Reliability of different Doppler ultrasound quantification methods and devices in the assessment of therapeutic response in arthritis

Transcription

1 Rheumatology 2008;47: Advance Access publication 2 August 2008 doi: /rheumatology/ken318 Reliability of different Doppler ultrasound quantification methods and devices in the assessment of therapeutic response in arthritis K. Albrecht 1, K. Grob 1, U. Lange 1,U.Müller-Ladner 1 and J. Strunk 1 Objective. To evaluate the metric qualities of power Doppler ultrasound (PDUS) for different quantification methods and devices in the assessment of arthritis during anti-inflammatory treatment Methods. Twenty-four patients with active arthritis and first-time treatment with the TNF- inhibitor adalimumab underwent sequential clinical, laboratory and US examination at baseline, weeks 2, 6 and 12. 2D and 3D PDUS was performed by two independent investigators. The images and movies were scored from 0 to 3 and the amount of colour pixels and voxels was calculated. In addition, the resistance index of a synovial artery was measured. Thirteen patients were examined with a second US machine. Results. Treatment response was already observed at week 2 with a significant reduction of 2D (P < 0.01) and 3D scores (P < 0.001). A moderate correlation to 28-joint disease activity score was found for 3D voxel count (r s ¼ 0.35, P < 0.001). Interobserver agreement was or ICC 0.8 for all methods except the resistance index (ICC ¼ 0.60). Intermachine agreement was ¼ 0.57 for 2D PDUS score. Conclusions. The study demonstrates good to excellent interobserver and moderate intermachine reliability of different PDUS assessment methods in a longitudinal open-label study. KEY WORDS: Power Doppler, Ultrasonography, 3D, Arthritis, Wrist. Introduction Power Doppler ultrasound (PDUS) is used increasingly to monitor the efficacy of a current treatment strategy. Changes in synovial perfusion in arthritic joints are predominantly evaluated using semi-quantitative scoring systems in which the intensity of the synovial blood flow is graded in a four-step scale [1 5]. Other groups perform a quantitative analysis, using the measurement of the resistance index or computer-aided calculations of the pixel count [6 11]. Recently, 3D PDUS movies have also been used for the assessment of a therapeutic response [12 14]. The OMERACT Ultrasound Special Interest Group has focused its work on defining a standardized consensus on acquisition, reading and interpretation of US imaging data in order to fulfil the components of the OMERACT filter [15, 16]. After a definition of standardized scans for each joint [17] and defining pathological lesions in patients with inflammatory arthritis [18], a systematic review of the US assessment of synovitis was performed using the literature published between 1966 and 2005 [19]. The collected information on the metric quality of US for the detection of synovitis revealed major shortcomings in the reliability, validity and responsiveness testing of US [20]. Studies, initiated through this Interest Group and other independent groups have also evaluated metrics of reliability in the meantime, most of them focusing on intrareader and interobserver reliability of binominal data (presence or absence of PDUS signals) and the semi-quantitative scoring method [21 23]. All of them confirmed earlier reliability data [2, 24]. However, with respect to quantitative assessment methods of PDUS findings, reliability data are rare [14, 25] and data concerning intermachine agreement is only published in one study from Koski et al. [26] showing differences in the smallest detectable flow in three US devices. Addressing this problem, the varying methods of PDUS assessment of synovial vascularity have been examined, presented together with a review of the available reliability data [27]. Subsequently, a prospective study was performed in order to compare these methods under the same preconditions and to investigate responsiveness, construct validity and reliability data for each of these methods. Patients and methods Patients Twenty-four consecutive arthritis patients, who received an initial treatment with the TNF- inhibitor adalimumab (40 mg subcutaneously every other week) at the Department of Rheumatology of the University of Giessen (Kerckhoff Clinic, Bad Nauheim, Germany) were included into the study. Treatment decision was independent from the study and followed the recommendation criteria of active arthritis despite prior treatment of two DMARDs including MTX. Further inclusion criteria were a clinically active arthritis as determined by soft-tissue swelling and tenderness in at least one peripheral joint, diagnosis of RA according to the ACR 1987 revised criteria or diagnosis of PsA according to Classification criteria for the diagnosis of Psoriatic Arthritis (CASPAR) criteria and age 18 yrs. Allowed concomitant drugs included stable doses of NSAIDs and steroids. The dosage remained unchanged during follow-up visits or was reduced if possible. Patients, who had prior IA steroid, radionuclide injection in the examined joint or operative intervention in the preceding 6 months, were excluded from the study. Clinical and US examination were performed in accordance with the Declaration of Helsinki. The study protocol was approved by the local ethics committee. A written informed consent was obtained from each patient. Study protocol Patients underwent clinical, laboratory and US examination during four sequential visits at baseline (visit 1), week 2 (visit 2), week 6 (visit 3) and week 12 (visit 4). Clinical and laboratory examination included the 28-joint disease activity score (DAS 28), the HAQ, a visual analogue scale (VAS: mm) for the patient s pain level, the measurement of the ESR by Westergren method and the serum level of CRP. To provide homogeneous data, the DAS 28 score was calculated for every patient including PsA patients. 1 Department of Rheumatology and Immunology, University of Giessen, Kerckhoff Klinik, Bad Nauheim, Germany. Submitted 11 April 2008; revised version accepted 7 July Correspondence to: K. Albrecht, Department of Rheumatology and Immunology, University of Giessen, Kerckhoff Klinik, Benekestr.2-8, Bad Nauheim, Germany. k.albrecht@kerckhoff-klinik.de 1521 ß The Author Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 1522 K. Albrecht et al. US imaging protocol At baseline, every patient was screened by US at bilateral wrists, MCP, PIP and MTP joints II V. Then the dominant joint that showed the most intensive Doppler activity was selected as a target joint for the follow-up examinations (Table 1). US examination was performed by one unexperienced (K.G., 15 US examinations including 3D PDUS prior to the study) and one experienced investigator (K.A., 3 yrs of clinical and scientific US), blinded to clinical and laboratory findings. Both investigators had been trained from the head ultrasonographer of the department (J.S.). US was performed with a linear array transducer with variable frequency from 5 to 12 MHz (L12-5/38 mm, HDI 5000, ATL/Philips, Bothell, WA, USA). Standardized scans according to the Guidelines for musculoskeletal ultrasound in rheumatology were performed [17]. Then the dorsal longitudinal scan of each joint was selected for scoring, measurement and for 3D movies to avoid large variation due to different acquisition of images. All joints were examined in extended position with a comfortable resting of the scanned extremity and of the scanning arm of the examiner to ensure as little pressure as possible from the transducer. Both investigators obtained and evaluated their own images independent of each other but agreed at baseline in the selection of the target joint. Synovial tissue and effusion Synovial tissue and effusion were not differentiated as defined by the OMERACT definitions for US pathology [18] because in our experience this discrimination shows a high grade of variation. It was measured in millimetres (mm) at the thickest point and graded on a semi-quantitative scale from 0 ¼ normal or no synovial thickening/effusion, 1 ¼ mild thickening/effusion, 2 ¼ moderate thickening/effusion to 3 ¼ intense thickening/effusion. To facilitate readability, only the word effusion is used in the Results section. 2D PDUS Power Doppler settings were standardized with a pulse repetition frequency (PRF) of 700 Hz. The gain was set to a CPU (colour per unit) of 79%, which was selected by manual elevation of the power Doppler US gain level until the colour box was almost uniformly filled with the first indication of colour and with only the minimum amount of the next highest signal just beginning to appear [28]. Representative 2D PDUS images with the highest amount of Doppler signals during the diastole were obtained. Doppler signals were estimated according to the 4-grade semiquantitative scoring system: 0 ¼ no Doppler signal/no blood flow; 1 ¼ single Doppler signals/mild blood flow, 2 ¼ various, confluent Doppler signals/moderate blood flow and 3 ¼ confluent Doppler signals with more than half of the visible synovium showing Doppler signals/intense blood flow. TABLE 1. Demographic data of patient collective Number of patients 24 Age, mean þ SD, yrs Sex (m/f) 7/17 Diagnosis (RA/PsA) 17/7 Disease duration, mean þ SD, yrs Target joints Carpus, n (%) 14 (58.3) MCP, n (%) 5 (20.8) PIP, n (%) 3 (12.5) MTP, n (%) 2 (8.3) Adalimumab monotherapy 7 (29.2) Combination therapy with MTX, n (%) 13 (54.2) LEF, n (%) 4 (16.6) Daily oral steroid use (prednisolone equivalent) <10 mg, n (%) 20 (83.3) <20 mg, n (%) 4 (16.6) NSAID use 18 (75%) Pixel count The digitally stored 2D PDUS images were transferred to a pixel counter program that was developed for the study [29]. A region of interest was defined with a fixed frame setting. Then the number of coloured pixels within the frame was calculated. Resistance index The spectral Doppler sampling area was placed over one intrasynovial artery and a spectral flow curve was obtained. The US unit identified the cardiac cycles as well as peak systolic and end diastolic flow, which was used to automatically calculate the resistance index [RI ¼ (peak systolic flow end diastolic flow)/ peak systolic flow]. The value of the RI ranges between 0 and 1. If there was no IA signal, the RI was defined as 1. 3D PDUS In the same scanning position, the transducer was mechanically moved in one direction from medial to lateral over the dorsal side of the joint (free-hand sweep) to obtain a sequence of 15 2D PDUS images [30]. The online 3D power Doppler function provided by the HDI 5000 vascular software was used to generate a 3D image of the peri-articular and IA blood vessels in which grey-scale information of the surrounding tissue was already subtracted. The acquired data were stored digitally on hard disc as cine loop. 3D movies were scored analogously to 2D PDUS images. Voxel count The 3D.avi files were transferred to Image J, a public domain image analysis program of the National Institutes of Health [31]. The connected voxel counter was used to calculate the coloured voxels [14]. Reliability data For all quantification parameters, both observers used their own 2D and 3D data sets to provide true interobserver agreement data. To provide intermachine reliability, 13 patients were examined subsequently at every visit from Investigator 1 with an Image Point (Hewlett Packard, Bad Homburg, Germany) and a 7.5 MHz linear array transducer. 2D PDUS images were obtained in the same scanning protocol as described for the HDI A cine loop of the examination was recorded by video tape. Afterwards, representative 2D images were selected and saved as.jpg files on hard disc. The performance of a single US examination in a given patient including 2D and 3D PDUS and the calculation of the pixel and voxel count required 25 min for each investigator and additionally 15 min for the second US investigation at the HP Image point. Statistical analysis Statistic analysis was performed with win STAT Õ excel version (R.Fitch Software, Bad Krozingen, Germany). Normal distribution was tested with Kolmogorov Smirnov test for continuous and chi-square test for discrete variables. Kruskal Wallis test was used to examine differences within the patient sample due to diagnosis or joint region. The comparison of findings at baseline and follow-up was evaluated using Wilcoxon s rank sum test. For the correlation between US and clinical variables, Spearman s rank test was used for non-parametric values and Pearson s correlation was used for normal distributed metric values. Reliability agreement data from ordinal data was calculated by overall agreement (percentage of exact agreement) and with weighted Cohen s -statistics [32]. A -value of was considered poor, fair, moderate, good and excellent [33]. Agreement between parametric values was tested with intraclass correlation (ICC) analysis of variance, model two and single-measure reliability [34]. The

3 Reliability of Doppler ultrasound 1523 coefficient represents concordance, in which 1 is for perfect agreement and 0 is for no agreement at all. The - and ICCvalues were obtained using the online calculation software provided by the Department of Obstetrics and Gynaecology of the Chinese University of Hong Kong [35]. Intermethod agreement was tested with Spearman s rank correlation. The level of significance was regarded as P < Results Patient collective Demographic data of the patient collective is presented in Table 1. Follow-up visits were interrupted after 6 weeks in two patients due to treatment side-effects and worsening of arthritis. For the same reason, in two additional patients, treatment strategy was TABLE 2. Clinical and US parameters during follow-up visits changed at week 12 (visit 4). The data from the two patients with incomplete follow-up was analysed. Kruskal Wallis test showed no significant differences between RA and PsA patients as well as between the different examined joint regions for all clinical and US parameters (P > 0.05). Responsiveness A significant decrease at week 2 was found for the 2D score (P < 0.01), 3D score (P < 0.001), pixel and voxel count (P < 0.05). At weeks 6 and 12, all US parameters showed a significant decrease. Statistics only for the wrist joint added up to very similar results. Mean values of clinical and US variables in the course of the follow-up visits are presented in Table 2. US treatment response from two patients is displayed in Fig. 1. Baseline Week 2 Week 6 Week 12 Variables ESR (mm/1st h) CRP (mg/l) DAS HAQ VAS (0 100) US parameters (target joint) Effusion score Effusion (mm) D Score RI D Pixel count D Score D Voxel count US parameters (wrist joint) Effusion score Effusion (mm) D Score RI D Pixel count D Score D Voxel count Mean values S.D., P < 0.05, P < 0.01, P < vs baseline (Wilcoxon s rank sum test). FIG. 1. Illustration of treatment response. 3D PDUS images of the wrist joint of Patients 7 and 8 from Investigator 1. A continuous decrease of Doppler signals from visits 1 to 4 could be observed in Patient 7. In Patient 8, the amount of Doppler signals remained unchanged until visit 3 and deteriorated at visit 4 in accordance with a clinical relapse of the disease.

4 1524 K. Albrecht et al. Construct validity Correlation between clinical and US parameters is distributed in Table 3. A moderate correlation was found between voxel count and DAS 28 (r s ¼ 0.35, P < 0.001) as well as to HAQ (r s ¼ 0.30, P < 0.01) and CRP (r s ¼ 0.29, P < 0.01). Reliability Interobserver agreement. In 24 patients, a subset of 65 visits was performed by two investigators. Interobserver agreement was ¼ 0.83 for effusion, ¼ 0.87 for 2D and ¼ 0.86 for 3D scores. The level of exact agreement was 69% for effusion, 79% for 2D and 71% for 3D scores, respectively. ICC values were 0.85 for effusion (mm), 0.60 for RI measurement, 0.80 for pixel and 0.88 for voxel count. Figure 2 shows an example of 3D PDUS interobserver agreement. Intermachine agreement. Four follow-up visits of 13 patients were used for intermachine agreement data (n ¼ 52). The -values were ¼ 0.58 for effusion and ¼ 0.57 for 2D score. ICC-value was 0.77 for effusion (mm) and 0.31 for pixel count. Follow-up images of the two US machines are presented in Fig. 3. All reliability data are displayed in Table 4. Discussion Demonstration of therapeutic response by PDUS has been reported using the 2D score [1, 3, 4, 36], pixel count [6, 7, 9 11, 37], resistance index [6, 9 11], 3D score [12] and voxel count [14] with earliest significant changes after 7 days [12]. Response parameters in our study showed a delayed alteration in synovial tissue/effusion and RI measurement compared with all other PDUS parameters. Especially 3D scoring showed a very early significant response after 2 weeks. Nevertheless, it has to be pointed out that the responsiveness of the 3D score to treatment is not better than the best available clinical measure, the DAS 28 score. The majority of studies addressing PDUS assessment have demonstrated a weak-to-moderate correlation with clinical and laboratory markers of inflammation [19]. Our study results indicate that 3D PDUS may provide a slightly better correlation TABLE 3. Construct validity data CRP DAS 28 HAQ Effusion score Effusion (mm) D score Pixel count D score Voxel count Spearman s correlation coefficient, P < 0.05, P < 0.01, P < to clinical markers as 2D PDUS. However, with respect to nearly equal values for interobserver agreement and only marginal earlier assessment of therapeutic response, 3D imaging cannot justify its efforts for daily clinical routine at the moment. It might gain in importance once the equipment is improved [38]. Good-to-excellent interobserver agreement values for scores and computer-aided quantification methods confirm prior reports even with regard to the problem of testing acquisition data with an inexperienced investigator. Both investigators have been trained in the same department though, which explains in part good interobserver results. The difficulties in measuring the resistance index are underlined, resulting in lower interobserver agreement mainly due to acquisition difficulties of detecting an IA artery during follow-up visits in a reasonable amount of time. Referring to all of the metric qualities, the semi-quantitative scores perform comparably with the software-generated quantification methods. With regards to the validity of studies performing semi-quantitative scoring as outcome measures for clinical trials, this is very important and encouraging for ongoing studies as the scoring method is distinctively easier to apply and not as timeconsuming as any software-generated quantification method. The most interesting data could be obtained for intermachine agreement. Here, the results show only a moderate agreement for the scoring method, which might become a substantial obstacle for the realization of multicentre studies. On the other hand, the score can easily be applied to images from different US devices. In contrast to that, even with the same frame settings, the quantitative amount of Doppler signals cannot be compared between different US devices, which is indicated by the fair intermachine reliability of the pixel count. As a high-end US machine was tested against an older device, agreement should improve when using high-end equipment in all centres though. The HP Image Point has a lower specification than the Philips HDI 5000 and employs a lower frequency transducer so that a loss of sensitivity for the detection of Doppler signals was not surprising and confirms the results from Koski et al. [26]. An external standardization of supplies by technical flow models could resolve the problem or at least reduce differences due to physical properties of the devices. The decision for follow-up of only one target joint was the use of 3D technique that is more time-consuming than 2D PDUS. According to our experiences, 3D movies display more precise results if we concentrated on one joint. As the 3D target joint follow-up has not been compared with multiple joint follow-up yet, we cannot exclude a selection bias due to spontaneous variation. Although a study of Scheel et al. [39] proposed the assessment of finger joints from the palmar view, we decided in general to use the dorsal view of all joints. On the basis of our experiences, Doppler signals are more frequently seen from the dorsal side of the joints but the study from Scheel et al. only investigated the assessment of synovitis/effusion but not Doppler signals. The recently introduced simplified ultrasonographic Power Doppler FIG. 2. Demonstration of interobserver agreement. 3D PDUS images of Patient 5 from Investigators 1 and 2 at visit 3 (week 6). The images are obtained from a free-hand sweep over the wrist joint in a dorsal longitudinal radio carpal scan showing good agreement in 3D score and voxel count.

5 Reliability of Doppler ultrasound 1525 FIG. 3. Intermachine agreement. Follow-up 2D PDUS images of Patient 19 from the two US devices Philips HDI 5000 and HP Image Point. Arrows indicate a dorsal longitudinal scan of the MTP II joint with present Doppler signals. TABLE 4. Interobserver and Intermachine agreement data Percentage of perfect Agreement ICC Interobserver agreement (n ¼ 65) Effusion score Effusion (mm) D score pixel count 0.80 resistance index D score voxel count 0.88 Intermachine agreement (n ¼ 52) Effusion score Effusion (mm) D score pixel count values and percentage of observed perfect agreement for ordinal, ICC for interval parameters with no significant difference between raters (P > 0.05). assessment from Naredo et al. [40] also contains the dorsal side of wrist and MCP joints. Summarizing the characteristics of the different PDUS assessment methods, the 2D score has the advantages of easy appliance and evaluation as well as a widespread availability. Within a short examination time, outcome qualities are almost similar to 3D PDUS and do not show significant limitations compared with quantitative analyses. With respect to the performance of multicentre studies, intermachine reliability data elucidate the importance of using equal US devices in order to minimize differences in the sensitivity of Doppler signal acquisition. Rheumatology key messages PDUS scores provide good interobserver and moderate intermachine agreement in the assessment of arthritis. Quantification and 3D imaging showed no significant improvement of reliability. Acknowledgements Funding: The study was funded in part by the Pitzer Foundation, Bad Nauheim and by Abbott Immunologies, Wiesbaden, Germany. Disclosure statement: The authors have declared no conflicts of interest. References 1 Fiocco U, Ferro F, Vezzu M et al. Rheumatoid and psoriatic knee synovitis: clinical, grey scale, and power Doppler ultrasound assessment of the response to etanercept. Ann Rheum Dis 2005;64: Filippucci E, Farina A, Carotti M, Salaffi F, Grassi W. Grey scale and power Doppler sonographic changes induced by intra-articular steroid injection treatment. Ann Rheum Dis 2004;63: Filippucci E, Iagnocco A, Salaffi F, Cerioni A, Valesini G, Grassi W. Power Doppler sonography monitoring of synovial perfusion at the wrist joints in patients with rheumatoid arthritis treated with adalimumab. Ann Rheum Dis 2006;65: Ribbens C, Andre B, Marcelis S et al. Rheumatoid hand joint synovitis: gray-scale and Power Doppler US quantifications following anti-tumor necrosis factor-alpha treatment: pilot study. Radiology 2003;229: Salaffi F, Carotti M, Manganelli P, Filippucci E, Giuseppetti GM, Grassi W. Contrastenhanced power Doppler sonography of knee synovitis in rheumatoid arthritis: assessment of therapeutic response. Clin Exp Rheumatol 2004;23: Takahashi A, Sato A, Yamadera Y et al. Doppler sonographic evaluation of effect of treatment with infliximab (Remicade) for rheumatoid arthritis. Mod Rheumatol 2005;15: Taylor PC, Steuer A, Gruber J et al. Comparison of ultrasonographic assessment of synovitis and joint vascularity with radiographic evaluation in randomized, placebo controlled study of infliximab therapy in early rheumatoid arthritis. Arthritis Rheum 2004;50: Teh J, Stevens K, Williamson L, Leung J, McNally EG. Power Doppler ultrasound of rheumatoid synovitis: quantification of therapeutic response. Br J Radiol 2003;76: Terslev L, Torp-Pedersen S, Qvistgaard E, Danneskiold-Samsøe B, Bliddal H. Estimation of inflammation by Doppler ultrasound: quantitative changes after intraarticular treatment in rheumatoid arthritis. Ann Rheum Dis 2003;62: Terslev L, Torp-Pedersen S, Qvistgaard E et al. Effects of treatment with etanercept (Enbrel, TNRF:Fc) on rheumatoid arthritis evaluated by Doppler ultrasonography. Ann Rheum Dis 2003;62: Shio K, Homma F, Kanno Y et al. Doppler sonographic comparative study on usefulness of synovial vascularity between knee and metacarpophalangeal joints for evaluation of articular inflammation in patients with rheumatoid arthritis treated by infliximab. Mod Rheumatol 2006;16:

6 1526 K. Albrecht et al. 12 Strunk J, Strube K, Müller-Ladner U, Lange U. Three-dimensional power Doppler ultrasonography confirms early reduction of synovial perfusion after intra-articular steroid injection. Ann Rheum Dis 2006;65: Strunk J, Strube K, Klingenberger P, Müller-Ladner U, Lange U. Two- and threedimensional Doppler sonographic evaluation of the effect of local cryotherapy on synovial perfusion in wrist arthritis. Rheumatology 2006;45: Strunk J, Strube K, Rumbaur C, Lange U, Müller-Ladner U. Interobserver agreement in two- and three-dimensional power Doppler Sonographic assessment of synovial vascularity during anti-inflammatory treatment in patients with rheumatoid arthritis. Ultraschall in Med 2007;28: Boers M, Brooks P, Strand CV, Tugwell P. The OMERACT filter for outcome measures in rheumatology. J Rheumatol 1998;25: Wakefield RJ, D Agostino M-A, Iagnocco AM et al. The OMERACT Ultrasound Group: status of current activities and research directions. J Rheumatol 2007;34: Backhaus M, Burmester GR, Gerber T et al. Guidelines for musculoskeletal ultrasound in rheumatology. Ann Rheum Dis 2001;60: Wakefield RJ, Balint PV, Szkudlarek M et al. Musculoskeletal ultrasound including definitions for ultrasonographic pathology. J Rheumatol 2005;32: Joshua F, Edmonds J, Lassere M. Power Doppler ultrasound in muskuloskeletal disease: a systematic review. Semin Arthritis Rheum 2006;36: Joshua F, Lassere M, Bruyn A et al. Summary findings of a systematic review of the ultrasound assessment of synovitis. J Rheumatol 2007;34: Scheel AK, Schmidt WA, Hermann KG et al. Interobserver reliability of rheumatologists performing musculoskeletal ultrasonography: results from a EULAR Train the trainers course. Ann Rheum Dis 2005;64: Naredo E, Moller I, Moragues C et al. Interobserver reliability in musculoskeletal ultrasonography: results from Teachers rheumatologist course. Ann Rheum Dis 2006;65: Koski JM, Saarakkala S, Helle M et al. Assessing the intra- and interreader reliability of dynamic ultrasound images in power Doppler ultrasonography. Ann Rheum Dis 2006;65: Szkudlarek M, Court-Payen M, Jacobsen S et al. Interobserver agreement in ultrasonography of the finger and toe joints in rheumatoid arthritis. Arthritis Rheum 2003;48: Qvistgaard E, Rogind H, Torp-Pedersen S, Terslev L, Danneskiold-Samsøe B, Bliddal H. Quantitative ultrasonography in rheumatoid arthritis: evaluation of inflammation by Doppler technique. Ann Rheum Dis 2001;60: Koski JM, Saarakkala S, Helle M, Hakulinen U, Heikkinen JO, Hermunen H. Power Doppler ultrasonography and synovitis: correlating ultrasound imaging with histopathological findings and evaluating the performance of ultrasound equipments. Ann Rheum Dis 2006;65: Albrecht K, Müller-Ladner U, Strunk J. Quantification of the synovial perfusion in rheumatoid arthritis using Doppler ultrasonography. Clin Exp Rheumatol 2007;25: Rubin JM, Adler RS, Fowlkes JB et al. Fractional moving blood volume: estimation with power Doppler US. Radiology 1995;197: Bals M. Ultraschall Pixel Counter CX (19 Jan 2008, date last accessed). 30 Strunk J, Lange U. Three-dimensional power Doppler sonographic visualization of synovial angiogenesis in rheumatoid arthritis. J Rheumatol 2004;31: Rasband WS. Image J. Bethesda, MA, USA: National Institutes of Health, (17 April 2008, date last accessed). 32 Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas 1960;20: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33: Portney LG, Watkins MP. Foundations of clinical research. Applications and practice. Norwalk, Connecticut: Appleton & Lange, 1993: Chang A. Stat tools site. Department of Obstetrics and Gynaecology of the Chinese University of Hong Kong. (2008, date last accessed). 36 Naredo E, Collado P, Cruz A et al. Longitudinal power Doppler ultrasonographic assessment of joint inflammatory activity in early rheumatoid arthritis: predictive value in disease activity and radiologic progression. Arthritis Rheum 2007;57: Hau M, Kneitz C, Tony HP, Keberle M, Jahns R, Jenett M. High resolution ultrasound detects a decrease in pannus vascularisation of small finger joints in patients with rheumatoid arthritis receiving treatment with soluble tumor necrosis factor alpha receptor (etanercept). Ann Rheum Dis 2002;61: Filippucci E, Iagnocco A, Meenagh G et al. Ultrasound imaging for the rheumatologist. VII. Ultrasound imaging in rheumatoid arthritis. Clin Exp Rheum 2007;25: Scheel AK, Herman KG, Kahler E et al. A novel ultrasonographic synovitis scoring system suitable for analysing finger joint inflammation in rheumatoid arthritis. Arthritis Rheum 2005;52: Naredo E, Rodriguez M, Campos C et al. Validity, reproducibility and responsiveness of a twelve-joint simplified power Doppler ultrasonographic assessment of joint inflammation in rheumatoid arthritis. Arthritis Rheum 2008;59: