Introduction

The imaging modalities used for assessment of painful hip and knee prostheses with suspected loosening include plain radiography, contrast arthrography, local bone scan and planar radionuclide arthrography. For assessment of the femoral component of the total hip prosthesis, a recent meta-analysis [1] suggested no significant difference in the accuracy of these modalities. The literature to date on radionuclide arthrography (RNA) has described the use of planar but not tomographic imaging.

Studies using planar RNA for assessment of acetabular cup loosening showed inferior accuracy figures [2–5], and concluded that it is not a suitable technique on its own. We were unable to find published accounts of RNA using SPECT with or without CT image fusion for the evaluation of acetabular cup loosening.

For assessment of total knee replacement loosening, the role of planar RNA was supported by a recent publication of Kitchener et al. [15]. Another earlier publication by Namasivayam et al. [6] suggested that there is possibly a complementary role for planar RNA. We were unable to find published accounts of RNA using SPECT with or without CT image fusion for evaluation of knee prosthesis loosening.

This is a retrospective audit of the use of RNA SPECT with anatomical localization provided by CT image fusion for evaluation of painful hip and knee prostheses suspected of mechanical loosening.

Materials and methods

Subjects

All patients referred for investigation of implant loosening between January 2005 and September 2007 were included in the study. Ethical approval was obtained from the Calvary Wakefield Research and Ethics Committee. Written informed consent was obtained from all subjects prior to recruitment. 119 patients were referred. Two patients were lost to follow-up and one was still awaiting clinical review at the conclusion of the audit period; these latter three subjects were excluded in the final analysis. One patient had both knees scanned. In total the clinical outcomes of 117 joints in 116 patients were available for examination.

The referrals for RNA SPECT/CT were based upon history, examination and preliminary radiological investigations, which included standardized X-rays that were interpreted for loosening using the DeLee and Charnley score for the acetabulum and Gruen zones of the femur for hips. Similarly, the presence of progressive periprosthetic lucent lines, or those greater than 2 mm in width adjacent to knee prosthetic components were thought to indicate loosening.

Imaging

Under fluoroscopic guidance, using a sterile technique, a 22-G spinal needle was inserted into the joint. Whenever possible, a joint aspiration of 2 ml of synovial fluid was sent for microbiological examination. The needle position was confirmed with radiographic contrast material (Isovue® 300, iopamidol 612 mg in 10 ml, Bracco Pty. Ltd., North Sydney, NSW, Australia) followed by injection of 20 MBq of 99mTc Calcium Phytate (Radpharm Scientific, Belconnen, ACT, Australia). The volume of the radiopharmaceutical was 3 ml for knees and 5 ml for hips. The patient was then mobilised and encouraged to bear weight on the injected joint.

Shortly after radiotracer administration planar images were acquired in the anterior, posterior, anterior oblique and posterior oblique projections for hips. Anterior, posterior, medial and lateral projections were used for knees. An anterior planar image acquired with a concurrent transmission 57Co sheet source was performed to aid anatomical localization. The planar images were repeated at 3 h post injection. Subsequently a SPECT scan was performed with CT co-registration using a hybrid gamma camera/CT scanner (GE Hawkeye VG).

The planar images were acquired on a single-headed (GE isocam 500 or Siemens e.cam) or a double-headed (GE Hawkeye VG) gamma camera using a LEHR collimator, 256 × 256 matrix, with acquisition time of 3 min per image.

The SPECT scan was performed with the dual-headed gamma camera, using a LEHR collimator, 180° body contoured orbit, 120 frames (11 s/frame), 128 × 128 matrix, and processed with a Butterworth filter (threshold 0.35, power 10) for the first 22 patients. Following a departmental upgrade, further studies were processed with iterative reconstruction (2 iterations, 10 subsets). For the purpose of RNA SPECT CT interpretation this upgrade was not felt to have materially changed the quality of the images reviewed.

The CT was a low-energy scan to provide an anatomical map that was not of diagnostic quality. The parameters of the CT included a single detector, step and shoot acquisition, effective mAs of 2.5, 140 kV, 10 mm/step, 14 s/step + 1 s totalling 10 min.

Scan reporting and follow up

All of the RNA SPECT/CTs were read by one experienced nuclear medicine physician, who was not blinded to the clinical history and previous imaging.

The planar images were reviewed on hard copies when formulating a report for the surgeon. All SPECT/CT studies were reported from images reviewed on the Xeleris 2 (GE healthcare) workstation. The triangulation tool was used to localise the bone/prosthesis interfaces on the anatomical map provided by the CT. The criterion used for diagnosis of prosthetic loosening was detection of the injected radiotracer within any part of the interface on the corresponding SPECT slices.

As the studies were performed in a clinical setting, the surgeons were not blinded to the results of the scintigraphic examinations. When issuing a formal clinical report, all images were reviewed but the final conclusion was determined by the SPECT/CT assessment. Retrospectively, for the purpose of comparison with SPECT/CT, planar images that were accessible were reassessed for loosening, using the same diagnostic criterion as the SPECT/CT. For this reassessment the planar images were reviewed electronically on the same workstation. The reader was blinded to the SPECT/CT results when rescoring the planar scans for loosening.

Patients were followed for 1 year after the scan. Operative evaluation was used as the gold standard for assessment of implant loosening. Patients too unfit to be considered for surgery were excluded from the analysis. In the 12 months of follow-up, patients who were not reoperated on were also excluded from the analysis.

A binary classification test was performed on the data to determine the sensitivity, specificity, positive (PPV) and negative (NPV) predictive values. A 95% confidence interval for each was also determined. McNemar’s test was used to detect a statistically significant difference between the SPECT/CT and planar results.

Results

117 joints (42 hips and 75 knees) were assessed––in 116 subjects, 1 of whom had both knees scanned.

Hips

Of the 42 hip arthroplasties one was a hemiarthroplasty (i.e. femoral stem only). The time from implantation of the hip prostheses was known for 39 subjects. The recorded durations ranged from 5 to 234 months (median 41 months).

Only 29 of the 42 subjects were reoperated on. 24 of these had a revision of their prostheses. As the sole patient with a hemiarthroplasty was among the reoperated group there were 29 femoral stems and 28 acetabular cups available for analyses. 20 of the 29 hip prostheses were implanted without the need for bone cement for fixation. The RNA SPECT CT assessments were compared with the surgical findings for all components of the hip prostheses. The results of this comparison are as outlined in Table 1.

Table 1 SPECT/CT versus planar assessment of hip prostheses
Full size table

Of the hips that required re operation, the planar images of 26 (including the 1 hemiarthroplasty) were available for assessment, and the results are as outlined in Table 1. The reader was unable to detect loosening in any of the acetabular cups on planar imaging. As such, the positive predictive value could not be determined.

Knees

Of the 75 knee replacements, there were 12 uni-compartmental prostheses and 63 total knee replacements. The duration of implantation of the knee prostheses was known for 71 subjects. The recorded durations ranged from 3 to 195 months (median 25 months).

34 of the 63 patients with total knee replacements underwent reoperation, resulting in the revision of 27 of the prostheses while 6 of these subjects had arthroscopies and one subject had a screw removed. Of the 12 unicompartmental knee replacements 10 underwent reoperation, 4 were revised and 6 had arthroscopic assessment. Only 16 (8 total and 8 unicompartmental replacements) of the 44 reoperated knee prostheses were cemented.

The RNA SPECT/CT assessments were compared with the surgical findings of all components of the knee prostheses and the results are outlined in Table 2.

Table 2 SPECT/CT versus planar assessment of knee prostheses
Full size table

Of the knee prostheses that required re operation, the planar images of 42 (including the 9 unicompartmental arthroplasties) were available for reassessment, and the results are as outlined in Table 2.

Joint aspirates of 1 knee and 2 hip prostheses were culture positive, consistent with bacterial infection resulting in a change of management to joint lavage and antibiotic suppression. Two other hip prostheses with positive cultures were thought to represent contamination.

Discussion

There is ample literature supporting the use of planar RNA in the evaluation of femoral stem loosening, with reported sensitivities of 76–90%, and specificities of up to 92% [7–13]. A more recent meta analysis by Temmerman et al. [1] indicated that for planar RNA the sensitivity of 85% and specificity of 83% are equivalent to the other imaging modalities of plain X-ray, contrast arthrography and local bone scan. In RNA assessment of the femoral stem there is no interference from the site of radiotracer injection within the joint pseudocapsule.

Unlike the femoral stem planar RNA assessment for acetabular cup loosening has multiple limiting factors. The first is the lack of anatomical landmarks to differentiate the joint space from the bone/metal interface of the cup. Second, there is marked interference from the injected radiotracer within the joint space Third, in two-dimensional planar images peri-prosthetic tracer accumulation located tangential to the curved margin of the cup in different projections is difficult to identify. Thus publications to date on planar RNA of acetabular cups show poor results. Temmerman et al. [3] in their series comparing with plain X-rays and digital subtraction arthrography, showed that planar RNA had a sensitivity of 57% and specificity of 67%. Planar RNA has not been sufficiently accurate to use as the sole mode of investigation in the evaluation of acetabular implant loosening [2, 4, 5]. In this audit the reader was unable to detect loosening in any of the acetabular cups using planar images.

We sought to evaluate the value of SPECT/CT. SPECT allowed a three-dimensional examination of the prostheses. The CT scan provided a reliable anatomical map to determine the exact location of the bone/metal interface in which to evaluate for radiotracer accumulation on the SPECT scan (Fig. 1). Any metallic artefacts on the low-energy CT were of no consequence as the images were not meant to be of diagnostic value. In our experience the metallic artefacts did not interfere with the ability to localise the bone/metal interface.

Fig. 1
figure 1

Radiotracer is demonstrated within the bone/metal interface of the acetabular cup of a left total hip replacement indicative of loosening

Full size image

Even though in SPECT scanning the increased patient distance from the camera resulted in a lower spatial resolution, the contrast resolution was enhanced with the tomographic reconstruction which helped in the detection of very low-grade radioactivity in the bone/prosthesis interface.

To the best of our knowledge, there is no published series on RNA with SPECT/CT for assessment of hip prostheses. In our series, the SPECT/CT results for the assessment of the femoral stem were equivalent to published data using planar imaging. Our SPECT/CT results were also equivalent to our planar imaging results.

For the assessment of the acetabular cup, our SPECT/CT results were better than the published results for planar imaging. Compared with our planar results, SPECT/CT was superior, and the difference was statistically significant. The strength of the SPECT/CT results appeared to be in their negative predictive value.

For suboptimal radiotracer administration, SPECT/CT helps. An extra articular injection gave a false impression of a loose femoral prosthesis in our subject with a hip hemiarthroplasty (Fig. 2) on planar images. The SPECT/CT localised the radiotracer to the psoas bursa, thus avoiding a false-positive assessment. The study was repeated.

Fig. 2
figure 2

The planar images in the anterior (a) and posterior (b) projections simultaneously acquired with a 57Co sheet source suggest loosening of the femoral prosthesis. However, the SPECT/CT (c) indicates that the injected tracer is located anterior to the joint, within the psoas bursa

Full size image

The same limitations of RNA planar imaging for the assessment of acetabular cups of hip prostheses apply for both components of the knee prostheses. To the best of our knowledge, there is no other published series on SPECT/CT RNA for knee prostheses.

The published literature on RNA planar imaging of knee prosthesis is sparse. Marks et al. [14] showed a sensitivity and specificity of 31 and 70% for the femoral component, and 8 and 82% for the tibial component, respectively. Navasivayam et al. [6] suggested that RNA may have a complementary role. In our series, SPECT/CT specificity was equivalent to that in previous publications whilst sensitivity was superior. For the femoral component, while our planar specificity was higher, SPECT/CT had a superior sensitivity and the difference was statistically significant. Our results for SPECT/CT were superior to planar for the tibial component, but the difference did not reach statistical significance.

A more recent publication by Kitchener et al. [15]. showed superior sensitivities and specificities of 93 and 86%, respectively, for TKRs and 82 and 92% for unicompartmental prostheses. However, they did not individualise the results for the femoral and tibial components. They were also not able to detect femoral component loosening unless a long femoral stem was present. Their incidence of femoral component loosening of 8% (2 of 26) was also lower than the 47% (9 of 19) in our series. While our accuracy figures were lower than those of Kitchener et al. our method was not predicated on the presence of a long prosthetic stem as SPECT/CT allowed the detection of a potential gap at the bone/prosthesis interface at the femoral (Fig. 4a, b) and tibial (Fig. 3) condyles.

Fig. 3
figure 3

Radiotracer is demonstrated at the bone/metal interface of the tibial component of the left knee replacement, anterior to the tibial stem

Full size image
Fig. 4
figure 4figure 4

Radiotracer is demonstrated in the bone/metal interface of the medial (a) and lateral (b) condyle of the right total knee replacement

Full size image

Similar to the results for the assessment of hip prostheses, the strength of knee SPECT/CT RNA was in its negative predictive value.

There were only ten unicompartmental knee prostheses. Due to the small number, a separate analysis was not presented.

Due to the limited number of subjects, a further breakdown of the analysis for cemented versus uncemented prostheses was not done. There is an impression of more false positives in uncemented prostheses except for the femoral stem of hip prostheses (i.e. for hips: 3 of the 5 false positive cups but 0 of the 2 false-positive femoral stems; for knees: 12 of 12 false-positive femoral and 4 of 5 false-positive tibial components). Such a contention, however, must be noted with caution as the numbers were too small to be meaningful. The higher number of false positives may potentially be explained by the fact that many uncemented implants only have partial bony ingrowth over their surface, leaving a fibrous tissue adjacent to the implant elsewhere. It is possible that the radiotracer can get into this fibrous portion, giving a false positive reading.

In five of our subjects (4 THRs and 1 TKR), the protocol joint aspirates culture were positive resulting in a change in the management of three patients. These cases highlight another benefit of RNA over other non-invasive imaging modalities in allowing microbiological assessment of joint fluid from patients with painful prostheses in whom infection is a differential diagnosis.

The diagnostic criterion used for loosening was the demonstration of radiotracer within the bone/prosthesis interface. This may be flawed in some circumstances. The pre loosening process of osteolysis may potentially create a recess for radiotracer accumulation. Also the metallic instrumentation may not perfectly fit into the native bone leaving a gap that could entrap the radiotracer. In our experience this was a cause of false-positive results in the posterior aspect of the femoral condyles of the knee prosthesis.

There were multiple limitations in this study design. First, it was retrospective. These were clinical studies, the surgeons were not blinded to the RNA SPECT/CT results, and thus there was the potential for bias. The chosen gold standard for loosening was operative assessment, which reduces but does not eliminate bias.

The second limitation was the small number of patients who went on for revision surgery.

The third limitation was that all studies were interpreted by one reader only. Inter- and intra-observer reproducibility of this method were not evaluated.

For a more definitive assessment of the accuracy of this technique, future studies with a prospective blinded design and with more than one reader are needed.

Conclusion

The addition of SPECT/CT using a hybrid gamma camera/CT scanner has the potential to improve the diagnostic value of planar radionuclide arthrography in the assessment of the acetabular component of the hip prosthesis and femoral component of the knee prosthesis for mechanical loosening. In this retrospective audit the main strength of SPECT/CT in the assessment of all components appeared to be its negative predictive value.