Joint prosthesis replacement is constantly growing due to the increase of life expectancy. Surgery aims at obtaining pain relief and improving joint functionality and quality of life (QoL), thus resulting in a significant social cost saving [1
]. However, uncommon complications may occur such as mechanical loosening or prosthetic joint infection (PJI). The incidence of PJI ranges between 2.0% and 2.5% for primary surgical replacement [2
] and up to 20% for revision procedures [3
]. The improvement of surgical techniques, the use of more biocompatible prosthetic materials and the optimization of therapeutic management in the post-operative period have reduced the incidence of clinically evident infections that may be easily detected with conventional radiology and biochemistry. The infections are often indolent, sub-acute or chronic thus delaying and complicating the diagnosis and further management. As consequence, clinicians, Nuclear Medicine (NM) physicians and radiologists deal with more challenging and complex cases.
Recently, the European Association of Nuclear Medicine (EANM), European Bone and Joint Infection Society (EBJIS), European Society of Radiology (ESR) and European Society of Clinical Microbiology and Infectious Diseases (ESCMID) published a joint guideline on the diagnosis of PJI aiming at providing diagnostic flow-charts for its management [4
]. Since each diagnostic approach has its pros and cons, as always, the correct diagnosis relies on the combination of medical examination, clinical history, laboratory tests, microbiology and histology and imaging modalities. The identification of the pathogen at microbiology still remains the corner stone for the differential diagnosis between aseptic loosening and PJI; however, bacterial cultures and leukocyte count from joint aspiration or, in alternative, biopsies are invasive procedures that may expose the patient to the development of an infection, and they may present false negative results during an antibiotic treatment [5
]. Radiology is useful to detect potential abnormalities involving the implant and/or surrounding bone and soft tissues, but its modalities are usually not able to discriminate between an infection and an aseptic loosening, since typical radiological signs of infections may appear in late stages of the process [4
]. Moreover, metal artefact may complicate the interpretation of Computed Tomography (CT) and Magnetic Resonance Imaging (MRI).
On the other hand, NM, with its functional imaging, offers several techniques and radiopharmaceuticals able to confirm or rule out the infection, even in the earlier stages, with high diagnostic accuracy [10
]. In this view, radiolabelled white blood cells (WBC) scintigraphy, with both [99m
Tc] and [111
In], is able to accurately differentiate a PJI from a mechanical aseptic loosening [12
]. Many efforts have been made by EANM in order to standardize labelling procedures [13
], acquisition protocols and interpretation criteria [15
], since the high accuracy of this modality relies on these crucial aspects. Once correctly acquired, with times corrected for isotope decay, the images have to be properly displayed by using a “total count” intensity scale and the same intensity threshold [15
] in order to reduce the observer bias. The interpretation is usually based on visual analysis of the uptake over time in the affected joint: an increased uptake between delayed and late images, in terms of intensity and/or extension, is indicative of an infection. Conversely, a decreased/stable uptake over time rules out the infection with very high negative predictive value (more than 90%–100% depending on the studies) [15
]. However, several factors, e.g., only a small change of uptake intensity or extent over time, or the presence of bone marrow (BM) interference that does not allow a correct evaluation of the affected bone, may complicate the visual analysis, thus making the diagnosis more challenging. When qualitative assessment is not sufficient, a semi-quantitative analysis has been proposed [18
], comparing target to background (T/B) ratio between delayed and late images and using the contralateral tissue as reference [17
], but a precise threshold of increase in T/B ratio over times has never been defined. It is well known that false positive results may occur in presence of physiologic BM expansion as a consequence of migration of WBC in a reticulo-endothelial system; therefore, a combination of bone marrow scintigraphy (BMS) and WBC scintigraphy is strongly suggested in order to improve the accuracy [19
]. However, this approach requires additional radiations to the patients.
Therefore, the purpose of this study is to explore whether semi-quantitative analysis could be able to solve doubtful cases thereby reducing the number of BMSs and reserving this combined approach only to selected cases.
Robust evidence in the literature exists on the role of radiolabelled WBC scintigraphy in detecting PJI [23
]. The recently published EANM guidelines have definitely clarified all the aspects able to reach a high accuracy in detecting an infection by using this imaging modality [15
]. Indeed, if correctly acquired, displayed and interpreted, planar images allow to accurately differentiation between an infection from a sterile inflammation, as it has also been demonstrated in other clinical scenarios [24
]. However, despite these precautions, some scintigraphies still remain doubtful basing only on visual analysis.
In some cases, qualitative analysis of planar images may be equivocal for the presence of a slight increase of uptake and/or extent at late images. In these cases, different NM physicians could classify the same scintigraphy as positive or negative, depending on their subjective visual interpretation. Another possible source of doubt is the presence of BM expansion that is frequently observed after a prosthetic joint replacement and it is the result of a physiological migration of granulocytes in the reticulo-endothelial system. Over time, the activity of bone marrow may remain stable, increase or decrease similarly to the suspected area, thus interfering with the qualitative assessment. In these cases, an additional scintigraphy with colloids is strongly recommended in order to obtain a scintigraphic map of BM activity and to rule out or confirm the infection [19
]. When the uptake of radio-colloids is spatially congruent with the activity observed at WBC scintigraphy, the hypothesis of an infection may be rejected, although FN cases may be observed in patients with small infection and concomitant physiological BM expansion (Figure 3
). Conversely, in presence of a mismatch, the uptake shown at WBC should be considered consistent with an infection. The role of this combined approach is well established nowadays and allows reducing the number of FP cases due to BM activation thus maximizing the accuracy of WBC scintigraphy. The reported accuracy of this dual approach, in general population, ranges from 83% to 98% for both hip and knee prosthesis [19
]; therefore, it is widely adopted in many centres despite the administration of an additional amount of radioactivity and an additional scan-time, which could result in being inconvenient for the patients.
Such a high diagnostic accuracy can also be achieved by using a correct image acquisition and interpretation procedure on planar gamma-camera images, and better if correlated by SPECT/CT [15
] without the use of a combined BM scan.
However, the final diagnostic accuracy of WBC scan in a reported study can be highly influenced by the number of doubtful cases in the studied population. This is very rarely reported in scientific papers and may largely affects the final accuracy.
Aiming at achieving an accurate diagnosis and sparing additional radioactive doses to the patients, a semi-quantitative analysis of planar gamma-camera WBC images could be helpful in equivocal cases, but its role has not been systematically addressed and a precise cut-off of increase in T/B ratios over time still does not exist. Pelosi and colleagues in 2004, performed semi-quantitative evaluation in patients with suspected PJI using the left iliac crest as reference [18
], thus requiring the inclusion of this region in the field of view in all scintigraphies. Of course, the choice of reference tissue, the size/shape and the placement of ROIs is operator-dependent and may strongly influence the results. In a large single-centre retrospective study on 297 patients with suspected muskoloskeletal infection (amongst which, 138 patients with PJI), authors analysed four different reference tissues: contralateral tissue, anterior superior iliac crest, ipsilateral bone marrow and contralateral bone marrow [17
]. In this large series, the choice of contralateral tissue as reference provided the best results. However, they did not define precise thresholds for the increase in uptake over time. In 2014, Erba et al. retrospectively reviewed 235 WBC scintigraphies for suspected muskoloskeletal infection acquired with a fixed-time acquisition protocol and with a time corrected for isotope decay protocol [16
]. They performed both qualitative and semiquantitative analysis by using different thresholds of increase in T/B ratios over time (5%; 10%, 20% and 25%) concluding that the best accuracy, sensitivity and NPV were obtained considering any percentage of increase of uptake over time.
In our study we applied automatic ROIs by fixing a threshold of 40% of maximum pixel on late images thus avoiding the operator bias and we found that, by progressively increasing the thresholds, the sensitivity decreases and specificity increases. As always, the best performance of a diagnostic test derives from a balance between sensitivity and specificity, but it also depends on the exact clinical indication. For example, for the assessment of antibiotic therapy efficacy after the diagnosis of an infection, we mainly need a sensitive test able to detect if the infection is still present or not. Therefore, for this particular purpose, we can speculate that the threshold of 5% could be used in order to select the highest number of TP patients that still require additional treatments. But it needs to be confirmed by larger studies.
On the other hand, in a diagnostic setting we need a good sensitivity, specificity and accuracy. In our population, the cut-off at 20% of increase of radioactivity over time showed the highest accuracy (80.3%) and a good compromise between sensitivity (66.7%) and specificity (85.4%). BMS showed significantly higher specificity and accuracy than cut-off at 5%, 10% and 15% and significantly higher sensitivity than cut-off at 30%, thus confirming the added value of this combined approach in doubtful cases but we did not find any significant difference comparing BMS and the cut-off at 20% of increase of radioactivity over time.
The best sensitivity and specificity achieved with a cut-off at 20% of increase of radioactivity over time (66.7% and 85.4%, respectively) are much lower than those generally reported for WBC scintigraphy, but we should underline that we analysed only doubtful cases at qualitative analysis.
This study has several limitations. First of all, not all the patients underwent surgery, bone biopsy or joint aspiration, since some clinicians preferred to perform a “wait and see” approach depending on the specific case. Therefore, in our study 13 patients were followed up for at least 18 months. With the aim to assess whether the inclusion of these patients could affect the sensitivity, specificity and accuracy of the different methods, we also performed a statistical analysis of data without including these 13 patients and we obtained the same results and reached same final conclusions.
Moreover, given the retrospective nature of this study, we could not be able to retrieve causative pathogen in all included patients and ten patients did not perform an additional BMS. However, our primary aim was to compare different thresholds of semi-quantitative analysis according to final diagnosis regardless to the aetiology. The secondary endpoint was to compare this method with BMS, that is commonly, but not necessarily in all cases, used in clinical practice.
Moreover, we did not analyse the added value of hybrid single photon emission computerized tomography/computerized tomography (SPECT/CT). Indeed, it is well known that the role of hybrid imaging is undeniable, and SPECT/CT can improve the localization and extent of infection [4
]. However, the correct diagnosis of presence/absence of infection can be achieved by considering the increase of activity over time at planar images and indeed, in our Centre, SPECT/CT acquisitions are usually performed at 20 h (for 30–40 sec/step) only in patients with positive planar images, not for diagnostic purpose, but in order to achieve a better anatomic overview of the infective process and to provide an accurate information on its extent and localization into the bone or soft tissues. However, we cannot exclude that a SPECT/CT might be helpful in clarifying the presence or absence of infection in doubtful planar images in our population.
Finally, the choice of ROIs and their placement is operator-dependant; therefore, we used a software able to automatically draw irregular ROIs on the target and to mirror them on the contralateral background aiming to make the methodology more reproducible. However, this approach was not possible in patients with doubtful WBC according to criterion B for the interference of BM activity on the evaluation of prosthesis. In these eleven cases we manually drawn the ROIs around the profile of whole prosthesis that was clearly visible on planar images but, of course, we cannot exclude minimal differences among different operators, thus introducing a possible bias.
Although no definitive conclusions may be drawn because of the small population, the cut-off at 20% seems to be a reliable tool to identify patients with an infection thus avoiding performing additional BMS. On the basis of these preliminary results, we propose a step-by step approach that is summarized in Figure 4
. In case of doubtful qualitative assessment, a semi-quantitative analysis should always be performed before programming a BMS. If an increase of less than 20% in T/B ratios between delayed and late images is observed, the patient should perform a BMS in order to confirm or rule out the infection. Conversely, if we observe an increase of more than 20% between delayed and late images, we may diagnose an infection with high accuracy, thus avoiding an additional exam to the patient. Of course, larger retrospective and prospective studies are mandatory to confirm our results. We also foresee to explore the role of semi-quantitative analysis in SPECT/CT images and when using 111
In-WBC instead of 99m