Background: The diagnostic value of scintigraphy in detecting sacroiliitis in patients with spondyloarthritis is not clear.
Objective: To assess the diagnostic value of scintigraphy in detecting sacroiliitis in ankylosing spondylitis (AS) and in patients with clinically probable sacroiliitis without x-ray changes.
Material and methods: A systematic literature research was performed in the Pubmed and Medline database up to August 2007. Articles in English and German on patients with established AS and clinically probable sacroiliitis without x-ray changes were selected. In addition, studies including patients with mechanical low back pain as a control group were searched. Pooled sensitivity, specificity and positive and negative likelihood ratios were calculated.
Results: In total 99 articles about scintigraphy were found. 25 articles were included into the analysis. Overall sensitivity for scintigraphy to detect sacroiliitis was 51.8% for patients with established AS (n = 361) and 49.4% for patients with probable sacroiliitis (n = 255). Sensitivity of scintigraphy in patients with AS with inflammatory back pain (indicating ongoing inflammation) was 52.7% (n = 112) and in patients with AS and suspected sacroiliitis with magnetic resonance imaging showing acute sacroiliitis (as a gold standard) was 53.2% (n = 62). In controls with mechanical low back pain specificity was 78.3% (n = 60) resulting in likelihood ratios not higher than 2.5–3.0.
Conclusion: These data as a result of a literature research suggest that scintigraphy of the sacroiliac joints is at most of limited diagnostic value for the diagnosis of established AS, including the early diagnosis of probable/suspected sacroiliitis.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Ankylosing spondylitis (AS) is a common chronic inflammatory disease with an estimated prevalence of 0.2–1.2%.1–5 Disease onset is typically in the third decade of life, and the disease is known to have a substantial socio-economic impact on patients and society.6 Besides non-steroidal anti-inflammatory drugs (NSAIDs), which can be considered as the cornerstone of drug therapy,7 tumour necrosis factor α blocking agents have convincingly demonstrated a strong and prompt effect on almost all features of AS.8–12 In particular, patients with AS with a short disease duration and good functional status are more likely to respond to tumour necrosis factor α blocking agents than patients with long-standing disease and impaired function.13 Thus, an early and reliable diagnosis of AS is becoming more and more important. However, there is still an unacceptably long delay between the onset of symptoms and time of diagnosis with an average delay of about 8–11 years.14 15 One major reason for this delay is that for the diagnosis of AS according to the modified New York criteria unequivocal radiographic sacroiliitis of at least grade 2 bilaterally or grade 3 unilaterally is needed.16
Magnetic resonance imaging (MRI) has been established recently to be the best method to detect active sacroiliitis, especially early in the course of the disease.17–20 However, scintigraphy is still widely used for this purpose, because MRI is relatively expensive and not everywhere available.
Early reports in the 1970s about the diagnostic role of scintigraphy to detect sacroiliitis were promising;21–24 however, subsequent studies have suggested that clear separation of active AS from controls is difficult.25–32 Thus, data about the diagnostic value of scintigraphy to detect acute inflammatory changes in the sacroiliac joint (SIJ) for the diagnosis of non-radiographic axial spondyloarthritis (early forms of axial spondyloarthritis) and radiographic axial spondyloarthritis (AS) are conflicting.
In this study we assessed the diagnostic property of scintigraphy for acute sacroiliitis in early and established AS by performing a systematic literature research.
Systematic literature research
A systematic literature research was performed. Pubmed and Medline were searched from inception until August 2007 using the keywords “ankylosing spondylitis” OR “early spondyloarthritis” OR “spondyloarthropathy” OR “early ankylosing spondylitis” in combination with “scintigraphy” AND “bone scan” AND “sacroiliitis” AND “imaging”. Articles in English and German were considered. In addition, a manual search of the reference lists of the articles found was performed.
Study selection and quality assessment
We analysed patient populations with established AS, with early pre-radiographic AS (“suspected sacroiliitis”) and with mechanical low back pain (MLBP). Established ankylosing spondylitis was defined according to the modified New York criteria.16 Furthermore, we included also earlier articles where AS was defined according to similar criteria if at least sacroiliitis grade 2 bilaterally or grade 3 unilaterally was present.33 Suspicion for sacroiliitis in the absence of typical x-ray changes was based on different findings such as clinical suspicion for AS with or without laboratory signs for inflammation, inflammatory back pain according to the Calin criteria, and tenderness over the SIJ. Only scintigraphy studies were considered for analysis in which sensitivity and/or specificity could be calculated. For the definition of a positive scintigraphy we selected only studies in which a clear description of reference values were given and quantitative scintigraphy had been performed by using the SIJ:sacrum ratio (SIJ:S ratio). We included only other studies not fulfilling these criteria if other quality criteria were presented. As far as control subjects are concerned we only selected controls with mechanical low back pain (MLBP) excluding controls other than MLBP, such as healthy controls or controls with other rheumatic diseases.
Data extraction and data analysis
Data on publication status, trial design and patient characteristics such as age, gender, HLA-B27 status, disease duration and radiotracer used were extracted by two experienced rheumatologists (IS, JC). Sensitivities and/or specificities as well as corresponding likelihood ratios were calculated. Scintigraphy findings were also compared with different parameters of disease activity such as elevated C-reactive protein (CRP) levels or positive MRI findings and between different sacroiliitis grades on x-ray. If possible, κ values were calculated.
Number of articles found in total, excluded and included
After research with an adequate combination of keywords in the databases and after a manual search of the literature we found a total of 99 articles (fig 1).
Altogether, 74 articles were excluded for different reasons such as double counting, insufficient description of grading, not well defined AS population, no possibility to calculate sensitivity (eg, only mean values given), results given on SIJs only and not on patients, results given on spine or peripheral joints instead of SIJs, case reports only, report focusing on technical details, only letter, comment or editorial.
Finally, 25 articles were included in our analysis. Of these 25 articles, 10 studies were about patients with AS only, eight studies about suspected sacroiliitis only, one study about MLBP controls only, one about AS and MLBP, two about suspected sacroiliitis and MLBP, two about AS and suspected sacroiliitis, and one about AS, suspected sacroiliitis and MLBP (fig 1).
Definition of positive scintigraphy findings
In only one study radioactive strontium-87m was used as a radiotracer.34 In all other 24 studies included in this analysis radioactive technetium-99m was used, mostly labelled to methylene-di-phosphonate. Quantitative scintigraphy was performed in 23 studies by calculating the ratio of radiotracer uptake over the SIJs compared with the sacrum as a reference point (SIJ:S ratio). Only two studies used other techniques by either doing a qualitative consensus reading35 or taking part of the ilium as a reference point instead of the sacrum (here the reference value was given as 280 (SD 90).36 In all other 23 studies that were included into the analysis reference values of the SIJ:S ratio were given for the definition of positive versus negative scintigraphy findings with a mean value for the cut-off of 1.39 (SD 0.18). In 12 these 23 studies the cut-off was even more precisely defined by calculating the mean radiotracer uptake in controls plus twice the standard deviation,24 28 34 36–44 here the mean tracer uptake was 1.48 (SD 0.18). Two studies even used different age groups to define reference values.42 45
Sensitivity of scintigraphy in established ankylosing spondylitis
The AS group included a total of 361 patients from 14 studies (table 1). Demographic data about disease duration, age, gender distribution and HLA-B27 are provided in table 2. A total of 187 (of 361) patients with established AS had a positive scintigraphic finding resulting in a sensitivity of 51.80% (fig 2A,B).
Sensitivity could be further calculated for those patients with AS with x-ray grade 2 or 3 findings (excluding grade 4) for 244 patients with AS from 11 studies23 28 35–37 39–41 46–48), 160 of these had positive findings by scintigraphy resulting in a sensitivity of 65.57%.
Sensitivity could also be calculated for 71 patients with AS with x-ray grade 4 findings (complete ankylosis of the SIJs) from 10 studies23 24 28 36 37 39–41 47 48). Only five of these 71 patients with AS had positive findings by scintigraphy, which translates into a sensitivity of 7.04%.
The most informative study is the one by Goei The et al.37 These investigators assessed 137 patients with inflammatory back pain and 31 controls. A total of 112 of these patients were diagnosed with AS according to the modified New York criteria. The cut-off for positive scintigraphy was well defined by calculating the mean uptake in controls plus twice the standard deviation. For all patients with AS with inflammatory back pain, a sensitivity of 52.7% was calculated, which is in concordance with the overall pooled sensitivity of about 52%. If only patients with AS with grade 2/3 findings were taken sensitivity in this study increased to 56.8%, and sensitivity for patients with AS with total ankylosis was 0%.
Sensitivity of scintigraphy in ankylosing spondylitis compared with magnetic resonance imaging, elevated C-reactive protein/erythrocyte sedimentation rate and clinical parameters
In six AS scintigraphy studies it was possible to extract data on the agreement between scintigraphic findings and other markers of disease activity such as MRI, laboratory and clinical disease activity parameters.
Comparison with magnetic resonance imaging
In two studies including a total of 37 patients with AS35 45 the pooled sensitivity of scintigraphy to detect acute sacroiliitis was 59.5% (22 of 37) compared with MRI as a gold standard (fig 3). Jee et al35 found that of 11 patients with AS with a positive MRI six patients with AS also had a positive scintigraphy, here a κ value (for the comparison of scintigraphy and MRI) of 0.33 could be calculated. In a recent study Inanc et al assessed 54 patients with spondyloarthropathy (SpA), 26 of them had AS and a positive MRI, and 16 also had a positive scintigraphy. In this study a κ value of 0.07 was found for the 54 patients with SpA.45
Comparison with laboratory disease activity parameters
In only one study sensitivity of scintigraphy in comparison with elevated laboratory parameters as a reference standard could be calculated. The 29 patients with AS in this study were divided into low and high disease activity subgroups according to laboratory parameters (ESR and CRP).54 Of 11 patients who had elevated ESR/CRP all 11 patients had a positive scintigraphy finding resulting in a sensitivity of 100%.
Comparison with clinical disease activity parameters
From three studies including 34 patients with AS it was possible to calculate a pooled sensitivity of 55.9% (19 of 34) for scintigraphy for the detection of sacroiliitis compared with clinical disease activity parameters that were defined as presence or absence of subjective symptoms (not otherwise specified),34 back pain55 and tenderness on stressing the SIJs.56 In one study by Ranawat et al55 a κ value of 0.27 could be calculated: in this study 11 of 17 patients with AS had back pain and 10 of 11 (90.9%) also had a positive scintigraphy finding; six of 17 patients with AS denied having back pain and two of six (33.3%) had a negative scintigraphy finding.
Five further studies also assessed the correlation between scintigraphy and markers of disease activity in AS; however, in these studies sensitivity or κ values could not be calculated. Three of these five studies showed negative results with no correlation between scintigraphy and clinical symptoms or laboratory parameters in a total of 84 patients with AS.25 57 58 Two studies showed statistical significant differences between the mean radiotracer uptake between clinically active versus inactive patients with AS (n = 16)30 and between active patients with AS according to ESR elevation (n = 7) and the control group (nine controls) in that study.59
Sensitivity of scintigraphy in patients with suspected sacroiliitis
In total 255 patients with suspected sacroiliitis were found in 13 studies (table 1). Demographic data about disease duration, age, gender distribution and HLA-B27 are provided in table 2. For the 255 patients with suspected pre-radiographic sacroiliitis a sensitivity of 49.41% was found.
Comparison of scintigraphy with magnetic resonance imaging in patients with suspected sacroiliitis
For patients with suspected sacroiliitis, two studies including 25 patients compared MRI with scintigraphy. A pooled sensitivity of 44.0% was calculated (11 of 25) (fig 3). Battafarano et al49 assessed eight patients with inflammatory back pain (and normal radiographs). MRI showed acute sacroiliitis in 100% of the patients (eight of eight). Scintigraphy was available in seven these eight patients and showed positive results in 71.0% (five of seven). Another study by Blum et al50 assessed 17 patients with inflammatory back pain with or without laboratory signs of inflammation and with normal radiographic findings. MRI showed a very high sensitivity of 94.1% (16 of 17) while scintigraphy only had a sensitivity of 35.3% (six of 17).
The sensitivity of scintigraphy to assess sacroiliitis compared with MRI in the combined group of patient with AS and suspected sacroiliitis (n = 62) was 53.2% (fig 3).
Specificity in patients with mechanical low back pain (control group)
In total, 60 patients with non-inflammatory (MBLP) back pain were found in five studies37 38 42 43 49 (table 1) including patients with “disc degeneration” (n = 12),37 “mechanical low back pain” (n = 11)49 and “non-inflammatory low back pain” (n = 27).42 43 Demographic data about age, gender distribution and HLA-B27 status are provided in table 2. Here, 47 of 60 controls had true negative findings in scintigraphy resulting in a specificity of 78.33%.
Likelihood ratio for established ankylosing spondylitis and suspected sacroiliitis
Combining sensitivity and specificity resulted in the following positive likelihood ratios (see fig 2B): The positive likelihood ratio (+LR) for the total group of patients with AS (n = 361), irrespective of the grading by x-ray, was 2.39. Subgroup analysis showed a +LR of 3.03 for patients with AS grade 2/3 and a +LR of 0.33 for patients with AS with grade 4 in x-ray. The +LR for the group with suspected sacroiliitis was 2.28. Negative likelihood ratio (–LR) analysis showed the following values: –LR was 0.62 for the whole group of patients with AS, 0.44 for AS grade 2/3, 1.19 for grade 4 and 0.65 for the group with suspected sacroiliitis (fig 2B).
The detection of acute sacroiliitis by whatever reliable, reproducible and affordable method can be considered to be the cornerstone for the early diagnosis of AS.60 Scintigraphy has been used for this purpose since the early 1970s,34 61 is nearly everywhere available and about 50% cheaper than an MRI investigation. However, especially as the MRI technique has successfully been introduced and applied in patients with SpA the question about the role of scintigraphy in a diagnostic approach of SpA has frequently been raised.
We and others have previously discussed the diagnostic value of scintigraphy in general;20 62 however, for the first time we now assessed the diagnostic property of scintigraphy in detecting sacroiliitis by calculating sensitivity, specificity and likelihood ratio.
A critical question about an analysis like this is which gold standard should be used to estimate the sensitivity for scintigraphy. X-rays do not detect acute inflammation but rather chronic changes as a consequence of inflammation. Thus, chronic changes as detected by x-rays can occur together with but also without active inflammation. Therefore, the very low sensitivity for scintigraphy in patients with radiographic sacroiliitis grade 4 (complete ankylosis) is not surprising because in most of these patients the presence of acute inflammation cannot longer be expected. Also clinical and laboratory parameters (such as CRP) alone have only a limited specificity for the diagnosis of AS/early AS.63 On this background, patients with radiographic sacroiliitis not more than grade 2 or 3 in combination with clinical signs of acute inflammation, such as the presence of inflammatory back pain, and patients with acute inflammation as shown by an MRI are, in our opinion, the most suitable groups for this purpose.
The overall sensitivity for scintigraphy in patients with AS was 52%. One study by Goei The et al37 investigated patients with AS who complained also about inflammatory back pain. The cut-off for positive scintigraphy was well defined by calculating the mean uptake in controls plus twice the standard deviation. For all patients with AS, a sensitivity of 52.7% was calculated, which is in concordance with the overall pooled sensitivity of about 52%. If only patients with AS with grade 2/3 findings were taken sensitivity in this study increased to 56.7%.
Our analysis of 37 patients with AS and 25 patients with suspected sacroiliitis in whom also an MRI was performed showed a pooled sensitivity of 59.5% and 44.0% (pooled 53.2%) for a positive scintigraphy for the detection of acute sacroiliitis, if acute lesions as shown by MRI were used as the gold standard. Again, this is close to the overall calculated sensitivity of 49% for a positive MRI when various clinical, laboratory and imaging activity parameters were used as a gold standard.
Taken together, a sensitivity of a positive scintigraphy for the detection of acute sacroiliitis is not higher than 50–55% and the specificity can be expected to be not higher than about 80%, resulting in a likelihood ratio of between 2.5 and 3.0 if scintigraphy is used for diagnostic purposes, which implies only a small increase in the probability of a diagnostic approach.
According to our recently proposed diagnostic algorithm,64 65 this would mean for example an increase from a pretest probability of about 14% (if a patient with chronic low back pain already fulfils the clinical criteria for inflammatory back pain) to about 35% in positive scintigraphy (and a decrease to about 6.5% in negative scintigraphy) or from about 59% (presence of inflammatory back plus positive HLA-B27) to about 83% post-test probability for the diagnosis of axial SpA in positive scintigraphy (and a decrease to about 38% in negative scintigraphy) if a likelihood ratio of about 2.8 is assumed.
In clinical practice different methods are used to assess sacroiliitis by scintigraphy.66 A quantitative assessment of tracer uptake in the SIJs is mostly performed in those patients who present with an abnormal qualitative finding suggestive of sacroiliitis. Other influencing factors that represent a bias are patient-related factors such as movement of the patient during the examination, metal implants,67 gender,68 age69 and technical factors such as time of assessment,41 distance between collimator and patient, and contamination of patients’ clothes with radioactive substances due to false application of radiotracer.67
Furthermore, the radiation exposure of scintigraphy has also to be taken into account. If a standard dosage of about 600 MBq of a technetium-99m-labelled radiotracer is used scintigraphy has a radiation exposure of about 4.8 mSv (milli Sievert),67 which is lower than the radiation of a computed tomography assessment of the SIJs (about 17 mSv)70 but higher than conventional x-rays of the SIJs (about 0.7 mSv).71
In conclusion, scintigraphy does not play a part in the diagnosis of acute sacroiliitis if an MRI can be performed and only a very limited role if an MRI is not available. An MRI has the advantage of no radiation exposure and of a likelihood ratio of at least 1063 when used for diagnosis, it allows an exact anatomical description of the inflamed structures, and, finally, not only acute but also chronic changes can be assessed by MRI.72 73
Competing interests: None.