Performance of ¹⁸F-sodium fluoride positron emission tomography with computed tomography to assess inflammatory and structural sacroiliitis on magnetic resonance imaging and computed tomography, respectively, in axial spondyloarthritis

Spondyloarthritis (SpA) is a group of chronic inflammatory rheumatic diseases that mainly affects the axial skeleton as well as the peripheral joints and enthesis [1–3]. The modified New York criteria define ankylosing spondylitis (AS) as the association of clinical criteria with radiological sacroiliitis (at least grade 2 bilateral or grade 3 unilateral) [4]. Since 2009, diagnosis of axial SpA has been based on the Assessment of SpondyloArthritis international Society (ASAS) classification criteria and can be made without radiological sacroiliitis, if sacroiliitis is detected on MRI, corresponding to the presence of bone marrow edema on STIR sequences [5, 6]. This inflammation can be detected before structural changes, which allows for an earlier diagnosis of axial SpA. Furthermore, it has been shown that syndesmophytes seem to preferentially develop at vertebral corners where both fatty lesions and bone marrow edema are present [7–9]. In fact, according to different studies, 57.4 to 94% of syndesmophytes develop in vertebral units without active inflammation [10–12]. Since the pathophysiological mechanisms leading to the syndesmophyte formation are not clearly identified on MRI, it appears relevant to evaluate new imaging techniques dedicated to bone remodeling. The diagnostic potential of nuclear imaging with positron emission tomography (PET) has been investigated in several inflammatory diseases, such as polymyalgia rheumatica, vasculitis, rheumatoid arthritis, and SpA [13, 14]. The relevance of PET imaging depends on the radiotracer used, and in axial SpA, Bruijnen and colleagues showed that the activity is better reflected by bone formation (using the ¹⁸F-fluorid radiotracer, which shows osteoblastic activity) rather than inflammation (with ¹⁸F-FDG and ¹¹C-PK11195, showing inflammation by glucose metabolism and neutrophil recruitment) [15, 16]. In fact, recent studies on ¹⁸F-fluoride PET have shown the potential for this technique to be used in the diagnosis of sacroiliitis [17, 18]. In the SIJ, the ¹⁸F-fluoride uptake is moderately associated with inflammatory lesions but not with structural damage on MRI [19] in a small sample of AS patients. The association between radiotracer uptake and the structural lesions on CT scan has not yet been evaluated.

The aim of our study was to compare increased sacroiliac joint (SIJ) uptake on ¹⁸F-fluoride sodium (¹⁸F-NaF) PET combined with CT scan with structural damages on CT scan according to a qualitative and quantitative approach in patients with axial SpA. We also evaluated the correlation between increased uptake of the SIJ on ¹⁸F-NaF PET/CT and the presence of inflammatory lesions on MRI.

Seven patients were classified with structural sacroiliitis on radiography. Among the 23 patients with active SpA, 20 (87%) presented a BASDAI ≥ 4, and the three other patients were undergoing NSAID treatment failure or had a contraindication for NSAIDs. Moreover, 52.2% of the patients presented with biological inflammation. After the study, 18 patients (78.3%) benefitted from tumor necrosis factor (TNF) blocker therapy.

Comparison between ¹⁸F-NaF PET/CT and CT scan (Fig. 2)

On a binary approach, there was no significant correlation between a positive PET and structural sacroiliitis on CT scan (p = 1). Among the 20 positive PET cases, there were 9 positive and 11 negative CT scans; 3 PET scans were negative, 1 with a positive CT scan and 2 others with negative CT scans.

For quantitative assessments, there was also no significant correlation between the PET activity score and erosion scores (ICC[IC95] = 0.30[− 0.13; 0.63]; p = 0.2) and ankylosis (ICC[IC95] = 0.07[− 0.35; 0.47]; p = 0.7). Similarly, the SUVmax SIJ/sacrum ratio was not correlated with erosion scores (ICC[IC95] = 0.25[− 0.18; 0.60]; p = 0.2) or ankylosis (ICC[IC95] = − 0.02[− 0.43; 0.40]; p = 0.9).

Comparison between ¹⁸F-NaF PET/CT and MRI (Fig. 3)

There was no significant correlation between a positive PET and inflammatory sacroiliitis on MRI on a binary approach (p = 0.2). All three negative PET scans were also negative on MRI, but only ten of the positive PET scans were positive on MRI, whereas the other ten were negative.

For quantitative assessments, there was a significant correlation between the PET activity score and the inflammation score (ICC[IC95] = 0.61[0.26; 0.82]; p = 0.001) and between the SUVmax SIJ/sacrum ratio and the inflammation score (CC[IC95] = 0.56[0.19; 0.79]; p = 0.004).

This study evaluating the performance of ¹⁸F-NaF PET/CT for the diagnosis of sacroiliitis in a population of axial SpA showed that there were twice as many positive PET scans as there were MRI and CT scans. This finding suggests that ¹⁸F-NaF PET may be more sensitive than MRI or CT scans for the detection of inflammatory and/or structural sacroiliitis by detecting early or scar lesions not visible on CT scans or MRI.

To our knowledge, this study is the first to compare ¹⁸F-NaF PET and CT scan for the diagnosis of structural sacroiliitis. In our study, there was no correlation between these two imaging techniques. Buchbender and colleagues already demonstrated the absence of a correlation between the ¹⁸F-NaF PET uptake and the presence of erosions, ankylosis, or sclerosis but only on MRI [21]. These results reinforce the idea that PET imaging can detect early bone remodeling processes before erosions or ankylosis is present.

Furthermore, we found a significant correlation between both the PET activity score and the SUVmax SIJ/sacrum ratio and the inflammation score on MRI. The concordance between ¹⁸F-NaF PET and MRI has also already been evaluated in different studies. Buchbender and colleagues [21] used hybrid ¹⁸F-Fluoride PET/MRI to show that bone marrow edema (BME) detected on MRI is associated with osteoblastic activity, while the combination of BME and fat deposition showed the highest ¹⁸F-F uptake. Fischer and colleagues [19] found that the correlation between the uptake detected on ¹⁸F PET and the inflammation on MRI was moderate for the SIJ (kappa = 0.64) and poor for the spine (kappa = 0.25). To explain these results, they suggested that ¹⁸F PET/CT might be able to detect the anabolic repair process leading to new bone formation regardless of the definite pathophysiological pathway.

In this study, we found high uptake values of ¹⁸F-NaF on SIJ in patients with axial SpA, without correlation with CT scan lesions. These results imply that ¹⁸F-NaF PET might detect an early bone remodeling process before structural lesions are visible on CT scan, suggesting the presence of an osteoid tissue not yet mineralized. Surprisingly, when erosions and ankylosis are present, this osteoid tissue seems to be minimal or absent, as it is not detected by ¹⁸F-NaF PET. However, we found a significant correlation between ¹⁸F-NaF PET and inflammation on MRI, even though it is observed in only half of the patients. According to these results, two pathophysiological pathways may be involved to explain the ossification process: one initiated by prior inflammation, and a second one without presence of inflammation on MRI.

In a recent study, biopsy procedures were used to collect material from PET-positive lesions in the spine for immunohistochemistry. These PET-positive lesions corresponded to osteoid formation and osteoclasts along with cell infiltrates in areas with both bone and connective tissue, which were largely absent in PET-negative lesions [22]. This study also showed that ¹⁸F-NaF PET/CT may detect changes in bone formation in AS during treatment, as one third of the PET-positive lesions in the spine disappeared after 12 weeks of TNF blocker treatment in their population of 12 patients with AS. In our study, patients continued their NSAID treatment but none of them were under biologic treatment for at least 3 months. Jarrett et al. demonstrated that NSAIDs have small influence on inflammatory lesions on MRI [23], but it still has to be evaluated on ¹⁸F-NaF PET.

The main strength of our study lies in its strong methodology. In fact, we based our results on a lecture by three different readers for CT scans, MRI, and ¹⁸F-NaF PET/CT. In other studies evaluating the performance of PET, the analysis of the imaging techniques was made by only one or two readers. We found that the inter-reader correlation for the qualitative assessment in PET was mediocre, which may be explained by the subjective nature of this evaluation on a complex joint with multiple bone superpositions. This issue suggests that a validation by three readers, with an agreement of at least two readers, seems to be more appropriate.

Moreover, in previous studies, the PET analysis was based only on a qualitative analysis [16, 19, 23] and/or on the SUVmax [18, 21], but an activity score has never been used, which may permit a better evaluation of the SIJ. Indeed, the inter-reader correlation was higher with a quantitative assessment when compared with a qualitative one, which justifies its utilization for a more accurate evaluation of the SIJ by PET imaging. In our study, we also calculated the ratio between the SUVmax for each SIJ and the SUVmax in the center of the sacrum, which makes it less dependent on the examination conditions, including the injected dose of radiotracer, but also on the individual uptake characteristics. The use of SUVpeak as a measuring tool instead of SUVmax could be relevant because it is less sensitive to noise. However, other studies on ¹⁸F-NaF PET/CT were based only on SUVmax, which is why we chose this assessment, since it permitted us to compare our results to those of the literature.

In our study, 7 AS and 16 non-radiographic SpA patients were included, which is different than what has been reported in the literature. In fact, in most of the studies, the evaluation of PET imaging is conducted on a population of AS [16, 18, 19, 21], but we found that PET imaging can be positive for non-radiographic sacroiliitis. In a study by Toussirot et al., no uptake of ¹⁸F-NaF PET was noted in SpA without structural sacroiliitis on radiography and also without inflammatory sacroiliitis on MRI [24]. Subgroup analysis between radiographic and non-radiographic SpA should be evaluated on a larger population in order to confirm these results.

We also found a significant correlation between the PET activity score and clinical activity scores (BASDAI and BASFI) but not with inflammation and structural changes, which is consistent with the results in the literature [23, 25, 26]. However, this outcome has to be taken with caution because it was not confirmed by the SUVmax and SUVmax SIJ/sacrum ratio. Moreover, spine assessment was not performed, limiting any comparison with clinical and biological parameters.

The performance of ¹⁸F-NaF PET/CT for the diagnosis of sacroiliitis has already been evaluated by Strobel and colleagues [18] in a population of 15 patients with active AS in comparison with 13 patients with non-traumatic mechanical low back pain for at least 3 months. In that study, they found a sensitivity, specificity, and accuracy of 80%, 77%, and 79%, respectively, for the diagnosis of sacroliitis, with a mean SUV SIJ/sacrum ratio of > 1.3 as a cut-off value, in comparison with radiography. Furthermore, it appeared that the sensitivity of PET imaging was better for grade 3 sacroiliitis (94%), which suggests that increased uptake might be correlated with post-inflammatory repair associated with osteoproliferation. In that study, as in ours, structural damages relative to osteoarthritic lesions such as osteophytes or intraarticular gas were not taken into consideration.

The main limitation of our study is the small number of examined patients, which requires the statistical analyses to be interpreted with caution. The quantitative approach, which had never been done before, aimed to compensate for this limitation.

Another limitation is the absence of a control group to assess the sensibility and specificity of ¹⁸F-NaF PET/CT for the diagnosis of sacroiliitis. Indeed, we did not consider the presence of structural damages caused by osteoarthritis which may be detected by ¹⁸F-NaF and may be the source of false positivity. Likewise, the possibility of overuse of SIJ or anatomical abnormalities such as accessory sacroiliac joint or lumbosacral transitional vertebrae, which can cause inflammation on MRI, was not evaluated. The impact of these modifications on ¹⁸F-NaF PET uptake could be evaluated in another study with a control group.

Also, all examinations were performed within a month, which is a short term, but does not formally avoid the potential risk of changes between exams in a same patient, even if we know by Jarrett et al. [23] that inflammation on MRI is quite stable at 6 months in SpA patients under conventional treatment.

In order to complete these results, a similar study could be conducted on the spine, evaluating the ¹⁸F-NaF PET/CT uptake, the bone marrow edema on vertebral corners on MRI, and the presence of erosions, syndesmophytes, and vertebral ankylosis on CT scans or X-rays.

Abnormal uptake of the SIJ on ¹⁸F-NaF PET is more frequent than inflammatory sacroiliitis on MRI (43.5%) and structural sacroiliitis on CT scan (43.5%) in a population of SpA patients. The comparison among ¹⁸F-NaF PET/CT, CT scans, and MRI found only a significant correlation between PET-positive lesions and inflammation assessed quantitatively but not with structural damages assessed on binary and quantitative approaches. Further studies with a control group and a larger sample are needed to evaluate the sensitivity and specificity of PET imaging.