Osteoarthritis has traditionally been imaged with conventional radiographs. However, in recent years, novel imaging techniques such as ultrasonography have been utilised to obtain a better understand of this disease. Although the application of ultrasonography to inflammatory diseases has been common and widespread, it has been applied to osteoarthritis less frequently.

Two recent systematic reviews by Joshua and colleagues1 2 examined the validity of ultrasonography as an outcome measure according to the principles of truth and discrimination; components of the OMERACT filter. The first addressed the validity and reproducibility of ultrasonography in assessing synovitis only;2 the second, power Doppler in musculoskeletal disease.1 These reviews demonstrated that most of the work validating ultrasonography has been undertaken in inflammatory diseases, such as rheumatoid arthritis, and has largely studied the hand, knee or ankle joints. A minority of work examined in these systematic reviews pertained to either synovitis or power Doppler signal in osteoarthritis. In the first review, 10 of the 54 manuscripts reviewed utilised ultrasonography to assess synovitis in osteoarthritis.2 Six of the 53 manuscripts reviewed in the second review article utilised Doppler signal in osteoarthritis.1

There are no published systematic reviews focusing on the application of ultrasonography to osteoarthritis. We wanted to examine the published literature to assess the role of ultrasonography in assessing structural pathology in osteoarthritis, and to examine the validity of ultrasonography as an assessment tool in osteoarthritis, with particular respect to the performance metrics of these tools. To do this, a systematic review was undertaken. The function of this review is to update the literature reviews by Joshua and colleagues,1 2 with a focus on osteoarthritis, and to broaden the search to include ultrasonography-detectable pathologies other than synovitis and Doppler signal, including tendon and ligament disorders, cartilage pathology and cortical pathology including osteophytosis. In addition, definitions of pathologies and scoring systems utilised in osteoarthritis were examined.

METHODS

Pubmed was searched for articles first published between 1955 and June 2008. The search was limited to humans and English language. The search terms were “[ultrasound or sonography] and osteoarthritis”. The titles and abstracts of the 244 manuscripts identified were reviewed. Medline was searched using [MESH subject heading “ultrasonography” or the keyword “ultrasonography”] and [MESH headings “osteoarthritis” or “osteoarthritis, knee” or “osteoarthritis, hip” or the keyword “osteoarthritis”]. The search was limited to humans and English language. A total of 148 articles was identified. Of the articles identified, 147 were duplicates, therefore the titles and abstracts of 245 articles were assessed with regard to inclusion and exclusion criteria. Articles were excluded if they were not original articles pertaining to the use of B-mode ultrasonography in the assessment of a joint in a cohort of subjects with a diagnosis of osteoarthritis at baseline. Review articles (n  =  48), case reports (n  =  15), letters (n  =  1), position statements (n  =  1), recommendations (n  =  2), practice audits (n  =  1), pictorial reviews (n  =  1), studies ex vivo (n  =  7) and second reports (n  =  2) were excluded. In addition, articles that utilised ultrasonography only for guiding injections and did not report any validity data or findings of the ultrasonography examination were excluded (n  =  6). Manuscripts utilising ultrasonography to measure only rotational angles were also excluded (n  =  3). Of the remaining articles, 58 did not assess a cohort with a diagnosis of osteoarthritis at baseline, 46 did not utilise B-mode ultrasonography and 16 did not examine a joint structure. An additional nine publications were identified by experts in the field and searching the bibliographies of recent review articles. Therefore 47 manuscripts were included in this review (see supplemental fig 1 available online only and table 1). Data were extracted and inserted into a spreadsheet developed for this review based on similar published reviews.1 2

This covered descriptive aspects of trial methodology, a description of the ultrasonography-detected findings in osteoarthritis cohorts, issues relating to the validity of ultrasonography in assessing osteoarthritis, the relationship between ultrasonography findings and symptoms of osteoarthritis and the clinical utility of ultrasonography in osteoarthritis.

The performance metrics were evaluated using criterion and construct validity, reliability and responsiveness to change. Criterion (or direct) validity is determined by comparing the technique with a gold standard.50 For the purpose of this review, this was considered a comparison against either direct macroscopic or microscopic visualisation of the pathology, for example by arthroscopy, examination during surgery, or histopathological examination. Construct (or indirect) validity is determined by comparing the technique against other modalities known to measure the same pathology; for example, comparing ultrasonography-detected synovitis against magnetic resonance imaging (MRI) or computed tomography (CT)-detected synovitis.50 Comparison against MRI, scintigraphy, conventional radiography, clinical examination, laboratory tests and bone mineral density were all considered measures of construct validity.

Reproducibility is intrinsic to both the validity of a technique as an outcome in clinical trials and also to its ability to demonstrate changes over time. Reproducibility is generally determined through examining inter and intra-observer reliability. For this review, both were subanalysed according to whether the assessments were made through repeated image acquisition or re-reading stored images. In addition, responsiveness to changes with time were also recorded, as these examine discrimination and also further address construct validity.50 A brief summary of the findings of each manuscript was included.

RESULTS

Characteristics of the studies

Forty-seven articles published between 1982 and 2008 were included in the review. The findings are summarised in table 1. The majority of studies were published after 2000. The knee has been examined more extensively than other joints, followed by the hip, hand, foot, tempromandibular joint and sternoclavicular joint. The definition of osteoarthritis was not consistent and was not specified in approximately half the papers. American College of Rheumatology criteria were often used to identify clinical disease. Radiographic criteria were also commonly used, using different Kellgren Lawrence or Altman grades to define the cohort. Other studies used diagnostic criteria specific to their study, such as a combination of clinical symptoms, signs, the American College of Rheumatology criteria and radiographic criteria. Some manuscripts used terms such as “clinical diagnosis” or “typical changes” without further clarification. It was also common for no definition to be provided.

Technical aspects of ultrasonography machines and image acquisition reported in the studies

The vast majority of studies employed grey-scale ultrasonography, and most (42, 89%) reported the transducer characteristics. Doppler, either power (six, 13%) or colour (three, 6%) were used in 10 studies, and contrast was examined in only one study. The Doppler specifications were reported in five, were unclear in one and were not reported in one manuscript.

The majority (40, 85%) of manuscripts provided some description of the probe and joint position during image acquisition; however, there was variability between studies imaging the same joint region as to how the images were acquired.

Pathologies imaged and scoring systems

The pathologies examined most commonly were effusion, followed by synovial thickening or hypertrophy, cartilage parameters, vascularity, Baker’s cysts, osteophytes, tendon and ligament abnormalities, meniscal changes, bursitis, erosions and panniculitis. Definitions of the imaging appearance of the pathology imaged were provided in approximately half of the studies, and again, no standard definition of pathology was used across the studies (tables 2, 3, 4 and 5). The ultrasonography appearance of cartilage, when defined, was generally considered a sonolucent or anechoic band overlying cortex. Cartilage thinning was the most common pathology examined, although clarity and sharpness was also measured, although definitions of these abnormalities were not given.6 26 Tendon and ligament pathologies were also rarely defined. Enthesitis was examined by one group,21 23 with definitions encompassing features of heterogeneous hypoechogeneicity, tendon thickening, cortical irregularities (erosions and enthesophytes) and oedema, although the Doppler signal was not examined in these studies. Cortical irregularities have similarly rarely been defined. Erosions have been defined in one study,29 and osteophytes in two studies.46 47 Synovial pathologies, including synovial hypertrophy, effusion and Doppler signal, were most often studied and usually defined. As a definition of synovial hypertrophy and effusion has been published by the OMERACT ultrasonography group,51 which can be used in future studies, the definitions used in previous published manuscripts are perhaps less interesting than other aspects of the imaging. For example, in reviewing the articles it became clear that there was no standardisation with regard to the positioning of the joint, planes in which images were obtained and the scoring of synovial pathologies. Eleven of the studies examining synovitis clearly differentiated between synovial hypertrophy and effusion, whereas in 12 studies they were either considered together or it was unclear. In addition, some studies required an arbitrary minimal thickness of synovial hypertrophy and effusion28 30 44 before considering the pathology to be present. The scoring systems used were usually reported, but again demonstrated great variety, being either dichotomous, ordinal or continuous (tables 2, 3, 4 and 5).

Validity of ultrasonography

Most studies addressed the construct validity of ultrasonography (n  =  27), with little examination of criterion validity (n  =  9). Two studies found reasonable correlation between ultrasonography-detected cartilage thickness and histological cartilage thickness6 9 and one study demonstrated reasonable correlation between ultrasonography-detected cartilage thickness and MRI.11 A paucity of information is available about the construct validity of ultrasonography-detected qualitative cartilage changes (table 2), and quantitative changes were limited to measurements of thickness, as unlike MRI, it is difficult to utilise ultrasonography to detected total volumes. Tendon and ligament changes were usually compared against clinical examination, with varying results. For example, little correlation was found between ultrasonography and clinical diagnoses of anserine tenobursitis,32 whereas there was good correlation between ultrasonography and clinical and radiographic changes of enthesitis at the shoulder and foot.21 23

The validity of ultrasonography in detecting cortical irregularities was infrequently studied (table 4), with ultrasonography being found to be more sensitive to osteophytosis than radiography in the small joints of the hand,46 but less sensitive to erosions.29 This was thought partly to be because osteophytes overhanging erosions may shadow underlying erosions preventing visualisation by ultrasonography. It may also be related to the positioning of the erosions. Whereas rheumatoid erosions tend to be peri-articular, osteoarthritis erosions, as seen radiographically, may be within the central portion of the joint and inaccessible with ultrasound.

Ultrasound performs comparably to MRI in detecting effusion, synovial hypertrophy and popliteal cysts (tables 2, 3, 4 and 5). The validity of ultrasonography-detected cartilage changes has only been assessed in comparison with MRI or histology at the knee joint (table 2). Ultrasonography was more sensitive and specific than clinical examination in detecting effusion and synovial hypertrophy, although this has been examined exclusively at the knee joint (table 5). The knee joint has also been the focus of comparison between ultrasonography-detected synovial pathology and MRI and arthroscopy.11 27 The ability of ultrasonography to detect synovitis changes has been examined at the hip,22 and fluid aspiration has been compared with ultrasonography-detected effusions in the hip and hand.15 39

No consistent relationship between clinical symptoms and ultrasonography-detected pathology is found in this review, although symptomatic joints tend to have more ultrasonography-detected pathology than controls/healthy joints.

Reproducibility of ultrasonography

A minority of studies reported any reproducibility data, although when reported it was reasonably good. Intra-reader acquisition was reported in three studies, intra-reader reporting was reported in four, inter-reader acquisition was reported in three and inter-reader reporting was reported in two.

Discriminate validity of ultrasonography

Only eight studies examined the ability of ultrasonography to detected changes over time. Those studies, the joints, interventions and pathologies studied are presented in table 6.

The general trends were a reduction in pathology with time after therapy, although only one of the studies was a randomised controlled trial, the others being observational case series.

DISCUSSION

This review demonstrates that since the start of the new millennium there has been increasing evidence of the application of ultrasonography to osteoarthritis. However, for ultrasonography to be fully useful in assessing therapies and responses, it first needs to be validated as an outcome tool. In this review, we have identified manuscripts that use ultrasonography to evaluate osteoarthritis and demonstrated that further work is required to validate ultrasonography in osteoarthritis.

Generally, the descriptions of ultrasonography technicalities, such as information about the machine and probe specifications and the position of the scan in obtaining images was adequately described. The quality of reporting of the pathologies imaged, their definitions and scoring was less well described and, when present, demonstrated marked heterogeneity between studies. There are no well accepted definitions of ultrasonography pathology in osteoarthritis, although definitions of synovial hypertrophy, effusion, tenosynovitis, enthesitis and erosion have been developed by the OMERACT ultrasonography group for use in inflammatory arthritis.51 These definitions were applied to osteoarthritis in some publications,46 47 but not routinely, which may reflect the fact that the recommendations were only published in late 2005. In addition, the validity of applying definitions developed for inflammatory arthritis to osteoarthritis needs consideration.

The scoring systems utilised were also not always described, and again demonstrated marked heterogeneity, generally being dichotomous, ordinal (based on qualitative, semiquantitative or quantitative domains) or continuous scales (such as simple numeric counts or measuring in millimetres). Most of the literature examined pathology in grey scale, with a paucity of publications utilising Doppler or contrast agents. The OMERACT ultrasonography group has recently been working towards recommendations for a scoring system for synovitis in inflammatory arthritis, which will soon be published. This is too new to see reflected in the published literature; however, again, whether this is applicable to osteoarthritis needs consideration.

Whereas ultrasonography appears to be more sensitive for the detection of synovitis in osteoarthritis than clinical examination, with reasonable sensitivity compared with MRI or histology, there is little evidence to confirm the validity of ultrasonography in detecting bony pathology in osteoarthritis, and the evidence regarding the detection of cartilage pathology is largely limited to the detection of focal cartilage thickness. The clinical utility of ultrasonography in detecting cartilage in vivo is questioned, as the physical properties of ultrasonography make load-bearing cartilage difficult to image reliably due to acoustic shadowing. This review has also highlighted a paucity of information on the responsiveness of ultrasonography in osteoarthritis and a lack of information about the feasibility of this imaging technique. Furthermore, there is a paucity of reliability data presented in the literature with regard to inter-reader and intra-reader reliability in image acquisition and the scoring of stored images.

This review has limitations. First, only two databases were searched, meaning that some manuscripts may have been missed. However, the two databases searched are arguably the most utilised in medical literature searches and the extensive duplication of manuscripts found was reassuring. Second, we limited the search to studies utilising ultrasonography in osteoarthritis, excluding studies that imaged joint pathologies in other joint diseases only. If the validity of ultrasonography in detecting synovial, cortical, cartilage and tendon changes in other joint diseases (ie, rheumatoid arthritis) can automatically be applied to osteoarthritis, then the scope of this review is limited. However, it may not be correct to assume that validity and reproducibility in one disease implies validity and reproducibility in another. These metrics are likely to be influenced by disease-specific factors, such as the degree of pathology, distribution of pathology, subtle differences in pathologies and response to therapy. For example, a manuscript examined in this review found ultrasonography less sensitive than radiography to cortical erosions in osteoarthritis of the small joints of the hand,29 whereas it is well accepted that ultrasonography is more sensitive to erosions in the small joints of the hand in rheumatoid arthritis.52 This is thought to be a result of osteophytes (a pathognomonic feature of osteoarthritis but not rheumatoid arthritis) obscuring ultrasonography visualisation of erosions in osteoarthritis.

A further issue to consider regarding this review is that its evaluation of the role of ultrasonography in osteoarthritis is limited by being systematic (with strict inclusion and exclusion criteria) and focusing on published evidence but excluding, for example, pictorial reviews that may provide insight into the way the ultrasonography appearance of pathology in osteoarthritis has been defined by some experts. The reason for excluding such reports was that although the definitions they included may have good face validity, the further validity or reliability of these definitions cannot be assessed from the published literature. Investigation of valuable information contained in such publications will be warranted in devising consensus definitions.

Another limitation (albeit a reflection of the published literature, rather than a methodological problem in this review) is that most of the studies included were undertaken with ultrasonography machines with now outdated technology. Modern imaging technology may have better sensitivity, specificity and further aid our understanding of osteoarthritis; it has recently been hypothesised that the pathology of the finger collateral ligaments may play a causal role in osteoarthritis,53 but these ligaments may have been difficult to identify with early high-resolution ultrasonography technology. This review may need updating in the near future, given that the OMERACT51 definitions were published relatively recently, machine technology is improving rapidly, and international organisations such as OMERACT and OARSI are developing research agendas focusing on ultrasonography in osteoarthritis.

Ultrasonography is an imaging technique that may be useful in the diagnosis and management of osteoarthritis, both in clinical trials and in practice. Application of this imaging methodology to osteoarthritis has aided the understanding of the disease process, the relationship between structure and symptoms and may aid in the assessment of future therapies. Whereas previous reviews have demonstrated reasonable validation of ultrasonography in inflammatory arthritis,1 2 further work is required to validate ultrasonography as an outcome tool in osteoarthritis.