Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
Contents lists available at ScienceDirect
Best Practice & Research Clinical Rheumatology journal homepage: www.elsevierhealth.com/berh
4
Imaging in early rheumatoid arthritis Fiona M. McQueen* Department of Molecular Medicine and Pathology, Faculty of Medical and Health Sciences, University of Auckland, 85 Park Rd, Grafton, Auckland, New Zealand
a b s t r a c t Keywords: Rheumatoid arthritis MRI Ultrasound CT scanning Imaging
Imaging in early rheumatoid arthritis (RA) has undergone extraordinary change in recent years and new techniques are now available to help the clinician diagnose and manage patients much more effectively than previously. While established modalities such as plain radiography (X-Ray) remain important, especially for detection of erosions and determining the progression of joint damage, there are many instances where ultrasound (US), magnetic resonance imaging (MRI) and computed tomography (CT) scanning provide added information. MRI and US are now used regularly by clinicians to help diagnose RA in the pre-radiographic stage as they offer improved visualisation of joint erosions. They also have the potential to provide prognostic information as MRI bone oedema/osteitis is linked to the later development of erosions and power Doppler ultrasound (PDUS) joint positivity is also a predictor of joint damage. Nuclear imaging techniques such as single photon emission computed tomography (SPECT) and positron emission tomography (PET) are also highly sensitive for detecting joint change in early RA and pre-RA but not yet used clinically mainly because of accessibility and radiation exposure. MRI, US, scintigraphy, SPECT and PET have all been shown to detect sub-clinical joint inflammation in patients in clinical remission, a state that is now the goal of most treat-to-target management strategies. Thus, imaging may be used to direct therapeutic decision making and MRI is also now being used in clinical trials to determine the impact of disease-suppressing therapy on the course of synovitis and osteitis. As is the case for all tests, it would be unwise to rely completely on any one imaging result, as false positives and negatives can occur for all modalities. Thus, the
* Tel.: þ64 9 3737999x86374; fax: þ64 9 3737677. E-mail address: [email protected]. 1521-6942/$ – see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.berh.2013.09.005
500
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
clinician needs to choose the most relevant and reliable imaging test, while also striving to minimise patient discomfort, radiation burden and economic impact. Ó 2013 Elsevier Ltd. All rights reserved.
Introduction Major advances have been made in the diagnosis and management of rheumatoid arthritis (RA) in the last two decades. Clinicians are now aware that patients do best if their disease is diagnosed early and effective disease-suppressing therapy is begun before permanent joint damage occurs. With more aggressive and effective treatment regimens has come a need to diagnose RA as early as possible, optimally at the time of first presentation, and to monitor accurately the effectiveness of management strategies so that these can be varied according to the ‘treat to target’ philosophy, aiming to achieve clinical remission. Imaging now plays a major role in these activities. Although plain radiography continues to be a mainstay for detection of early bone erosion and cartilage damage, it does not go far enough for the current requirements and so there has been a proliferation of newer, more sensitive modalities including ultrasound (US), magnetic resonance imaging (MRI), computed tomography (CT) and positron emission tomography (PET), to meet the needs of the modern rheumatologist. However, it is important to understand that each of these modalities has strengths and weaknesses and the unwary clinician could definitely face pitfalls. Reader reliability can be variable, which in the context of early disease can make the difference between a positive and a negative diagnosis of RA; joint coverage differs substantially between the different modalities; and there are also issues surrounding exposure to ionising radiation, feasibility and cost that need to be borne in mind when selecting the appropriate modality. This review aims to summarise the latest imaging advances in the field of early RA, including the role played by each modality, firstly in diagnosis and secondly in monitoring disease activity and damage. Insights provided by imaging into disease pathogenesis will also be discussed as will the emerging role of imaging in clinical trials. The clinical utility of each imaging modality will be reviewed including a discussion of cost and accessibility, which may limit the use of some of these techniques to centres where special facilities and expertise are available. Plain radiography Most clinicians have grown up with the concept that plain radiography is integral to the practice of rheumatology and indeed this continues to be the case. X-rays (XR, plain radiography) of the hands (including wrists) and feet is the imaging investigation that virtually all patients require at first presentation to determine baseline joint integrity. Repeated imaging is usually performed over the course of their disease to monitor damage progression. There is a far greater base of experience amongst rheumatologists and radiologists and a much larger literature devoted to plain radiography than to any other imaging modality in RA. Other advantages include ease of access, wide coverage of important joint regions, newer digitised formats that allow easy retrieval and comparison of images longitudinally and relative low cost. Disadvantages include exposure to ionising radiation, which although relatively low for one set of XRs can cumulate over time with a potential impact of patient longevity [1] and, most importantly, a lack of sensitivity for detecting early joint damage and inability to image the inflammatory processes within the joint that precede damage [2]. XRs in the diagnosis of RA Plain radiography is not particularly helpful in making a diagnosis of RA in the majority of patients. XRs of hands and feet are abnormal at first presentation in only 15–30% of patients who eventually fulfil the diagnostic criteria for RA [3,4]. This figure differs depending on symptom duration prior to presentation and is tending to be lower with more modern cohorts presenting earlier [4]. When
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
501
abnormalities are present in early RA, these most often take the form of nonspecific soft-tissue swelling and periarticular osteopaenia, neither of which are sufficient to confirm a clinical diagnosis. The critical radiographic sign of typical marginal erosions is most often first observed in the feet, particularly at the fifth metatarsophalangeal (MTP) joint, probably because radiographs image this region particularly well [5]. The wrist is an important site of early erosion on MRI [4] but is poorly imaged by X-rays and often not informative, because of the complex anatomy and overlapping shadows produced by a twodimensional modality (Fig. 1). Having said this, the radiographic detection of bone erosions in a patient with early inflammatory arthritis is of major clinical importance as it greatly increases the likelihood of an RA diagnosis and has serious prognostic implications, immediately putting the patient into the category of having aggressive, damaging disease. Thus, XR is a test with low sensitivity but high specificity for making a diagnosis of RA and for this reason been incorporated into both sets of American College of Rheumatology (ACR) diagnostic criteria [6,7]. XRs to monitor damage progression in early RA As far as monitoring the progression of damage over time, XR remains the current gold-standard imaging modality. There are a number of good reasons for this. First, as already stated it is widely
Fig. 1. X-Ray and MRI of the hands/dominant wrist in a 68 year old woman with early RA (8 month history). A) X-Ray shows a small subchondral lucency at the radius (arrow), no definite rheumatoid erosions B) T1w coronal 3T-MRI right wrist shows erosions involving radius, scaphoid and lunate (circle) and pole of hamate (arrow). C) Matching coronal proton density T2w MRI scan shows florid bone oedema/osteitis within the lunate (arrowheads) and erosion at pole of hamate (arrow). Slightly different cut does not show other erosions.
502
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
available and offers good coverage of multiple joints likely to be affected by RA. Second, it is quite feasible and indeed standard practice to image these joints on multiple occasions to determine the progression of erosions and joint space narrowing (JSN) indicating cartilage damage [3]. Just how often XRs need to be repeated is up to the discretion of the clinician but a yearly frequency, especially during the first few years of disease when erosive damage progresses most rapidly, is common practice. [8] The use of the Sharp van der Heidje (SvdH) score for quantifying erosions and JSN reinforces the role of XR in the setting of randomised clinical trials (RCTs). This imaging outcome measure has been extensively validated and very high reliability is normally quoted in major clinical trials. In the ATTRACT (Anti-TNF Therapy in RA with Concomitant Therapy) trial investigating the efficacy of infliximab in preventing erosive progression, inter-reader reliability was high with intraclass correlation coefficients (ICCs) of 0.84–0.92 for status scores [9]. Guillemin et al. compared several radiographic scoring systems for quantifying rheumatoid joint damage and found that inter-rater reproducibility was highest with the SvdH score compared with earlier systems such as the Larsen method, but Bland–Altman graphs showed a decrease in reader concordance where there was more severe damage [10]. The smallest detectable difference for change was 3.5% of the maximum score. Recently, Knevel et al. investigated whether XR of one joint area would be sufficient for evaluating erosiveness, but found that if only the hands were imaged (and the feet omitted), 24–40% of RA patients were incorrectly classified as non-erosive [11]. Thus, despite its drawbacks, plain radiography performs well as a test to determine damage progression in RA and this is because of several important factors. XRs profile cortical bone very well producing a clear margin that may be interrupted if an erosion is present or become more closely apposed to a neighbouring cortex if the joint is narrowed. From an imaging point of view, this is advantageous as it means that even a relatively small change can be detected with quite good reliability (leaving aside the technical issues of joint imaging which are largely dealt with by adherence to standardised radiographic views). When these measurements are repeated many times over (a total of 44 joints are imaged for erosions and 40 joints for JSN in the SvdH score) in a disease where joint damage appears to progress at a similar rate at all sites, a useful clinical score is produced. Nevertheless, all of this comes to naught in the patient whose disease is as-yet subradiographic and it is in this setting that other imaging modalities assume serious importance. Tomosynthesis Tomosynthesis is a new technique developed from conventional tomography and involves collecting a number of projected images at different angles with a digital detector, allowing reconstruction (by scrolling though images) at arbitrary depths [12]. This improves detection of erosions especially at complex sites such as the wrist, because there is less projectional overlapping than on standard radiographs. Canella et al. compared tomosynthesis with plain radiography for the detection of wrist erosions in 40 RA patients, using multidetector CT as a gold standard [13]. Significantly more erosions were shown with tomosynthesis than with radiography (a total of 232 detected by CT vs 199 by tomosynthesis vs 140 by plain radiography, p < 0.0001) in this group of RA patients of whom 40% had early disease (scanned within 2 years of presentation). With CT as a reference, the sensitivity and specificity of tomosynthesis were 77.6% and 89.9%, respectively, and for radiography they were 53.9% and 92%, respectively [13]. Thus, tomosynthesis is a relatively low-cost option that has yet to be fully appreciated by the rheumatology community and which greatly increases the pickup of erosions in RA patients, when compared with plain radiography. Further studies into its potential use as a diagnostic aid are warranted. CT scanning There have been relatively few CT studies in the context of early RA but it is an important player in the imaging team and could be very helpful for assisting diagnosis at first presentation, for the same reasons as tomosynthesis. Like plain radiography, it profiles cortical bone extremely well and is generally regarded as the gold standard for imaging erosions against which other modalities are compared [13– 15]. Multidetector helical CT produces very high-quality images which can be stored in a digitised
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
503
format and compared with later images to determine erosion progression. Perry et al. compared CT with MRI at RA wrists and found that although they were closely comparable, CT outperformed MRI in erosion detection [15] (Fig. 2). Dohn et al. confirmed this at the metacarpophalangeal (MCP) joints [16] and then went on to use CT as a means to monitor erosion progression in patients on anti-tumour necrosis factor (anti-TNF) therapy [17]. CT scanning does entail exposure to ionising radiation but the biological impact of this is likely to be relatively low as only the extremities are examined. It does provide less coverage than plain radiography as usually only one joint area is scanned, and given the results of Knevel et al. cited above [11], this could be a significant drawback. It is worth noting that reliability for scoring CT scans in the tomosynthesis study was extremely high with inter-reader ICCs of 0.96–0.99 [13], making this a highly reproducible ‘end’ point which could be clinically useful in early RA. CT techniques to investigate bone loss Microfocal CT (micro-CT) is a high-resolution technique that allows volumetric assessment of bone mineral density and it has been used extensively by Schett et al. to investigate the bone changes associated with erosion in RA. One study from this group included RA patients (16 of whom had recent-onset disease) and 30 healthy controls [18] and showed that while small erosions were observed in controls as well as RA patients, lesions >1.9 mm in diameter were highly specific for RA. RA erosions were mostly found along the radial aspect of the metacarpal heads. This group has gone on to study this modality as a
Fig. 2. MRI and CT scan axial images from a 59 year old male with RA showing matched erosions. A) and B) axial T1w MRI images showing erosions at the 5th metacarpal base on 2 adjacent slices (arrows), C) and D) high resolution CT scans of the same region confirm the same erosions with greater clarity indicating the breach in the bony cortex.
504
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
means to assess the response to the anti-interleukin-6 (anti-IL-6) biological disease-modifying antirheumatic drug (bDMARD), tocilizumab, which induced limited erosion repair in larger sclerotic lesions [19]. Another CT-based imaging technique that has been applied to the investigation of periarticular osteopaenia in early RA is CT osteoabsorptiometry [20]. Mineralisation at the MCP joints was significantly reduced in all groups of RA patients studied, including those with early disease, compared with controls (p < 0.004). A French group has also studied volumetric bone mineral density (vBMD) using a peripheral quantitative CT (high-resolution-peripheral quantitative CT, HR-pQCT) system and confirmed the involvement of the trabecular bone compartment in periarticular osteopaenia [21]. Currently, these extensions of CT are research tools only and are unlikely to be used in a clinical context. MRI scanning MRI is ideally suited to visualising pathology in early RA. This is because it combines tomographic capability with the capacity to image bony structures and cartilage, as well as soft tissues and fluid. It produces images by detecting signal from Hþ ions as they are exposed to a powerful magnetic field and are forced to reorientate their spin direction following the application of electromagnetic pulses [22]. This means that the images are produced in a completely different way from radiographic modalities, which depend upon the attenuation (blocking) of X-rays as they pass through tissue, thus providing clear detail of Caþþ-containing structures such as bone but minimal information about soft tissues. The detection of Hþ ions by MRI means that tissues with high concentrations of water register a high signal on T2-weighted (T2w) and short tau inversion recovery (STIR) sequences, making this an ideal modality for detecting free fluid as well as regions of inflammation. In RA, this means the easy detection of synovitis, tenosynovitis and synovial effusions, and also the virtually unique capacity to image inflammation within bone in the form of bone marrow oedema/osteitis as discussed below. The option to scan after infusion of a paramagnetic gadolinium-containing contrast agent allows further definition of regions of active inflammation where there is enhanced vascularity. These post-contrast T1weighted (T1w) sequences are often produced with machine settings to suppress the signal produced by fat (and therefore referred to as fat saturated (FS)) so that the contrast-enhanced tissues are better displayed. MRI insights into RA pathology MRI bone-marrow oedema in RA is due to the presence of an inflammatory infiltrate within subchondral trabecular bone. This inflammation within the bone marrow has been termed osteitis and it replaces the normal tissue that has a high component of fat, leading to a dramatic change in MRI signal characteristics [23]. However, the bone oedema appearance on MRI scans is nonspecific and does not necessarily indicate a uniform underlying pathology. Any process that results in increased cellularity or vascularity within subcortical bone will cause MRI bone oedema and this is florid in osteomyelitis [24] and also frequently accompanies fracture [25]. Osteonecrosis or osteoarthritis (OA) is also associated with MRI bone oedema and these pathologies may co-exist with RA [26]. In OA, increased Hþ signal within bone is due to replacement of fat-containing marrow by fibrotic repair tissue as well as regions of haemorrhage and inflammation [27]. A considerable body of evidence now exists to indicate that MRI bone oedema has special significance in RA as it is not only common (occurring in 40–60% of cases) but is also an adverse prognostic sign being associated with an aggressive, erosive disease phenotype [28]. Imaging/histological studies have been performed where sites of bone oedema were identified on pre-surgical MRI scans in RA patients undergoing joint replacement. Subsequent histological examination of resected bone revealed an inflammatory infiltrate in the subarticular region [29,30]. This was characterised by immunostaining as lymphoplasmacytic, containing T and B lymphocytes, plasma cells and macrophages adjacent to osteoclasts. Intense staining for RANKL (receptor of activated nuclear factor-kappa B ligand) was noted [31]. There were many similarities to the rheumatoid synovitis lesion, including follicular clusters of B cells reminiscent of germinal centres [26]. This work supported the proposal that the osteitis lesion represents a second focus of pathology in RA [32] and could drive the erosive process [28,33]. Therefore, the finding of MRI bone oedema, especially at the wrist, which seems to be a particularly informative site, has importance in early RA.
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
505
MRI in the diagnosis of RA In 1998, McQueen et al. reported that 45% of a cohort of early RA patients, recruited within 6 months of the onset of symptoms, had evidence of bone erosion on MRI scanning of the dominant wrist while only 15% had radiographic erosions [4]. A follow-up study tracking the progress of these MRI erosions revealed that 21% and 26% were observed on XR, 1 and 2 years later, respectively [34]. Although initially it was suspected that this disparity might mean that MRI erosions were often not ‘true’ erosions, later data suggested that the more likely explanation was a deficiency in the XR detection of erosions at the wrist. Dohn et al. clarified this by comparing CT, MRI and XR of the second-to-fifth MCP joints of one hand for the pick-up of erosions in 17 RA patients and four healthy controls. These different modalities detected 77, 62 and 12 erosions, respectively, indicating that CT and MRI are 5–6 times more sensitive than XR for erosion detection [14]. The influence of time is also important in RA and Ostergaard et al. found that new erosions were detectable on wrist MRI, a median of 2 years earlier than they were apparent on XR [35]. This group also used 0.2-T extremity-MRI (E-MRI) to show that conventional radiography was unable to detect small MRI erosions involving 1 h, hand involvement and MRI bone oedema and was correct in 82% of the 27 patients who developed RA after 1 year. Similar findings were reported by Tamai et al. in their cohort with undifferentiated arthritis [42,43]. All the 22 patients who had MRI bone oedema at the wrist and were anti-citrullinated peptide/protein antibody (ACPA) positive progressed to RA at 1 year, with a positive predictive value (PPV) of 100%. Narvaez found that MRI synovitis with erosions and/or bone oedema predicted a later diagnosis of RA in a similar group of patients. In their study, MRI scan findings correctly predicted the onset of RA in 31/33 patients, equating with a sensitivity of 100% and a specificity of 78%, compared with ACPA positivity which alone gave a sensitivity of 23% and a specificity of 100% [44]. To look even earlier at the pre-RA group of anti-cyclic citrullinated peptide-positive (antiCCPþ) patients with arthralgia but no true synovitis, Krabben et al. recently reported a correlation between the MRI inflammation score (RAMRIS synovitis plus bone oedema scores) and the presence of ACPA, echoing the observations of Tamai et al. [42] and suggesting that these could be linked during disease initiation [45]. MRI to predict joint damage progression in early RA The association between MRI bone oedema and the subsequent development of radiographic erosions in RA was first reported by McQueen et al., in 2003 [38]. These patients had baseline MRI scans of the wrist at presentation as described above [4] and were then re-examined radiographically after 6 years. On univariate regression, the baseline MRI bone oedema score (but not the synovitis score) was predictive of XR erosion and JSN scores, separately, and combined as the total SvdH score. An optimal model incorporating MRI features and levels of inflammatory markers explained 59% of the variance in the 6-year total Sharp score. Site-specific analysis revealed that MRI bone erosion was more likely to be detected after 6 years if bone oedema was present at baseline, with an odds ratio (OR) of 6.5. Interestingly, baseline bone oedema was also found to be a predictor of physical function after 8 years [46]. Fig. 3 shows an example of bone oedema involving the scaphoid evolving to bone erosion over a 3-year period in a typical early RA patient.
506
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
Fig. 3. T1w MRI scans from a 58 year old man with early RA indicating the transition from bone oedema/osteitis to erosion A) Coronal image showing bone oedema at the scaphoid, RAMRIS score 2/3 (arrow) B) Bone oedema confirmed on axial image C) Matching scan 3 years later showing that a large erosion has developed at this site D) axial image confirming erosion.
Other cohorts of early RA patients have been studied in a similar way with very similar results, including a Norwegian study reported by Haavardsholm et al. [47] who found the baseline bone oedema score to be an independent predictor of both XR and MRI erosive progression. Hetland et al. reported on a Danish cohort that was part of the CIMESTRA (Cyclosporine, Methotrexate, Steroid in RA) clinical trial investigating the effects of a cyclosporine/methotrexate combination in 130 RA patients [48]. These authors reported at the 2-year point that the baseline MRI bone oedema score was the only independent predictor of change in the total SvdH score, and alone explained 41% of the variation in radiographic progression. MRI synovitis, C-reactive protein (CRP), Disease Activity Score (DAS28) and anti-CCP status were not independent risk factors. These findings were confirmed when the group was restudied after 5 years [49]. More recently, Boyesen et al. reported on another cohort of 50 RA patients and found that both baseline and 1-year cumulative measures of MRI synovitis and bone marrow oedema independently predicted 3-year radiographic progression [50]. Looking at the issue from the other direction, data from the New Zealand cohort [4,34,38,46] also revealed that very low levels of MRI inflammation at presentation had a high negative predictive value of 86% for wrist erosions after 2 years, indicating the potential role of a negative MRI to predict a benign outcome, with implications for management [34]. To summarise this section, all studies have suggested that high levels of MRI joint inflammation in early RA, especially bone oedema/osteitis, are associated with a more aggressive and erosive course of disease. The question now arises, should a baseline MRI scan be part of the routine work-up for these patients?
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
507
MRI and cartilage damage in early RA Over the last 3 years, MRI has also helped elucidate the pathways leading to cartilage damage in RA. A recent study describing an MRI cartilage scoring system for use at the rheumatoid wrist included a group of 22 early-RA patients (
Contents lists available at ScienceDirect
Best Practice & Research Clinical Rheumatology journal homepage: www.elsevierhealth.com/berh
4
Imaging in early rheumatoid arthritis Fiona M. McQueen* Department of Molecular Medicine and Pathology, Faculty of Medical and Health Sciences, University of Auckland, 85 Park Rd, Grafton, Auckland, New Zealand
a b s t r a c t Keywords: Rheumatoid arthritis MRI Ultrasound CT scanning Imaging
Imaging in early rheumatoid arthritis (RA) has undergone extraordinary change in recent years and new techniques are now available to help the clinician diagnose and manage patients much more effectively than previously. While established modalities such as plain radiography (X-Ray) remain important, especially for detection of erosions and determining the progression of joint damage, there are many instances where ultrasound (US), magnetic resonance imaging (MRI) and computed tomography (CT) scanning provide added information. MRI and US are now used regularly by clinicians to help diagnose RA in the pre-radiographic stage as they offer improved visualisation of joint erosions. They also have the potential to provide prognostic information as MRI bone oedema/osteitis is linked to the later development of erosions and power Doppler ultrasound (PDUS) joint positivity is also a predictor of joint damage. Nuclear imaging techniques such as single photon emission computed tomography (SPECT) and positron emission tomography (PET) are also highly sensitive for detecting joint change in early RA and pre-RA but not yet used clinically mainly because of accessibility and radiation exposure. MRI, US, scintigraphy, SPECT and PET have all been shown to detect sub-clinical joint inflammation in patients in clinical remission, a state that is now the goal of most treat-to-target management strategies. Thus, imaging may be used to direct therapeutic decision making and MRI is also now being used in clinical trials to determine the impact of disease-suppressing therapy on the course of synovitis and osteitis. As is the case for all tests, it would be unwise to rely completely on any one imaging result, as false positives and negatives can occur for all modalities. Thus, the
* Tel.: þ64 9 3737999x86374; fax: þ64 9 3737677. E-mail address: [email protected]. 1521-6942/$ – see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.berh.2013.09.005
500
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
clinician needs to choose the most relevant and reliable imaging test, while also striving to minimise patient discomfort, radiation burden and economic impact. Ó 2013 Elsevier Ltd. All rights reserved.
Introduction Major advances have been made in the diagnosis and management of rheumatoid arthritis (RA) in the last two decades. Clinicians are now aware that patients do best if their disease is diagnosed early and effective disease-suppressing therapy is begun before permanent joint damage occurs. With more aggressive and effective treatment regimens has come a need to diagnose RA as early as possible, optimally at the time of first presentation, and to monitor accurately the effectiveness of management strategies so that these can be varied according to the ‘treat to target’ philosophy, aiming to achieve clinical remission. Imaging now plays a major role in these activities. Although plain radiography continues to be a mainstay for detection of early bone erosion and cartilage damage, it does not go far enough for the current requirements and so there has been a proliferation of newer, more sensitive modalities including ultrasound (US), magnetic resonance imaging (MRI), computed tomography (CT) and positron emission tomography (PET), to meet the needs of the modern rheumatologist. However, it is important to understand that each of these modalities has strengths and weaknesses and the unwary clinician could definitely face pitfalls. Reader reliability can be variable, which in the context of early disease can make the difference between a positive and a negative diagnosis of RA; joint coverage differs substantially between the different modalities; and there are also issues surrounding exposure to ionising radiation, feasibility and cost that need to be borne in mind when selecting the appropriate modality. This review aims to summarise the latest imaging advances in the field of early RA, including the role played by each modality, firstly in diagnosis and secondly in monitoring disease activity and damage. Insights provided by imaging into disease pathogenesis will also be discussed as will the emerging role of imaging in clinical trials. The clinical utility of each imaging modality will be reviewed including a discussion of cost and accessibility, which may limit the use of some of these techniques to centres where special facilities and expertise are available. Plain radiography Most clinicians have grown up with the concept that plain radiography is integral to the practice of rheumatology and indeed this continues to be the case. X-rays (XR, plain radiography) of the hands (including wrists) and feet is the imaging investigation that virtually all patients require at first presentation to determine baseline joint integrity. Repeated imaging is usually performed over the course of their disease to monitor damage progression. There is a far greater base of experience amongst rheumatologists and radiologists and a much larger literature devoted to plain radiography than to any other imaging modality in RA. Other advantages include ease of access, wide coverage of important joint regions, newer digitised formats that allow easy retrieval and comparison of images longitudinally and relative low cost. Disadvantages include exposure to ionising radiation, which although relatively low for one set of XRs can cumulate over time with a potential impact of patient longevity [1] and, most importantly, a lack of sensitivity for detecting early joint damage and inability to image the inflammatory processes within the joint that precede damage [2]. XRs in the diagnosis of RA Plain radiography is not particularly helpful in making a diagnosis of RA in the majority of patients. XRs of hands and feet are abnormal at first presentation in only 15–30% of patients who eventually fulfil the diagnostic criteria for RA [3,4]. This figure differs depending on symptom duration prior to presentation and is tending to be lower with more modern cohorts presenting earlier [4]. When
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
501
abnormalities are present in early RA, these most often take the form of nonspecific soft-tissue swelling and periarticular osteopaenia, neither of which are sufficient to confirm a clinical diagnosis. The critical radiographic sign of typical marginal erosions is most often first observed in the feet, particularly at the fifth metatarsophalangeal (MTP) joint, probably because radiographs image this region particularly well [5]. The wrist is an important site of early erosion on MRI [4] but is poorly imaged by X-rays and often not informative, because of the complex anatomy and overlapping shadows produced by a twodimensional modality (Fig. 1). Having said this, the radiographic detection of bone erosions in a patient with early inflammatory arthritis is of major clinical importance as it greatly increases the likelihood of an RA diagnosis and has serious prognostic implications, immediately putting the patient into the category of having aggressive, damaging disease. Thus, XR is a test with low sensitivity but high specificity for making a diagnosis of RA and for this reason been incorporated into both sets of American College of Rheumatology (ACR) diagnostic criteria [6,7]. XRs to monitor damage progression in early RA As far as monitoring the progression of damage over time, XR remains the current gold-standard imaging modality. There are a number of good reasons for this. First, as already stated it is widely
Fig. 1. X-Ray and MRI of the hands/dominant wrist in a 68 year old woman with early RA (8 month history). A) X-Ray shows a small subchondral lucency at the radius (arrow), no definite rheumatoid erosions B) T1w coronal 3T-MRI right wrist shows erosions involving radius, scaphoid and lunate (circle) and pole of hamate (arrow). C) Matching coronal proton density T2w MRI scan shows florid bone oedema/osteitis within the lunate (arrowheads) and erosion at pole of hamate (arrow). Slightly different cut does not show other erosions.
502
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
available and offers good coverage of multiple joints likely to be affected by RA. Second, it is quite feasible and indeed standard practice to image these joints on multiple occasions to determine the progression of erosions and joint space narrowing (JSN) indicating cartilage damage [3]. Just how often XRs need to be repeated is up to the discretion of the clinician but a yearly frequency, especially during the first few years of disease when erosive damage progresses most rapidly, is common practice. [8] The use of the Sharp van der Heidje (SvdH) score for quantifying erosions and JSN reinforces the role of XR in the setting of randomised clinical trials (RCTs). This imaging outcome measure has been extensively validated and very high reliability is normally quoted in major clinical trials. In the ATTRACT (Anti-TNF Therapy in RA with Concomitant Therapy) trial investigating the efficacy of infliximab in preventing erosive progression, inter-reader reliability was high with intraclass correlation coefficients (ICCs) of 0.84–0.92 for status scores [9]. Guillemin et al. compared several radiographic scoring systems for quantifying rheumatoid joint damage and found that inter-rater reproducibility was highest with the SvdH score compared with earlier systems such as the Larsen method, but Bland–Altman graphs showed a decrease in reader concordance where there was more severe damage [10]. The smallest detectable difference for change was 3.5% of the maximum score. Recently, Knevel et al. investigated whether XR of one joint area would be sufficient for evaluating erosiveness, but found that if only the hands were imaged (and the feet omitted), 24–40% of RA patients were incorrectly classified as non-erosive [11]. Thus, despite its drawbacks, plain radiography performs well as a test to determine damage progression in RA and this is because of several important factors. XRs profile cortical bone very well producing a clear margin that may be interrupted if an erosion is present or become more closely apposed to a neighbouring cortex if the joint is narrowed. From an imaging point of view, this is advantageous as it means that even a relatively small change can be detected with quite good reliability (leaving aside the technical issues of joint imaging which are largely dealt with by adherence to standardised radiographic views). When these measurements are repeated many times over (a total of 44 joints are imaged for erosions and 40 joints for JSN in the SvdH score) in a disease where joint damage appears to progress at a similar rate at all sites, a useful clinical score is produced. Nevertheless, all of this comes to naught in the patient whose disease is as-yet subradiographic and it is in this setting that other imaging modalities assume serious importance. Tomosynthesis Tomosynthesis is a new technique developed from conventional tomography and involves collecting a number of projected images at different angles with a digital detector, allowing reconstruction (by scrolling though images) at arbitrary depths [12]. This improves detection of erosions especially at complex sites such as the wrist, because there is less projectional overlapping than on standard radiographs. Canella et al. compared tomosynthesis with plain radiography for the detection of wrist erosions in 40 RA patients, using multidetector CT as a gold standard [13]. Significantly more erosions were shown with tomosynthesis than with radiography (a total of 232 detected by CT vs 199 by tomosynthesis vs 140 by plain radiography, p < 0.0001) in this group of RA patients of whom 40% had early disease (scanned within 2 years of presentation). With CT as a reference, the sensitivity and specificity of tomosynthesis were 77.6% and 89.9%, respectively, and for radiography they were 53.9% and 92%, respectively [13]. Thus, tomosynthesis is a relatively low-cost option that has yet to be fully appreciated by the rheumatology community and which greatly increases the pickup of erosions in RA patients, when compared with plain radiography. Further studies into its potential use as a diagnostic aid are warranted. CT scanning There have been relatively few CT studies in the context of early RA but it is an important player in the imaging team and could be very helpful for assisting diagnosis at first presentation, for the same reasons as tomosynthesis. Like plain radiography, it profiles cortical bone extremely well and is generally regarded as the gold standard for imaging erosions against which other modalities are compared [13– 15]. Multidetector helical CT produces very high-quality images which can be stored in a digitised
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
503
format and compared with later images to determine erosion progression. Perry et al. compared CT with MRI at RA wrists and found that although they were closely comparable, CT outperformed MRI in erosion detection [15] (Fig. 2). Dohn et al. confirmed this at the metacarpophalangeal (MCP) joints [16] and then went on to use CT as a means to monitor erosion progression in patients on anti-tumour necrosis factor (anti-TNF) therapy [17]. CT scanning does entail exposure to ionising radiation but the biological impact of this is likely to be relatively low as only the extremities are examined. It does provide less coverage than plain radiography as usually only one joint area is scanned, and given the results of Knevel et al. cited above [11], this could be a significant drawback. It is worth noting that reliability for scoring CT scans in the tomosynthesis study was extremely high with inter-reader ICCs of 0.96–0.99 [13], making this a highly reproducible ‘end’ point which could be clinically useful in early RA. CT techniques to investigate bone loss Microfocal CT (micro-CT) is a high-resolution technique that allows volumetric assessment of bone mineral density and it has been used extensively by Schett et al. to investigate the bone changes associated with erosion in RA. One study from this group included RA patients (16 of whom had recent-onset disease) and 30 healthy controls [18] and showed that while small erosions were observed in controls as well as RA patients, lesions >1.9 mm in diameter were highly specific for RA. RA erosions were mostly found along the radial aspect of the metacarpal heads. This group has gone on to study this modality as a
Fig. 2. MRI and CT scan axial images from a 59 year old male with RA showing matched erosions. A) and B) axial T1w MRI images showing erosions at the 5th metacarpal base on 2 adjacent slices (arrows), C) and D) high resolution CT scans of the same region confirm the same erosions with greater clarity indicating the breach in the bony cortex.
504
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
means to assess the response to the anti-interleukin-6 (anti-IL-6) biological disease-modifying antirheumatic drug (bDMARD), tocilizumab, which induced limited erosion repair in larger sclerotic lesions [19]. Another CT-based imaging technique that has been applied to the investigation of periarticular osteopaenia in early RA is CT osteoabsorptiometry [20]. Mineralisation at the MCP joints was significantly reduced in all groups of RA patients studied, including those with early disease, compared with controls (p < 0.004). A French group has also studied volumetric bone mineral density (vBMD) using a peripheral quantitative CT (high-resolution-peripheral quantitative CT, HR-pQCT) system and confirmed the involvement of the trabecular bone compartment in periarticular osteopaenia [21]. Currently, these extensions of CT are research tools only and are unlikely to be used in a clinical context. MRI scanning MRI is ideally suited to visualising pathology in early RA. This is because it combines tomographic capability with the capacity to image bony structures and cartilage, as well as soft tissues and fluid. It produces images by detecting signal from Hþ ions as they are exposed to a powerful magnetic field and are forced to reorientate their spin direction following the application of electromagnetic pulses [22]. This means that the images are produced in a completely different way from radiographic modalities, which depend upon the attenuation (blocking) of X-rays as they pass through tissue, thus providing clear detail of Caþþ-containing structures such as bone but minimal information about soft tissues. The detection of Hþ ions by MRI means that tissues with high concentrations of water register a high signal on T2-weighted (T2w) and short tau inversion recovery (STIR) sequences, making this an ideal modality for detecting free fluid as well as regions of inflammation. In RA, this means the easy detection of synovitis, tenosynovitis and synovial effusions, and also the virtually unique capacity to image inflammation within bone in the form of bone marrow oedema/osteitis as discussed below. The option to scan after infusion of a paramagnetic gadolinium-containing contrast agent allows further definition of regions of active inflammation where there is enhanced vascularity. These post-contrast T1weighted (T1w) sequences are often produced with machine settings to suppress the signal produced by fat (and therefore referred to as fat saturated (FS)) so that the contrast-enhanced tissues are better displayed. MRI insights into RA pathology MRI bone-marrow oedema in RA is due to the presence of an inflammatory infiltrate within subchondral trabecular bone. This inflammation within the bone marrow has been termed osteitis and it replaces the normal tissue that has a high component of fat, leading to a dramatic change in MRI signal characteristics [23]. However, the bone oedema appearance on MRI scans is nonspecific and does not necessarily indicate a uniform underlying pathology. Any process that results in increased cellularity or vascularity within subcortical bone will cause MRI bone oedema and this is florid in osteomyelitis [24] and also frequently accompanies fracture [25]. Osteonecrosis or osteoarthritis (OA) is also associated with MRI bone oedema and these pathologies may co-exist with RA [26]. In OA, increased Hþ signal within bone is due to replacement of fat-containing marrow by fibrotic repair tissue as well as regions of haemorrhage and inflammation [27]. A considerable body of evidence now exists to indicate that MRI bone oedema has special significance in RA as it is not only common (occurring in 40–60% of cases) but is also an adverse prognostic sign being associated with an aggressive, erosive disease phenotype [28]. Imaging/histological studies have been performed where sites of bone oedema were identified on pre-surgical MRI scans in RA patients undergoing joint replacement. Subsequent histological examination of resected bone revealed an inflammatory infiltrate in the subarticular region [29,30]. This was characterised by immunostaining as lymphoplasmacytic, containing T and B lymphocytes, plasma cells and macrophages adjacent to osteoclasts. Intense staining for RANKL (receptor of activated nuclear factor-kappa B ligand) was noted [31]. There were many similarities to the rheumatoid synovitis lesion, including follicular clusters of B cells reminiscent of germinal centres [26]. This work supported the proposal that the osteitis lesion represents a second focus of pathology in RA [32] and could drive the erosive process [28,33]. Therefore, the finding of MRI bone oedema, especially at the wrist, which seems to be a particularly informative site, has importance in early RA.
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
505
MRI in the diagnosis of RA In 1998, McQueen et al. reported that 45% of a cohort of early RA patients, recruited within 6 months of the onset of symptoms, had evidence of bone erosion on MRI scanning of the dominant wrist while only 15% had radiographic erosions [4]. A follow-up study tracking the progress of these MRI erosions revealed that 21% and 26% were observed on XR, 1 and 2 years later, respectively [34]. Although initially it was suspected that this disparity might mean that MRI erosions were often not ‘true’ erosions, later data suggested that the more likely explanation was a deficiency in the XR detection of erosions at the wrist. Dohn et al. clarified this by comparing CT, MRI and XR of the second-to-fifth MCP joints of one hand for the pick-up of erosions in 17 RA patients and four healthy controls. These different modalities detected 77, 62 and 12 erosions, respectively, indicating that CT and MRI are 5–6 times more sensitive than XR for erosion detection [14]. The influence of time is also important in RA and Ostergaard et al. found that new erosions were detectable on wrist MRI, a median of 2 years earlier than they were apparent on XR [35]. This group also used 0.2-T extremity-MRI (E-MRI) to show that conventional radiography was unable to detect small MRI erosions involving 1 h, hand involvement and MRI bone oedema and was correct in 82% of the 27 patients who developed RA after 1 year. Similar findings were reported by Tamai et al. in their cohort with undifferentiated arthritis [42,43]. All the 22 patients who had MRI bone oedema at the wrist and were anti-citrullinated peptide/protein antibody (ACPA) positive progressed to RA at 1 year, with a positive predictive value (PPV) of 100%. Narvaez found that MRI synovitis with erosions and/or bone oedema predicted a later diagnosis of RA in a similar group of patients. In their study, MRI scan findings correctly predicted the onset of RA in 31/33 patients, equating with a sensitivity of 100% and a specificity of 78%, compared with ACPA positivity which alone gave a sensitivity of 23% and a specificity of 100% [44]. To look even earlier at the pre-RA group of anti-cyclic citrullinated peptide-positive (antiCCPþ) patients with arthralgia but no true synovitis, Krabben et al. recently reported a correlation between the MRI inflammation score (RAMRIS synovitis plus bone oedema scores) and the presence of ACPA, echoing the observations of Tamai et al. [42] and suggesting that these could be linked during disease initiation [45]. MRI to predict joint damage progression in early RA The association between MRI bone oedema and the subsequent development of radiographic erosions in RA was first reported by McQueen et al., in 2003 [38]. These patients had baseline MRI scans of the wrist at presentation as described above [4] and were then re-examined radiographically after 6 years. On univariate regression, the baseline MRI bone oedema score (but not the synovitis score) was predictive of XR erosion and JSN scores, separately, and combined as the total SvdH score. An optimal model incorporating MRI features and levels of inflammatory markers explained 59% of the variance in the 6-year total Sharp score. Site-specific analysis revealed that MRI bone erosion was more likely to be detected after 6 years if bone oedema was present at baseline, with an odds ratio (OR) of 6.5. Interestingly, baseline bone oedema was also found to be a predictor of physical function after 8 years [46]. Fig. 3 shows an example of bone oedema involving the scaphoid evolving to bone erosion over a 3-year period in a typical early RA patient.
506
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
Fig. 3. T1w MRI scans from a 58 year old man with early RA indicating the transition from bone oedema/osteitis to erosion A) Coronal image showing bone oedema at the scaphoid, RAMRIS score 2/3 (arrow) B) Bone oedema confirmed on axial image C) Matching scan 3 years later showing that a large erosion has developed at this site D) axial image confirming erosion.
Other cohorts of early RA patients have been studied in a similar way with very similar results, including a Norwegian study reported by Haavardsholm et al. [47] who found the baseline bone oedema score to be an independent predictor of both XR and MRI erosive progression. Hetland et al. reported on a Danish cohort that was part of the CIMESTRA (Cyclosporine, Methotrexate, Steroid in RA) clinical trial investigating the effects of a cyclosporine/methotrexate combination in 130 RA patients [48]. These authors reported at the 2-year point that the baseline MRI bone oedema score was the only independent predictor of change in the total SvdH score, and alone explained 41% of the variation in radiographic progression. MRI synovitis, C-reactive protein (CRP), Disease Activity Score (DAS28) and anti-CCP status were not independent risk factors. These findings were confirmed when the group was restudied after 5 years [49]. More recently, Boyesen et al. reported on another cohort of 50 RA patients and found that both baseline and 1-year cumulative measures of MRI synovitis and bone marrow oedema independently predicted 3-year radiographic progression [50]. Looking at the issue from the other direction, data from the New Zealand cohort [4,34,38,46] also revealed that very low levels of MRI inflammation at presentation had a high negative predictive value of 86% for wrist erosions after 2 years, indicating the potential role of a negative MRI to predict a benign outcome, with implications for management [34]. To summarise this section, all studies have suggested that high levels of MRI joint inflammation in early RA, especially bone oedema/osteitis, are associated with a more aggressive and erosive course of disease. The question now arises, should a baseline MRI scan be part of the routine work-up for these patients?
F.M. McQueen / Best Practice & Research Clinical Rheumatology 27 (2013) 499–522
507
MRI and cartilage damage in early RA Over the last 3 years, MRI has also helped elucidate the pathways leading to cartilage damage in RA. A recent study describing an MRI cartilage scoring system for use at the rheumatoid wrist included a group of 22 early-RA patients (
