusions. Vibration testing can deliver more accurate information on the stability of total hip replacements than radiographs in the clinical setting, despite being in the early stages of development. Relevance This study shows that 70 more patients may be provided with the correct diagnosis and 46 fewer patients may be undiagnosed each year, when using vibrometry as opposed to radiographs. In view of the relative disparity between the level of development between the two techniques and the encouraging results hitherto presented, it is felt that by improving vibration testing it may supersede radiographs in the detection of prosthesis loosening. Ó 2001 Elsevier Science Ltd. All rights reserved. Keywords: Total hip replacement; Loosening; Diagnosis; Vibration
1. Introduction Total hip replacement (THR) arthroplasty is currently the second most commonly performed elective surgical procedure in the UK. The most serious common complication of a THR is loosening of the prosthetic stem and cup; at 10 yr postoperatively, an estimated 0±7% of femoral components and 22±29% of acetabular components are loose [1,2]. Physical examination and clinical history have been shown to be inadequate in diagnosing loosening of THRs [3], so clinicians must rely heavily on the investigative techniques available. However, studies on the same
*
Corresponding author. E-mail address: [email protected] (J.L. Cunningham).
techniques by dierent authors have produced vastly dierent levels of accuracy (e.g. radiographs have been quoted as having accuracy's ranging from 50 [4] to 100% [5]), questioning the reliability of these investigations and the impending diagnosis based upon them. With the cost of a revision THR at $30,000 [6], there is the need for a reliable diagnostic tool for the dierentiation of secure and loose THRs. This report has examined the potential role of vibration (or vibrometry) as a non-invasive method for ful®lling this task. Previous studies into the potential use of this technique have involved the application of a sinusoidal input force to the lateral epicondyle of the femur, and examination of the output signal recorded at the greater trochanter. In vitro studies found the sensitivity of this technique was very poor in the detection of early
0268-0033/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 8 - 0 0 3 3 ( 0 1 ) 0 0 0 0 2 - X
316
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
loosening, the presence of a ®brous membrane or cement fractures [7,8]. However, a pilot clinical study showed this test to be 100% sensitive and 100% speci®c, both where arthrography was falsely negative and where clinical evaluation and radiographs were falsely positive [9]. The aim of this study was to conduct a trial on the eectiveness of vibrometry in the clinical setting. A comparison of the accuracy of this technique with other diagnostic tests provides information regarding the potential role for this technique in the clinical environment.
2. Methods Southmead Medical Research Ethics Committee gave ethical approval for this study, and informed consent was obtained from each patient before testing was carried out. 2.1. Vibration testing apparatus Patients for vibration testing lay on a bed on their side, exposing the lateral aspect of the femur that contained the prosthesis. The leg rested upon the contralateral leg, removing any muscular tension and allowing it to be vibrated with ease (see Fig. 1). A vibrator (Ling Dynamic Systems V201, Royston, Herts, UK) ®tted with a 19 mm spherical tip was held over the lateral epicondyle of the femur. Care was taken to ensure good bony contact in order to minimise high frequencies being ®ltered out by soft tissues, and to avoid distortion of the signal due to intermittent contact with the bone. The vibrator was driven by a sweep/ function generator (Thurlby Thandar TG230, Hun-
tingdon, Cambs, UK) connected to a power ampli®er (Ling Dynamic Systems PA25E, Royston, Herts, UK). The input waveform taken directly from the vibrator, was ampli®ed (PCB XDCR 480 D06, Depew, New York, USA), before being displayed on a two channel oscilloscope (Tektronix TAS220, Beaverton, Oregon, USA). The output waveform was detected by a low mass accelerometer (PCB Piezotronics 303A03, Depew, New York, USA), held over the greater trochanter of the femur, and care was again taken to obtain the best bony contact possible. The accelerometer was connected to an ampli®er (PCB-XDCR 480 D06, Depew, New York, USA), and then to the oscilloscope. From this display, any ®nal adjustments to the position of the vibrator or accelerometer were made in order to obtain a strong signal. Input and output data were recorded onto digital audio tapes (DAT) by a digital audio tape recorder (Pioneer D-05, Slough, Bucks, UK), for analysis at a later date. The position of the tape was noted before and after testing was carried out, so that the location of each patient's data on the tape was known. The input frequency was swept from 0 to 1000 Hz and back to 0 Hz over 15 times; greater than 3 min of data were recorded from each patient, to allow a reasonable average for the response to be obtained. 2.2. Patients for vibration testing Vibration data were collected from 23 patients (subjects) admitted to hospital for a revision THR, with clinical and radiographic signs indicative of loosening. The stability of their prosthesis was determined intraoperatively one day after vibration testing by a surgeon unaware of the vibrometry ®ndings. Vibration data were also collected from 10 patients who had undergone a primary THR several months previously, and who thus served as controls. Their prostheses were classi®ed as secure on the basis of the time since their primary operation together with clinical and radiographic criteria. The clinical criteria used in this study were assessment of pain, gait and range of motion; radiographic criteria included a tight ®t of the prosthesis within the bone [10] and the absence of bone cement or bone interface radiolucencies [1], or osteolysis. 2.3. Analysis of wave data
Fig. 1. Photograph of vibration testing. The vibrator is applied to the knee, and held at 90° to the femur in both planes. The accelerometer is applied to the subcutaneous greater trochanter.
In accordance with the previous work in this ®eld [7±9], two aspects of the vibration data were examined: 1. Purity of the output wave: This was examined by playing the DAT tape directly into a computer running Wavepak version 2.32 (Computational Systems, Knoxville, TN, USA), allowing examination of the
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
components of the output wave by fast fourier transform (FFT), in order to reveal the presence of harmonics. Each harmonic represents a component of a wave that is not sinusoidal, therefore being indicative of a loose prosthesis. The absence of harmonics implies a wave that is sinusoidal, and therefore generated from excitation of a secure prosthesis. 2. Number of resonant peaks: The number of resonant peaks in the output wave was examined by digitising the data (CTRAN version 2.1a, SSG, Washington DC, USA) and producing an amplitude/ frequency graph using Matlab for Windows 95 (Natick, MA, USA). The same procedure on the input wave data con®rmed a constant amplitude of the vibrator. Three blinded assessors who had no prior knowledge of this diagnostic technique assessed all vibrometry data. The criteria for this assessment (taken from the previous work in this ®eld [7±9]) was summarised for the assessors in an information sheet which included a method of evaluating the data presented to them, ensuring consistency and accuracy of interpretation. This assessment was made from a combined examination of the amplitude response graph and the FFT analysis. The criteria were: From the FFT analysis: · No harmonics or harmonics with amplitude of 25% or less that of the fundamental frequency ! secure prosthesis. It was felt that an amplitude below 25% would contribute suciently little to the output wave to making it inseparable from noise. · A harmonic with amplitude 50% or more of that of the fundamental frequency (especially applicable to the ®rst harmonic) ! loose prosthesis. It was felt that an amplitude of 50% or above must represent a signi®cant component of the output wave and could not be noise. · Five harmonics or more ! loose prosthesis. From the amplitude response analysis: · One resonant frequency ! secure prosthesis. · Two or more resonant frequencies ! loose prosthesis. The values of 25% and 50% were arbitrary and chosen for ease of interpretation. 2.4. Comparison of vibrometry with other diagnostic tests The stability of the prosthesis of each patient that underwent vibration testing was made by three blinded clinicians (all of registrar grade) based solely on the patient's radiographs. This evaluation was then compared with the same criteria used in determining the accuracy of vibrometry, allowing comparison of the accuracy of current investigations, with the accuracy of vibrometry for the same patients.
317
3. Results Patient details are summarised in Table 1. The mean age of subjects was 65.7 (SD 10.4) and the mean time since the patient's primary operation was 135.5 months (SD 64.8). Three patients who gave consent could not go through with vibrometry, due to the discomfort induced by the vibrator, or the inability to lie on their side. The mean age of controls was 58.1 (SD 17) and the mean time since the patient's primary operation was 21.9 months (SD 25.8). 3.1. Data Examples of the data obtained from vibrometry are illustrated in Figs. 2±5. 3.2. Comparison of vibrometry and radiographs Statistics regarding the accuracy of vibrometry and radiographs are illustrated in Fig. 6. Fig. 6 suggests that both vibrometry and radiographs can dierentiate between loose and secure prostheses in the clinical setting to a high degree of accuracy. Such a comparison is made particularly valid in that vibrometry and radiograph data are taken from the same patients. It is worth noting however the largest dierences between these techniques: · Vibrometry is 20% more sensitive than radiographs (80% compared to 60%). · Radiographs fail to deliver a de®nitive diagnosis in 21% of patients compared to 8% with vibrometry, a dierence of 13%. Combining the two diagnostic tests did not improve the accuracy of diagnosis. 3.3. Capacity for vibrometry to dierentiate loosening of dierent components Close examination of Table 1 suggests that loosening of the acetabular component is more dicult to detect than that of the femoral component or both components using this technique; only 50% of loose acetabular cups were correctly diagnosed using vibrometry. This suggests that loosening of dierent components signi®cantly in¯uences the mechanical properties of the system, and alludes to the possibility that the varying response obtained may present suciently dierent vibrometry data to allow dierentiation between loosening of dierent components. This was tested using the following components of the response: · The number of resonant frequencies. · The amplitude of ®rst harmonic. · The amplitude of the second harmonic. · The average amplitude of the remaining harmonics. · The number of remaining harmonics.
318
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
Table 1 Results of assessment of vibrometry data and radiographsa Patient number
Age at vibrometry test (yr)
Summary of patient details 1 42 2 78
a
Time since primary operation (months)
Intraoperative ®nding
Vibrometry accurate? (mean appraisal)
Radiographs accurate? (mean appraisal)
93 204
Loose femoral stem Loose acetabular component Loose femoral stem Notes unavailable NA (secure prosthesis) Notes unavailable NA (secure prosthesis) NA (secure prosthesis) Loose acetabular component Loose femoral stem NA (secure prosthesis) NA (secure prosthesis) NA (secure prosthesis) Loose acetabular component Loose femoral stem NA (secure prosthesis) Loose acetabular component Loose femoral stem Loose acetabular component NA (secure prosthesis) Loose femoral stem Loose acetabular component Loose acetabular component Loose acetabular component Loose femoral component Loose acetabular component Loose femoral component Loose acetabular component Loose femoral component Loose acetabular component NA (secure prosthesis) Notes missing Loose acetabular component Loose acetabular component Loose femoral component NA (secure prosthesis) Operation cancelled Loose acetabular component
Yes Yes
Yes No
? Indiscernible ? No Yes Yes
? Yes ? Yes Yes Yes
Yes Yes Yes Yes Yes Yes Yes
Yes Yes Yes D No Yes Indiscernible
Yes Yes Yes
Indiscernible Yes No
Indiscernible Yes
No Yes
Yes
Indiscernible
No
Yes
Yes Yes ? No Yes
Yes X-rays unavailable ? Indiscernible Yes
Yes ? No
X-rays unavailable X-rays unavailable X-rays unavailable
3 4 5 6 7 8
61 51 76 44 71 64
27 19 120
9 10 11 12 13 14 15
58 69 74 82 81 81 81
4 91 10 205 53 29 276
16 17 18
70 63 69
100 3 126
19 20
69 82
117 192
21
61
120
22
62
96
23 24 25 26 27
68 25 71 63 56
137 14 ? 18 113
28 29 30
45 80 74
10 ± 198
12
NA ± not applicable; D ± disagreement between assessors; Text in italics refers to control patients.
(All amplitudes were expressed as a percentage of the amplitude of the fundamental frequency, in order to gain an indication of the relative contributions of other harmonics to the formation of the output wave). The above values from all of the loose femoral components were compared with the values from all of the loose acetabular components and all of the values where both components were loose. The Mann±Whitney U test (Minitab for Windows 95, Minitab, State College, Pennsylvania, USA) showed there to be no signi®cant dierence at the P 0:05 level between any of the values examined, despite a visible trend. This result could be due to the small sample size.
4. Discussion More than 50,000 THR arthroplasties are performed in the UK and more than 400,000 are performed in the US each year. The most serious common complication of THR is loosening of the prosthetic stem and cup, the diagnosis of which is heavily dependent on investigative techniques [3] that continue to prove unreliable. With the cost of a revision THR at $30,000 [6], there is the need for a reliable diagnostic tool for the dierentiation of secure and loose THRs. This report has examined the potential role of vibration as a non-invasive method for ful®lling this task.
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
319
Fig. 2. A good example of an amplitude response graph from a secure prosthesis (patient 14). Only one resonant frequency can be seen (see arrow).
Fig. 3. A good example of an amplitude response graph from a loose prosthesis (patient 15). Two resonant frequencies can be seen (see arrows).
320
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
Fig. 4. A good example of an FFT analysis from a secure prosthesis (patient 9). Note the lack of harmonics of signi®cant amplitude in the output signal (top trace). The ®rst harmonic (1st) has amplitude less than 50% of the fundamental frequency (FF). The bottom trace shows the driving frequency (DF) of the vibrator.
Fig. 5. A good example of an FFT analysis from a loose prosthesis (patient 8). Note the large number of harmonics in the output signal (>5), and the size of their amplitudes in relation to that of the FF (top trace). The ®rst harmonic (1st) has amplitude greater than 50% that of the fundamental frequency. The bottom trace shows the driving frequency (DF) of the vibrator.
Vibration testing has been used in engineering for many years to determine the integrity of laminated structures such as boat hulls. The THR may itself be viewed as a laminated system, composed of bone, cement and prosthesis. Vibration or excitation of a securely implanted system produces an output char-
acteristic of a single unit having one resonant frequency and no harmonics (as the output wave is pure). Following loosening (or delamination), the separated elements of the composite respond separately, producing sustained superimposed output signals that may be separated by FFT analysis (into
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
321
Fig. 6. Comparison of statistics for radiographs and vibrometry. Positive predictive value (PPV) and negative predictive value (NPV) will change with the prevalence of loosening. `Indiscernible' represents the inability to dierentiate between a loose or a secure prosthesis; `disagreement' shows where assessors disagreed on the stability of the prosthesis. The sum of indiscernible and disagreement is interpreted as the percentage of occasions where a diagnosis is not made.
harmonics) and amplitude analysis (showing two or more resonant frequencies). Observations such as these drawn from in vitro studies [7±9] allude to the possibility that the dierence in response may be suciently great to provide a useful diagnostic tool in the clinical setting. In a single blind cohort study comparing patients with a THR experiencing hip pain symptomatic of loosening, with patients showing evidence of a secure prosthesis, this study has found that vibrometry is 20% more sensitive and can diagnose 13% more patients than radiographs. If 50,000 THRs are performed each year, and 7% of femoral components become loose in 10 yr [2], these ®gures correlate to 70 more patients being provided with the correct diagnosis and 46 fewer patients being undiagnosed each year, when using vibrometry as opposed to radiographs. These ®gures have wide ranging implications for both the patient and the health care provider, concluding that vibration testing can produce valuable diagnostic information on the stability of hip prostheses. Previous work in this ®eld has examined the capacity for vibrometry to diagnose loosening of the femoral component [7,8]. This is the ®rst study to look at the eectiveness of vibrometry where there is loosening of the acetabular component, or both components, in addition to that of the femoral component. It has been shown that a correct diagnosis of loosening is most often obtained when the femoral component is loose, or when both components are loose. Loose acetabular components are the most dicult to detect with this technique, accurately diagnosing 50% of patients, possibly because the loosened component is further from the source of excitation. Although this ®gure may sound disappoint-
ing, Greiner et al. [11] quote a sensitivity for arthrography of 48% and Zilkens et al. [4] quote a sensitivity for radiographs of 50%, in investigation of loosening of the acetabular component. It is therefore felt that this ®gure is promising for the current stage of development of this technique. The high incidence of acetabular component loosening when compared with that of femoral component loosening [1,2] increases the demand for a diagnostic tool, pro®cient in the dierential diagnosis of loosening of this component. Unfortunately, the ability of vibrometry to dierentiate between loose components was found to be statistically insigni®cant using selected wave criteria, despite a visible trend. It is likely that this can be attributed to the small sample size, and it is anticipated that a correlation will be found when the sample size is increased. 4.1. Limitations Vibrometry does bring with it some important limitations. These include: · It is unsuitable for patients who cannot lie on their side or who experience discomfort or pain induced by the vibrator, especially at low frequencies. In this study, 10% of subjects fell into this category. · In patients with knee prostheses, diagnosis by vibrometry must be interpreted with caution because the stability of the knee prosthesis and the eect of conduction of the signal through two prostheses is at present unknown. Further to this, the high level of accuracy of vibrometry and radiographs quoted above may in part be attributed to the fact that patients who entered the trial
322
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
were at either end of the broad clinical spectrum. Controls exhibited all signs of a successful THR; subjects showed all the signs of gross, mechanical loosening. A longitudinal study could be used to examine the eectiveness of vibrometry throughout the clinical spectrum, con®rming the suggestion that loosening may be objectively graded through a correlation with the number of harmonics, their size and the number of resonant frequencies (the greater the magnitude, the looser the implant). It is in the accurate diagnosis of cases that are currently a diagnostic challenge, that vibrometry may prove invaluable. Loosening is often radiographically diagnosed on progressive X-ray changes over time. As changes of vibrometry over time were not available, clinicians were only presented with one X-ray, so as not to bias diagnosis by radiographs, possibly explaining the high percentage of ``indiscernible'' radiographs and the disagreement between clinicians. Whilst a diagnosis of loosening is often dicult to make with a single X-ray, it may be made with comparable ease with a single vibrometry measurement, and it is in assessing this ``snapshot in time'' that the vibrational technique may also be particularly valuable. This in vivo trial has also highlighted some important dierences from the in vitro work of Li et al. [7,8] and Rosenstein et al. [9], which are not transferable to the clinical setting: (i) The presence of small harmonics in patients with secure prostheses is likely to be the result of a realistic transformation of the in vitro data into the clinical setting. The small amplitude of these harmonics relative to that of the fundamental frequency is felt to be indicative of their minor contributions to the wave. (ii) This study has shown the amplitude of the ®rst harmonic to be of signi®cant value in interpretation of the data. 4.2. Improving vibrometry Ongoing work in this department is looking at improved methods of vibrator and accelerometer application and novel methods of computerised data processing in analysis of the output wave. With improvements such as these it is anticipated that vibrometry may supersede radiographs in the diagnosis of prosthesis loosening. 4.3. Summary This study has shown that despite being in the early stages of development, vibrometry: (i) can diagnose loosening of hip prostheses to a high degree of accuracy in the clinical setting, being 20%
more sensitive, and diagnosing 13% more patients than radiographs; (ii) can provide an accurate diagnosis where no previous patient data is available. Diagnosis by radiographs is dependent on X-ray changes over time; (iii) is at least as accurate as current techniques in the diagnosis of loosening of the acetabular component. This is currently the greatest diagnostic challenge in this ®eld. (iv) shows early signs of being capable of distinguishing loosening of dierent components of the THR (although this was statistically insigni®cant in this study). (v) has the potential to grade the extent of loosening (as the parameters used in diagnosis are objective) providing extra information in cases that are currently a diagnostic challenge.
Acknowledgements The authors wish to thank Southmead Hospital Research Foundation for their ®nancial support, and Professor I.D. Learmonth and colleagues from the Department of Orthopaedic Surgery for their clinical assistance and participation. Many thanks also to Dr. N. Lieven, Mr. J.S. Clayton from the Department of Aerospace Engineering for their technical help and advice.
References [1] Harris WH. Osteolysis and particle disease in hip replacement. A review. Acta Othop Scand 1994;65:113±1231. [2] Weinstein SL, Buckwalter JA, editors. The adult hip; treatment of hip diseases. In: Turek's orthopaedics. Principles and their applications. JB Lippincott Company 1994;549±72. [3] Simank HG, Brocai DR, Rpde S, Lukoschek M. Diagnosis of hip prosthesis loosening ± sensitivity of clinical parameters. Z Orthop Ihre Grenzgeb 1998;136:39±43. [4] Zilkens KW, Wicke A, Zilkens J, Bull U. Nuclear imaging in loosening of hip endoprostheses. Arch Orthop Traum Surg 1988;107:288±92. [5] Lieberman JR, Huo MH, Schneider R, Salvati EA, Rodi S. Evaluation of painful hip arthroplasties. Are technetium bone scans necessary?. J Bone Joint Surg [Br] 1993;75B:475±8. [6] Lavernia CJ, Drakeford MK, Tsao AK, Gittelsohn A, Krackow KA, Hungerford DS. Revision and primary hip and knee arthroplasty a cost analysis. Clin Orthop 1995;311:136±41. [7] Li PLS, Jones NB, Gregg PJ. Vibration analysis in the detection of total hip prosthetic loosening. Med Eng Phys 1996;18:596±600. [8] Li PLS, Jones NB, Gregg PJ. Loosening of total hip arthroplasty. Diagnosis by vibration analysis. J Bone Joint Surg [Br] 1995;77B:640±4. [9] Rosenstein AD, McCoy GF, Bulstrode CJ, McLardy-Smith PD, Cunningham JL, Turner-Smith AR. The dierentiation of loose and secure femoral implants in total hip replacement using a
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323 vibrational technique: an anatomical and pilot clinical study. Proc Instn Mech Engr 1989;203:77±81. [10] Crenshaw AH, editor. Arthroplasty of hip. In: Cambell's operative orthopaedics. 7th ed. CV Mosby 1987;1213±501.
323
[11] Greiner B, Braun M, Becker W. Value of the plain roentgen image and arthrography in diagnosis of loosening of cemented hip edoprostheses. Z Orthop Ihre Grenzgeb 1997;135:292±6.
www.elsevier.com/locate/clinbiomech
Accurate diagnosis of hip prosthesis loosening using a vibrational technique A.P. Georgiou a, J.L. Cunningham b,* a
Department of Orthopaedic Surgery University of Bristol, Bristol Royal In®rmary, Bristol BS2 8HW, UK b Department of Mechanical Engineering, University of Bath, Bath BA2 7AY, UK Received 6 October 2000; accepted 10 October 2001
Abstract Objective. To examine the potential role of vibration testing as a non-invasive method of diagnosing loosening of total hip replacements in the clinical setting. Design. Single blind cohort study in two hospitals. Background. Diagnosing loosening of total hip replacements is heavily dependent on investigative techniques that are unreliable. Previous studies into the use of vibration testing have produced con¯icting results. Methods. Comparison of vibration testing and radiographs in patients with a total hip replacement experiencing hip pain symptomatic of loosening, with patients showing evidence of a secure prosthesis. Results. Vibration testing has a sensitivity of 80% and a speci®city of 89%. The positive predictive value was 92% and the negative predictive value was 73%; it was unable to produce a de®nitive diagnosis in 8% of patients. When compared with radiographs from the same patients, vibration testing was shown to be 20% more sensitive and able to diagnose 13% more patients. Conclusions. Vibration testing can deliver more accurate information on the stability of total hip replacements than radiographs in the clinical setting, despite being in the early stages of development. Relevance This study shows that 70 more patients may be provided with the correct diagnosis and 46 fewer patients may be undiagnosed each year, when using vibrometry as opposed to radiographs. In view of the relative disparity between the level of development between the two techniques and the encouraging results hitherto presented, it is felt that by improving vibration testing it may supersede radiographs in the detection of prosthesis loosening. Ó 2001 Elsevier Science Ltd. All rights reserved. Keywords: Total hip replacement; Loosening; Diagnosis; Vibration
1. Introduction Total hip replacement (THR) arthroplasty is currently the second most commonly performed elective surgical procedure in the UK. The most serious common complication of a THR is loosening of the prosthetic stem and cup; at 10 yr postoperatively, an estimated 0±7% of femoral components and 22±29% of acetabular components are loose [1,2]. Physical examination and clinical history have been shown to be inadequate in diagnosing loosening of THRs [3], so clinicians must rely heavily on the investigative techniques available. However, studies on the same
*
Corresponding author. E-mail address: [email protected] (J.L. Cunningham).
techniques by dierent authors have produced vastly dierent levels of accuracy (e.g. radiographs have been quoted as having accuracy's ranging from 50 [4] to 100% [5]), questioning the reliability of these investigations and the impending diagnosis based upon them. With the cost of a revision THR at $30,000 [6], there is the need for a reliable diagnostic tool for the dierentiation of secure and loose THRs. This report has examined the potential role of vibration (or vibrometry) as a non-invasive method for ful®lling this task. Previous studies into the potential use of this technique have involved the application of a sinusoidal input force to the lateral epicondyle of the femur, and examination of the output signal recorded at the greater trochanter. In vitro studies found the sensitivity of this technique was very poor in the detection of early
0268-0033/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 8 - 0 0 3 3 ( 0 1 ) 0 0 0 0 2 - X
316
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
loosening, the presence of a ®brous membrane or cement fractures [7,8]. However, a pilot clinical study showed this test to be 100% sensitive and 100% speci®c, both where arthrography was falsely negative and where clinical evaluation and radiographs were falsely positive [9]. The aim of this study was to conduct a trial on the eectiveness of vibrometry in the clinical setting. A comparison of the accuracy of this technique with other diagnostic tests provides information regarding the potential role for this technique in the clinical environment.
2. Methods Southmead Medical Research Ethics Committee gave ethical approval for this study, and informed consent was obtained from each patient before testing was carried out. 2.1. Vibration testing apparatus Patients for vibration testing lay on a bed on their side, exposing the lateral aspect of the femur that contained the prosthesis. The leg rested upon the contralateral leg, removing any muscular tension and allowing it to be vibrated with ease (see Fig. 1). A vibrator (Ling Dynamic Systems V201, Royston, Herts, UK) ®tted with a 19 mm spherical tip was held over the lateral epicondyle of the femur. Care was taken to ensure good bony contact in order to minimise high frequencies being ®ltered out by soft tissues, and to avoid distortion of the signal due to intermittent contact with the bone. The vibrator was driven by a sweep/ function generator (Thurlby Thandar TG230, Hun-
tingdon, Cambs, UK) connected to a power ampli®er (Ling Dynamic Systems PA25E, Royston, Herts, UK). The input waveform taken directly from the vibrator, was ampli®ed (PCB XDCR 480 D06, Depew, New York, USA), before being displayed on a two channel oscilloscope (Tektronix TAS220, Beaverton, Oregon, USA). The output waveform was detected by a low mass accelerometer (PCB Piezotronics 303A03, Depew, New York, USA), held over the greater trochanter of the femur, and care was again taken to obtain the best bony contact possible. The accelerometer was connected to an ampli®er (PCB-XDCR 480 D06, Depew, New York, USA), and then to the oscilloscope. From this display, any ®nal adjustments to the position of the vibrator or accelerometer were made in order to obtain a strong signal. Input and output data were recorded onto digital audio tapes (DAT) by a digital audio tape recorder (Pioneer D-05, Slough, Bucks, UK), for analysis at a later date. The position of the tape was noted before and after testing was carried out, so that the location of each patient's data on the tape was known. The input frequency was swept from 0 to 1000 Hz and back to 0 Hz over 15 times; greater than 3 min of data were recorded from each patient, to allow a reasonable average for the response to be obtained. 2.2. Patients for vibration testing Vibration data were collected from 23 patients (subjects) admitted to hospital for a revision THR, with clinical and radiographic signs indicative of loosening. The stability of their prosthesis was determined intraoperatively one day after vibration testing by a surgeon unaware of the vibrometry ®ndings. Vibration data were also collected from 10 patients who had undergone a primary THR several months previously, and who thus served as controls. Their prostheses were classi®ed as secure on the basis of the time since their primary operation together with clinical and radiographic criteria. The clinical criteria used in this study were assessment of pain, gait and range of motion; radiographic criteria included a tight ®t of the prosthesis within the bone [10] and the absence of bone cement or bone interface radiolucencies [1], or osteolysis. 2.3. Analysis of wave data
Fig. 1. Photograph of vibration testing. The vibrator is applied to the knee, and held at 90° to the femur in both planes. The accelerometer is applied to the subcutaneous greater trochanter.
In accordance with the previous work in this ®eld [7±9], two aspects of the vibration data were examined: 1. Purity of the output wave: This was examined by playing the DAT tape directly into a computer running Wavepak version 2.32 (Computational Systems, Knoxville, TN, USA), allowing examination of the
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
components of the output wave by fast fourier transform (FFT), in order to reveal the presence of harmonics. Each harmonic represents a component of a wave that is not sinusoidal, therefore being indicative of a loose prosthesis. The absence of harmonics implies a wave that is sinusoidal, and therefore generated from excitation of a secure prosthesis. 2. Number of resonant peaks: The number of resonant peaks in the output wave was examined by digitising the data (CTRAN version 2.1a, SSG, Washington DC, USA) and producing an amplitude/ frequency graph using Matlab for Windows 95 (Natick, MA, USA). The same procedure on the input wave data con®rmed a constant amplitude of the vibrator. Three blinded assessors who had no prior knowledge of this diagnostic technique assessed all vibrometry data. The criteria for this assessment (taken from the previous work in this ®eld [7±9]) was summarised for the assessors in an information sheet which included a method of evaluating the data presented to them, ensuring consistency and accuracy of interpretation. This assessment was made from a combined examination of the amplitude response graph and the FFT analysis. The criteria were: From the FFT analysis: · No harmonics or harmonics with amplitude of 25% or less that of the fundamental frequency ! secure prosthesis. It was felt that an amplitude below 25% would contribute suciently little to the output wave to making it inseparable from noise. · A harmonic with amplitude 50% or more of that of the fundamental frequency (especially applicable to the ®rst harmonic) ! loose prosthesis. It was felt that an amplitude of 50% or above must represent a signi®cant component of the output wave and could not be noise. · Five harmonics or more ! loose prosthesis. From the amplitude response analysis: · One resonant frequency ! secure prosthesis. · Two or more resonant frequencies ! loose prosthesis. The values of 25% and 50% were arbitrary and chosen for ease of interpretation. 2.4. Comparison of vibrometry with other diagnostic tests The stability of the prosthesis of each patient that underwent vibration testing was made by three blinded clinicians (all of registrar grade) based solely on the patient's radiographs. This evaluation was then compared with the same criteria used in determining the accuracy of vibrometry, allowing comparison of the accuracy of current investigations, with the accuracy of vibrometry for the same patients.
317
3. Results Patient details are summarised in Table 1. The mean age of subjects was 65.7 (SD 10.4) and the mean time since the patient's primary operation was 135.5 months (SD 64.8). Three patients who gave consent could not go through with vibrometry, due to the discomfort induced by the vibrator, or the inability to lie on their side. The mean age of controls was 58.1 (SD 17) and the mean time since the patient's primary operation was 21.9 months (SD 25.8). 3.1. Data Examples of the data obtained from vibrometry are illustrated in Figs. 2±5. 3.2. Comparison of vibrometry and radiographs Statistics regarding the accuracy of vibrometry and radiographs are illustrated in Fig. 6. Fig. 6 suggests that both vibrometry and radiographs can dierentiate between loose and secure prostheses in the clinical setting to a high degree of accuracy. Such a comparison is made particularly valid in that vibrometry and radiograph data are taken from the same patients. It is worth noting however the largest dierences between these techniques: · Vibrometry is 20% more sensitive than radiographs (80% compared to 60%). · Radiographs fail to deliver a de®nitive diagnosis in 21% of patients compared to 8% with vibrometry, a dierence of 13%. Combining the two diagnostic tests did not improve the accuracy of diagnosis. 3.3. Capacity for vibrometry to dierentiate loosening of dierent components Close examination of Table 1 suggests that loosening of the acetabular component is more dicult to detect than that of the femoral component or both components using this technique; only 50% of loose acetabular cups were correctly diagnosed using vibrometry. This suggests that loosening of dierent components signi®cantly in¯uences the mechanical properties of the system, and alludes to the possibility that the varying response obtained may present suciently dierent vibrometry data to allow dierentiation between loosening of dierent components. This was tested using the following components of the response: · The number of resonant frequencies. · The amplitude of ®rst harmonic. · The amplitude of the second harmonic. · The average amplitude of the remaining harmonics. · The number of remaining harmonics.
318
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
Table 1 Results of assessment of vibrometry data and radiographsa Patient number
Age at vibrometry test (yr)
Summary of patient details 1 42 2 78
a
Time since primary operation (months)
Intraoperative ®nding
Vibrometry accurate? (mean appraisal)
Radiographs accurate? (mean appraisal)
93 204
Loose femoral stem Loose acetabular component Loose femoral stem Notes unavailable NA (secure prosthesis) Notes unavailable NA (secure prosthesis) NA (secure prosthesis) Loose acetabular component Loose femoral stem NA (secure prosthesis) NA (secure prosthesis) NA (secure prosthesis) Loose acetabular component Loose femoral stem NA (secure prosthesis) Loose acetabular component Loose femoral stem Loose acetabular component NA (secure prosthesis) Loose femoral stem Loose acetabular component Loose acetabular component Loose acetabular component Loose femoral component Loose acetabular component Loose femoral component Loose acetabular component Loose femoral component Loose acetabular component NA (secure prosthesis) Notes missing Loose acetabular component Loose acetabular component Loose femoral component NA (secure prosthesis) Operation cancelled Loose acetabular component
Yes Yes
Yes No
? Indiscernible ? No Yes Yes
? Yes ? Yes Yes Yes
Yes Yes Yes Yes Yes Yes Yes
Yes Yes Yes D No Yes Indiscernible
Yes Yes Yes
Indiscernible Yes No
Indiscernible Yes
No Yes
Yes
Indiscernible
No
Yes
Yes Yes ? No Yes
Yes X-rays unavailable ? Indiscernible Yes
Yes ? No
X-rays unavailable X-rays unavailable X-rays unavailable
3 4 5 6 7 8
61 51 76 44 71 64
27 19 120
9 10 11 12 13 14 15
58 69 74 82 81 81 81
4 91 10 205 53 29 276
16 17 18
70 63 69
100 3 126
19 20
69 82
117 192
21
61
120
22
62
96
23 24 25 26 27
68 25 71 63 56
137 14 ? 18 113
28 29 30
45 80 74
10 ± 198
12
NA ± not applicable; D ± disagreement between assessors; Text in italics refers to control patients.
(All amplitudes were expressed as a percentage of the amplitude of the fundamental frequency, in order to gain an indication of the relative contributions of other harmonics to the formation of the output wave). The above values from all of the loose femoral components were compared with the values from all of the loose acetabular components and all of the values where both components were loose. The Mann±Whitney U test (Minitab for Windows 95, Minitab, State College, Pennsylvania, USA) showed there to be no signi®cant dierence at the P 0:05 level between any of the values examined, despite a visible trend. This result could be due to the small sample size.
4. Discussion More than 50,000 THR arthroplasties are performed in the UK and more than 400,000 are performed in the US each year. The most serious common complication of THR is loosening of the prosthetic stem and cup, the diagnosis of which is heavily dependent on investigative techniques [3] that continue to prove unreliable. With the cost of a revision THR at $30,000 [6], there is the need for a reliable diagnostic tool for the dierentiation of secure and loose THRs. This report has examined the potential role of vibration as a non-invasive method for ful®lling this task.
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
319
Fig. 2. A good example of an amplitude response graph from a secure prosthesis (patient 14). Only one resonant frequency can be seen (see arrow).
Fig. 3. A good example of an amplitude response graph from a loose prosthesis (patient 15). Two resonant frequencies can be seen (see arrows).
320
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
Fig. 4. A good example of an FFT analysis from a secure prosthesis (patient 9). Note the lack of harmonics of signi®cant amplitude in the output signal (top trace). The ®rst harmonic (1st) has amplitude less than 50% of the fundamental frequency (FF). The bottom trace shows the driving frequency (DF) of the vibrator.
Fig. 5. A good example of an FFT analysis from a loose prosthesis (patient 8). Note the large number of harmonics in the output signal (>5), and the size of their amplitudes in relation to that of the FF (top trace). The ®rst harmonic (1st) has amplitude greater than 50% that of the fundamental frequency. The bottom trace shows the driving frequency (DF) of the vibrator.
Vibration testing has been used in engineering for many years to determine the integrity of laminated structures such as boat hulls. The THR may itself be viewed as a laminated system, composed of bone, cement and prosthesis. Vibration or excitation of a securely implanted system produces an output char-
acteristic of a single unit having one resonant frequency and no harmonics (as the output wave is pure). Following loosening (or delamination), the separated elements of the composite respond separately, producing sustained superimposed output signals that may be separated by FFT analysis (into
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
321
Fig. 6. Comparison of statistics for radiographs and vibrometry. Positive predictive value (PPV) and negative predictive value (NPV) will change with the prevalence of loosening. `Indiscernible' represents the inability to dierentiate between a loose or a secure prosthesis; `disagreement' shows where assessors disagreed on the stability of the prosthesis. The sum of indiscernible and disagreement is interpreted as the percentage of occasions where a diagnosis is not made.
harmonics) and amplitude analysis (showing two or more resonant frequencies). Observations such as these drawn from in vitro studies [7±9] allude to the possibility that the dierence in response may be suciently great to provide a useful diagnostic tool in the clinical setting. In a single blind cohort study comparing patients with a THR experiencing hip pain symptomatic of loosening, with patients showing evidence of a secure prosthesis, this study has found that vibrometry is 20% more sensitive and can diagnose 13% more patients than radiographs. If 50,000 THRs are performed each year, and 7% of femoral components become loose in 10 yr [2], these ®gures correlate to 70 more patients being provided with the correct diagnosis and 46 fewer patients being undiagnosed each year, when using vibrometry as opposed to radiographs. These ®gures have wide ranging implications for both the patient and the health care provider, concluding that vibration testing can produce valuable diagnostic information on the stability of hip prostheses. Previous work in this ®eld has examined the capacity for vibrometry to diagnose loosening of the femoral component [7,8]. This is the ®rst study to look at the eectiveness of vibrometry where there is loosening of the acetabular component, or both components, in addition to that of the femoral component. It has been shown that a correct diagnosis of loosening is most often obtained when the femoral component is loose, or when both components are loose. Loose acetabular components are the most dicult to detect with this technique, accurately diagnosing 50% of patients, possibly because the loosened component is further from the source of excitation. Although this ®gure may sound disappoint-
ing, Greiner et al. [11] quote a sensitivity for arthrography of 48% and Zilkens et al. [4] quote a sensitivity for radiographs of 50%, in investigation of loosening of the acetabular component. It is therefore felt that this ®gure is promising for the current stage of development of this technique. The high incidence of acetabular component loosening when compared with that of femoral component loosening [1,2] increases the demand for a diagnostic tool, pro®cient in the dierential diagnosis of loosening of this component. Unfortunately, the ability of vibrometry to dierentiate between loose components was found to be statistically insigni®cant using selected wave criteria, despite a visible trend. It is likely that this can be attributed to the small sample size, and it is anticipated that a correlation will be found when the sample size is increased. 4.1. Limitations Vibrometry does bring with it some important limitations. These include: · It is unsuitable for patients who cannot lie on their side or who experience discomfort or pain induced by the vibrator, especially at low frequencies. In this study, 10% of subjects fell into this category. · In patients with knee prostheses, diagnosis by vibrometry must be interpreted with caution because the stability of the knee prosthesis and the eect of conduction of the signal through two prostheses is at present unknown. Further to this, the high level of accuracy of vibrometry and radiographs quoted above may in part be attributed to the fact that patients who entered the trial
322
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323
were at either end of the broad clinical spectrum. Controls exhibited all signs of a successful THR; subjects showed all the signs of gross, mechanical loosening. A longitudinal study could be used to examine the eectiveness of vibrometry throughout the clinical spectrum, con®rming the suggestion that loosening may be objectively graded through a correlation with the number of harmonics, their size and the number of resonant frequencies (the greater the magnitude, the looser the implant). It is in the accurate diagnosis of cases that are currently a diagnostic challenge, that vibrometry may prove invaluable. Loosening is often radiographically diagnosed on progressive X-ray changes over time. As changes of vibrometry over time were not available, clinicians were only presented with one X-ray, so as not to bias diagnosis by radiographs, possibly explaining the high percentage of ``indiscernible'' radiographs and the disagreement between clinicians. Whilst a diagnosis of loosening is often dicult to make with a single X-ray, it may be made with comparable ease with a single vibrometry measurement, and it is in assessing this ``snapshot in time'' that the vibrational technique may also be particularly valuable. This in vivo trial has also highlighted some important dierences from the in vitro work of Li et al. [7,8] and Rosenstein et al. [9], which are not transferable to the clinical setting: (i) The presence of small harmonics in patients with secure prostheses is likely to be the result of a realistic transformation of the in vitro data into the clinical setting. The small amplitude of these harmonics relative to that of the fundamental frequency is felt to be indicative of their minor contributions to the wave. (ii) This study has shown the amplitude of the ®rst harmonic to be of signi®cant value in interpretation of the data. 4.2. Improving vibrometry Ongoing work in this department is looking at improved methods of vibrator and accelerometer application and novel methods of computerised data processing in analysis of the output wave. With improvements such as these it is anticipated that vibrometry may supersede radiographs in the diagnosis of prosthesis loosening. 4.3. Summary This study has shown that despite being in the early stages of development, vibrometry: (i) can diagnose loosening of hip prostheses to a high degree of accuracy in the clinical setting, being 20%
more sensitive, and diagnosing 13% more patients than radiographs; (ii) can provide an accurate diagnosis where no previous patient data is available. Diagnosis by radiographs is dependent on X-ray changes over time; (iii) is at least as accurate as current techniques in the diagnosis of loosening of the acetabular component. This is currently the greatest diagnostic challenge in this ®eld. (iv) shows early signs of being capable of distinguishing loosening of dierent components of the THR (although this was statistically insigni®cant in this study). (v) has the potential to grade the extent of loosening (as the parameters used in diagnosis are objective) providing extra information in cases that are currently a diagnostic challenge.
Acknowledgements The authors wish to thank Southmead Hospital Research Foundation for their ®nancial support, and Professor I.D. Learmonth and colleagues from the Department of Orthopaedic Surgery for their clinical assistance and participation. Many thanks also to Dr. N. Lieven, Mr. J.S. Clayton from the Department of Aerospace Engineering for their technical help and advice.
References [1] Harris WH. Osteolysis and particle disease in hip replacement. A review. Acta Othop Scand 1994;65:113±1231. [2] Weinstein SL, Buckwalter JA, editors. The adult hip; treatment of hip diseases. In: Turek's orthopaedics. Principles and their applications. JB Lippincott Company 1994;549±72. [3] Simank HG, Brocai DR, Rpde S, Lukoschek M. Diagnosis of hip prosthesis loosening ± sensitivity of clinical parameters. Z Orthop Ihre Grenzgeb 1998;136:39±43. [4] Zilkens KW, Wicke A, Zilkens J, Bull U. Nuclear imaging in loosening of hip endoprostheses. Arch Orthop Traum Surg 1988;107:288±92. [5] Lieberman JR, Huo MH, Schneider R, Salvati EA, Rodi S. Evaluation of painful hip arthroplasties. Are technetium bone scans necessary?. J Bone Joint Surg [Br] 1993;75B:475±8. [6] Lavernia CJ, Drakeford MK, Tsao AK, Gittelsohn A, Krackow KA, Hungerford DS. Revision and primary hip and knee arthroplasty a cost analysis. Clin Orthop 1995;311:136±41. [7] Li PLS, Jones NB, Gregg PJ. Vibration analysis in the detection of total hip prosthetic loosening. Med Eng Phys 1996;18:596±600. [8] Li PLS, Jones NB, Gregg PJ. Loosening of total hip arthroplasty. Diagnosis by vibration analysis. J Bone Joint Surg [Br] 1995;77B:640±4. [9] Rosenstein AD, McCoy GF, Bulstrode CJ, McLardy-Smith PD, Cunningham JL, Turner-Smith AR. The dierentiation of loose and secure femoral implants in total hip replacement using a
A.P. Georgiou, J.L. Cunningham / Clinical Biomechanics 16 (2001) 315±323 vibrational technique: an anatomical and pilot clinical study. Proc Instn Mech Engr 1989;203:77±81. [10] Crenshaw AH, editor. Arthroplasty of hip. In: Cambell's operative orthopaedics. 7th ed. CV Mosby 1987;1213±501.
323
[11] Greiner B, Braun M, Becker W. Value of the plain roentgen image and arthrography in diagnosis of loosening of cemented hip edoprostheses. Z Orthop Ihre Grenzgeb 1997;135:292±6.