Validation and reproducibility of ultrasonography in the detection of synovitis in the knee: A comparison with arthroscopy and clinical examination




Accurate detection of synovitis is important in both the diagnosis and outcome assessment of arthritis. This study was undertaken to assess the validity and reproducibility of ultrasonography (US) as a means of detecting synovitis in the knee, by comparing US findings with findings of arthroscopy and clinical examination.


Sixty consecutive patients with knee pain due to various arthritides had a clinical examination and US of their knee performed immediately prior to arthroscopy. All 3 assessments were performed by different clinicians who were blinded to the results obtained with the other modalities. US and clinical examination were compared with arthroscopically detected synovitis as the gold standard. Data from a subset of patients were used for calculating the inter- and intrareader reproducibility of US results, using a standard dichotomous (absence/presence of synovitis) as well as a graded (absence/grade of synovitis) scoring system.


With the use of arthroscopy as the gold standard, US had a higher sensitivity (98% versus 85%), specificity (88% versus 25%), accuracy (97% versus 77%), positive predictive value (98% versus 88%), and negative predictive value (88% versus 20%) compared with clinical examination. The Cohen kappa values for inter- and intrareader reproducibility of US for distinguishing between presence and absence of synovitis were 0.71 and 0.85, respectively (P < 0.05 for both). The weighted kappa values for distinguishing grade of synovitis were 0.65 for inter- and 0.74 for intrareader reproducibility. The kappa value for intrareader reproducibility of arthroscopy results was 0.88.


Ultrasonography is a valid and reproducible technique for detecting synovitis in the knee, and is more accurate than clinical examination. It may be valuable as a tool in studies investigating pain, diagnosis, and treatment response in knee arthritis.

Synovitis is the usual presenting sign of inflammatory arthritis and has a central role in the progression of joint damage in rheumatoid arthritis (RA) (1). Synovitis is also frequently present in osteoarthritis (OA), especially with advanced structural damage (2), and has recently been associated with the degree of pain (3) and predicted progression of cartilage loss (4) in the knee. Clinical assessment of the knee for synovitis has poor sensitivity and reproducibility (5–7), and a more reliable method of detecting and quantifying synovitis is therefore desirable.

Arthroscopy allows direct visualization of the synovial membrane and structures within the joint compartment. It is frequently used as a method of quantifying synovitis of the knee, and has been validated against histologic findings in both OA and inflammatory arthritis (8, 9). In addition, recent evidence suggests that synovitis visible on arthroscopy is a predictor of progression of OA (4). Magnetic resonance imaging (MRI) has been shown to be a sensitive tool for detecting synovitis when compared with arthroscopy (10), but issues relating to cost may affect patient access.

The use of ultrasonography (US) is increasing because it is relatively inexpensive, noninvasive, does not involve exposure to ionizing radiation, and as a result can be used repeatedly on the same patient in multiple anatomic areas. It therefore has the potential for use in assessing the extent of synovitis and changes in the synovial volume over time. However, there are few validation studies of US and its ability to detect synovitis in large joints. A relatively small number of studies have compared synovitis detected by gray-scale US with MRI (11) and with arthroscopy (12–14), although the latter studies assessed only patients with inflammatory arthritis with persistent knee synovitis who were referred for synovectomy (a group who may be expected to have a greater degree of synovitis). A recent study assessed the relatively new technique of power Doppler US in the knee, comparing its results with histopathologic findings of synovial membrane vascularity in 23 patients with RA or OA who were undergoing arthroplasty; the findings suggested good correlation (15).

Whereas US has demonstrated good reproducibility for detecting erosions in the metacarpophalangeal joint (16) and assessing cartilage thickness in the knee (17), there are no published data on its reproducibility in the assessment of knee synovitis. In this study, US and clinical examination were compared with arthroscopy for the detection of synovitis in a large group of patients with knee pain. We also assessed the inter- and intrareader reproducibility of US, as well as intrareader reproducibility of arthroscopy, in a subset of patients.



Consecutive patients with knee pain who were referred for arthroscopy of the knee were included in the study. All patients had a clinical examination and US performed by different physicians, blinded to one another's findings, prior to arthroscopy. The indications for referral for arthroscopy were either diagnostic or therapeutic, and arthroscopy was performed on the more severely affected knee only, in all patients. Sixty knees from 60 patients were included in the study (30 female patients, 30 right knees). The mean age of the patients was 52 years (range 18–79), and clinical diagnoses included OA (n = 19), RA (n = 16), seronegative spondylarthropathy (n = 14), oligoarthritis (n = 4), gout (n = 4), and anterior knee pain (n = 3). The local research ethics committee gave approval for the study, and informed consent was obtained from all patients before the procedures.

Clinical examination.

All patients had a standard clinical examination of the index knee performed by the assessing physician, who documented the presence or absence of synovitis, defined as the presence of an effusion (positive patella tap result or bulge sign) or palpable synovial thickening.


An ATL (Advanced Technology Laboratories, Bothel, WA) HDI 3000 US machine with a linear-array 10–5-MHz “hockey stick” transducer was used to examine the knee on the same day as (but prior to) arthroscopy. Three areas of the knee, the medial compartment (MC), lateral compartment (LC), and suprapatellar pouch (SPP), were assessed. The MC was defined as the area medial to the patella running inferiorly past the medial joint line to the inferomedial aspect of the joint capsule, the LC as the area lateral to the patella running inferiorly past the lateral joint line to the inferolateral aspect of the joint capsule, and the SPP as the area above the patella. The retropatellar space was not included in the comparison because this space cannot be fully visualized with US.

One physician (ZK) performed US on all patients; in the last 10 patients, US was also performed by an additional physician (RJW). Both were experienced in performing US and were blinded to the indication for the arthroscopy and clinical examination findings, as well as the findings of the other ultrasonographer.

Ultrasonographic synovitis was defined as the presence of a thickened synovial membrane, identified as the hypoechoic structure adjacent to the anechoic joint effusion and more hyperechoic fatty tissue (Figure 1). The patient was examined initially with the knee fully extended, then partially flexed to 30°, and each compartment was scanned in transverse and longitudinal planes. Moving the knee from the extended to the partially flexed position allowed for better visualization of all 3 compartments, but also caused movement of any synovial fluid if present, which helped create a better acoustic interface. Small amounts of fluid were detected by contraction of the quadriceps muscle and confirmed by transducer ballottement (Figure 2). Images of each compartment were recorded and stored on acetates, with only the patient's study number and compartment identified.

Figure 1.

Ultrasonography in a patient with rheumatoid arthritis: longitudinal view of the lateral compartment with the knee extended. Hypoechoic synovial membrane (+, ×, ∗) with villi-like protrusions adjacent to the anechoic joint effusion, and more hyperechoic fatty tissue, are seen. SF = synovial fluid; LC = lateral condyle.

Figure 2.

Ultrasonography, with effusion demonstrated by transducer ballottement: longitudinal view of the suprapatellar pouch with the knee flexed. A, Image obtained under normal transducer pressure. B, Increased transducer pressure or ballottement. Arrows indicate apposition of the synovial membrane (∗) under pressure from the transducer. SF = synovial fluid; MC = medial condyle.

Synovitis was documented using a standard dichotomous scale (either present or absent) for each of the compartments, with a semiquantitative assessment for degree of synovitis also recorded (normal = no synovitis; mild = flat, thickened synovium; moderate = thickened synovium with few villi-like protrusions; severe = marked thickening with multiple villi-like protrusions). The presence or absence of synovial fluid was also documented.

Interreader reproducibility was calculated by comparing the findings of the 2 ultrasonographers in the 10 patients (30 compartments) who were examined independently and sequentially on the same day. Intraobserver reproducibility was calculated using stored images of 30 randomly chosen compartments reviewed 1 month after the end of the study period.


A standard procedure was used to assess all knees. Arthroscopy was performed with the patient under local anesthesia, using a 2.7-mm rigid arthroscope (Storz, Heidelberg, Germany) and a heavy 5-mm drainage cannula. The physician performing the arthroscopy (MQ) was blinded with regard to the findings of clinical examination and US. Prior to insertion of the arthroscope, the knee was distended with 40–60 ml of normal saline. The arthroscope was inserted through the inferolateral portal with the drainage cannula in the supralateral portal, and the knee was lavaged with normal saline throughout the procedure. The knee was assessed in a structured manner, with the SPP visualized first, followed by the MC and LC.

Arthroscopically evident synovitis was defined as the presence of synovial membrane hypertrophy, documented as either increased granularity or villous hypertrophy (Figure 3) in each compartment. Increased granularity is seen as a milder form of synovitis, with uniform low-grade thickening of the membrane as opposed to villous hypertrophy, in which there is increased thickness of the membrane with villi-like protrusions into the joint cavity. A 100-mm visual analog scale (VAS) to allow for comparison with an overall assessment of the knee, including the retropatellar space, was also completed by the arthroscopist. Images were captured digitally (Vision Digital Management System 635; Dyonics, Andover, MA), and intrareader reproducibility was calculated using 30 randomly chosen images reviewed 1 month after the end of the study period.

Figure 3.

Arthroscopic images. A, Normal synovium. B, Increased granularity, with the classic “cobblestone” appearance of the synovium, in a patient with rheumatoid arthritis. C, Corresponding arthroscopy of the rheumatoid arthritis patient in Figure 1, demonstrating villous hypertrophy.

Statistical analysis.

Sensitivity, specificity, and positive and negative predictive values of US and clinical examination were calculated using arthroscopy, and then arthroscopic VAS, as the gold standard. These values were then also calculated comparing clinical examination results with US as the gold standard. Intrareader reproducibility for arthroscopy (0 = no synovitis; 1 = granularity; 2 = villous hypertrophy) as well as inter- and intrareader reproducibility for the standard dichotomous US scoring method (0 = no synovitis; 1 = synovitis) were calculated using the Cohen kappa test. The reproducibility of the semiquantitative US scoring method for distinguishing grade of synovitis (0 = no synovitis; 1 = mild synovitis; 2 = moderate synovitis; 3 = severe synovitis) was calculated using a standard weighted kappa statistic.



Arthroscopy could be used to assess 175 of 180 compartments (97%) in 60 patients. Access was difficult in 5 compartments due to plicae or adhesions. With arthroscopy, synovitis was detected in 52 of 60 patients (87%) and 119 of 175 compartments (68%) (differentiated as increased granularity in 43 compartments and villous hypertrophy in 76).

Ultrasonography versus arthroscopy.

With US, synovitis was detected in 52 of 60 patients (87%), and synovial fluid was detected in all of these plus 3 other patients. Synovitis was detected in 121 of 175 compartments (69%) assessed by arthroscopy and in 4 of 5 of the compartments that could not be assessed by arthroscopy.


Of the 52 patients with arthroscopically detected synovitis (“positive arthroscopy result”), the US result was positive in 51 (98%); of the 8 with negative arthroscopy results, the US result was negative in 7 (88%) (i.e., there was 1 false-positive and 1 false-negative result with US). Using arthroscopy as the gold standard, the sensitivity and positive predictive value for US were both 98%, and the specificity and negative predictive value were both 88%, resulting in an overall accuracy of 97% (Table 1). Synovial fluid was detected by US in 55 of 60 patients (92%), including all 52 patients with a positive arthroscopy result plus 3 others (including the 1 who had a negative arthroscopy result for synovitis and a false-positive US result).

Table 1. Results obtained with ultrasound versus arthroscopy (used as the gold standard), clinical assessment versus arthroscopy, clinical assessment versus ultrasound, ultrasound versus arthroscopy VAS, and clinical assessment versus arthroscopy VAS, in the identification of synovitis*
 US vs. AR (60 patients)CA vs. AR (60 patients)CA vs. US (60 patients)US vs. VAS (60 patients)CA vs. VAS (60 patients)US vs. AR (175 compartments)
  • *

    VAS = visual analog scale; US = ultrasound; AR = arthroscopy; CA = clinical assessment; PPV = positive predictive value; NPV = negative predictive value.

Sensitivity, %988587968590
Specificity, %882538862975
PPV, %988890989088
NPV, %882030752078
Accuracy, %977780957885


Synovitis was detected by US in 107 of 119 compartments with a positive arthroscopy result (90%) and was negative by US in 42 of 56 compartments in which no synovitis was documented at arthroscopy (75%). This resulted in a positive predictive value of 88%, a negative predictive value of 78%, and an accuracy of 85% (Table 1). There was insignificant variation in sensitivity between the 3 compartments (data not shown). US was more sensitive at detecting villous hypertrophy (70 of 76 [92%]) compared with increased granularity (37 of 43 [86%]) as documented by arthroscopy.

Clinical examination versus arthroscopy.

Fifty of the 60 patients (83%) were judged to have synovitis by clinical examination. Using arthroscopy as the gold standard for synovitis detection, clinical examination correctly identified synovitis in 44 of 52 patients (85%); 8 patients had arthroscopically documented synovitis that was undetected by clinical examination, including the 1 patient with a false-negative US result. Two of 8 patients (25%) in whom no synovitis was found by arthroscopy were also judged to have no synovitis by clinical examination (i.e., there were 6 false-positive results); synovial fluid was detected by US in 2 of these 6 patients (including the 1 who had a false-positive US result). The positive and negative predictive values of clinical examination were 88% and 20%, respectively, resulting in an accuracy of 77% (Table 1).

Clinical examination versus ultrasonography.

Compared with US, results of clinical examination for synovitis were positive in 45 of 52 patients (87%) and negative in 3 of 8 (38%). This resulted in a positive predictive value of 90%, a negative predictive value of 30%, and an accuracy of 80% for clinical examination versus US (Table 1). All 7 of the patients with US-detected synovitis not found on clinical examination had arthroscopically evident synovitis.

Arthroscopic VAS.

The VAS scores for overall arthroscopically evident synovitis (including the retropatellar space) ranged from 0 mm to 99 mm, with a mean of 34 mm. The VAS score was positive (i.e., >0) in 53 of 60 patients (88%); 52 of these 53 (98%) had arthroscopically evident synovitis in the compartments under comparison, and 1 patient had isolated retropatellar granular hypertrophy and a VAS of 13 mm.

US detected synovitis in 51 of the 53 patients (96%) who had positive VAS scores, with false-negative results in the above-mentioned patient with retropatellar disease (VAS 13 mm) and in another patient who had a VAS score of 7 mm (mean 10 mm). Both patients with false-negative US results had synovial fluid detected on US. The US result was negative in 6 of 7 patients (86%) who had VAS scores of 0. The positive and negative predictive values of US compared with arthroscopy VAS for overall synovitis were 98% and 75%, respectively, resulting in an accuracy of 95% (Table 1).

Synovitis was judged to be present by clinical examination in 45 of 53 patients with a positive VAS score (85%), including the 1 patient with isolated retropatellar disease and a VAS score of 13 mm. The 8 patients with a positive VAS score and a (false)-negative clinical assessment for synovitis had VAS scores between 5 mm and 47 mm (mean 21). Synovitis was not detected by clinical examination in 2 of the 7 patients (29%) with a VAS score of 0. The positive and negative predictive values of clinical examination compared with arthroscopy VAS for overall synovitis were 90% and 20%, respectively, resulting in an accuracy of 78% (Table 1).


The kappa value for intraobserver reproducibility for arthroscopy was 0.88. The kappa values using the dichotomous scale for the detection of synovitis with US were 0.71 and 0.85 for inter- and intraobserver reproducibility, respectively (P < 0.05 for both). The weighted kappa values for agreement of the semiquantitative US grading system were 0.65 and 0.74 for inter- and intraobserver reproducibility, respectively.


This is the largest study undertaken to date to validate US-detected knee synovitis in patients with a spectrum of disease and synovitis severity. We used arthroscopy as the gold standard, while also assessing its reproducibility and comparing US with clinical examination. The results suggest that US can detect and accurately localize synovitis in the knee, and is more sensitive than clinical examination. This is the first study to document the reproducibility of US-detected synovitis in the knee, and the results suggest that US is a technique with high inter- and intraobserver reproducibility.

Compared with arthroscopy, US of the knee has the advantage of being noninvasive and enabling visualization of extracapsular features such as the collateral ligaments, Baker's cysts, and quadriceps and patellar tendons. We have demonstrated a high positive predictive value for detection of synovitis by US, suggesting that a US examination of the knee may well eliminate the need for further, often more expensive and invasive, investigations. However, arthroscopy still has an important role, particularly in differentiating between the inflammatory arthritides, performing tissue sampling under direct visualization, and assessing the retropatellar space or the cruciate ligaments. The fact that only 1 knee had isolated retropatellar granular hypertrophy (with a VAS for overall synovitis of 13 mm) suggests that in this group of patients there was little additional information regarding synovitis in the retropatellar space. That this space cannot be fully visualized with US may be less important for the detection of synovitis.

With US, synovitis was detected in 7 of the 8 patients (88%) in whom clinical assessment failed to detect synovitis that was seen arthroscopically. The 1 patient with a false-negative US result had a very low arthroscopy synovitis score (VAS 7 mm), compared with an average VAS score of 21 mm for the patients with false-negative results on clinical examination. This is consistent with previous reports suggesting that low-grade synovitis cannot be detected by clinical examination (5). The clinical relevance of this lower-grade synovitis may be particularly important in OA, where the degree of synovitis may be less, but still an important cause of symptoms (3) and predictor of damage (4). The findings of this study also suggest that US is more sensitive at detecting villous hypertrophy than simple granularity of the synovial membrane, which is to be expected given that granularity represents an early, more subtle process with less synovial thickening compared with villous hypertrophy.

A small amount of synovial fluid may be seen in the normal knee joint with US and was present in all patients with arthroscopically evident synovitis, including the 1 patient with a false-negative US result and the other patient with isolated retropatellar synovitis, suggesting that increased vigilance for synovitis is warranted if synovial fluid is detected. Synovial fluid was also detected by US in 3 patients without synovitis on arthroscopy, but in only 1 of these 3 was synovitis incorrectly diagnosed with US and clinical examination, suggesting that synovial fluid has no major impact on the false-positive diagnosis of synovitis using these techniques.

This study demonstrated very good inter- and intraobserver reproducibility of US in the detection of the presence or absence of synovitis in the knee, and intraobserver reproducibility for grading synovitis. This suggests that US can be used with confidence in longitudinal studies by either individual or multiple readers for determining the presence or absence of synovitis, and by an individual reader for determining the grade of synovitis. The level of interobserver agreement using the semiquantitative grading scale, although good, suggests that in studies involving multiple readers, other methods, such as measurement of synovial thickness, may need to be considered.

There are some limitations to this study. The use of macroscopic findings on arthroscopy as the gold standard could be questioned due to the dearth of studies comparing this evaluation with histologic changes (8, 9, 18, 19). Earlier studies (18, 19) had contrasting outcomes, with one demonstrating poor (18) and the other demonstrating good (19) correlation, although neither took into account the variation of histologic findings that can be seen within a single joint (8, 9, 20–22). Studies addressing this variation, using an assessment technique similar to that in the present study, demonstrated good correlation between site-specific macroscopic findings and histologic analysis (8, 9). Our group has also demonstrated good correlation between global arthroscopy VAS for synovitis and immunohistologic results (CD4 levels) (23), as well as synovial fluid matrix metalloproteinase 1 (MMP-1), MMP-3, and CD3 levels (24). Furthermore, the current study also confirms the excellent intraobserver reproducibility with arthroscopy.

Finally, the clinical applicability of any gold standard is related to its pathogenic relevance, without which its prognostic capacity is diminished (25). Currently, there is no firm evidence that correlates histologic findings with clinical outcomes, in particular, joint damage. Also there appears to be only poor correlation between histologic findings and levels of C-reactive protein (26), an often-used surrogate for disease activity that cumulatively correlates well with disease progression (27, 28). In contrast, there is a growing body of evidence that synovial volume correlates with subsequent development of joint erosions (29, 30). Arthroscopy allows direct visualization of the synovium, and current scoring methods offer a semiquantitative assessment of severity, and therefore, disease amount. Although there is a paucity of studies demonstrating a correlation between arthroscopically evident synovitis and damage (4), it appears that one can conclude this by association.

Due to the difficulty in the anatomic delineation of the 3 compartments, there may well be some overlap of compartments between those visualized by US and those visualized by arthroscopy. For example, if a patient had synovitis extending medially and laterally from the SPP, the arthroscopist may have identified this as being in the SPP only, whereas the ultrasonographer may have judged this as extending into both the medial and lateral compartments, hence scoring 2 false-positives. The better results for patients as a whole compared with individual compartments (Table 1) may reflect this.

There was a selection bias in the present study in that all patients were referred for arthroscopy because of knee pain, and it is therefore more likely that they would have been thought to have synovitis on clinical examination. Furthermore, the relatively low number of joints without arthroscopically evident synovitis (n = 8) makes it difficult to extrapolate the specificity of US, although it is clear that, even in this group in which the majority of patients had synovitis, clinical examination is relatively inaccurate.

Clinical examination assessed the knee as a whole and did not differentiate between the retropatellar space and the other compartments, but only 1 patient had isolated retropatellar disease, and US remained more sensitive than clinical examination when compared with arthroscopy VAS for overall synovitis, which would have included this space. In this study the definition of clinical synovitis deliberately did not include joint tenderness (periarticular tenderness in the knee is common, especially in degenerative and entheseal disease). Any increase in sensitivity as a result of including tenderness may have also been accompanied by an increase in false-positive diagnosis, resulting in an even lower specificity and negative predictive value. Newer techniques such as power Doppler or contrast-enhanced Doppler may improve the accuracy of US by differentiating between hypervascular (active) and fibrotic (inactive) synovial thickening or pannus.

This study demonstrates that US is a valid and reproducible tool for the detection of synovitis in the knee joint and confirms previous reports that it is superior to clinical examination. Improved rates of detection of synovitis with US, and the role of synovitis in the symptoms and progression of knee arthritis, suggest that this technique should be used increasingly in order to improve disease diagnosis and management (1, 4). This has important implications with regard to the future practice and training of rheumatologists. Further studies are needed to validate more recent US techniques, the ability of US to detect changes following therapy, and the cost-effectiveness of US compared with other assessment techniques.