The smallest detectable difference and sensitivity to change of magnetic resonance imaging and radiographic scoring of structural joint damage in rheumatoid arthritis finger, wrist, and toe joints: A comparison of the omeract rheumatoid arthritis magnetic resonance imaging score applied to different joint combinations and the sharp/van der heijde radiographic score

Authors


Abstract

Objective

To compare 2 magnetic resonance imaging (MRI) approaches and radiographic evaluation according to the Sharp/van der Heijde method with respect to sensitivity to change in joint destruction in patients with rheumatoid arthritis (RA).

Methods

Thirty-five RA patients and 9 healthy controls underwent MRI and radiography on 2 occasions 1 year apart. Conventional radiographs of the hands, wrists, and forefeet were evaluated according to the Sharp/van der Heijde method. MRIs of unilateral wrist and second through fifth metacarpophalangeal (MCP) joints (“few-joints approach”) and of bilateral wrist and MCP joints plus unilateral metatarsophalangeal (MTP) joints (“many-joints approach”) were assessed for bone erosions according to the scoring system recommended by the OMERACT (Outcome Measures in Rheumatology Clinical Trials) group. The smallest detectable differences (SDDs) of the radiography and MRI scores were computed based on reevaluation of one-third of the study population.

Results

Progressive joint destruction, i.e., an increase in score after the followup period, was observed more frequently with the MRI “many-joints approach” (30 subjects) and “few-joints approach” (25 subjects) than with the Sharp/van der Heijde radiographic method (9 subjects) (P < 0.001 by chi-square analysis). No significant difference between the MRI approaches was observed. When only subjects with a change greater than the SDD were considered, progression was revealed with the MRI “many-joints approach,” the MRI “few-joints approach,” and radiography in 15, 13, and 5 RA subjects, respectively. With both MRI approaches, significantly more subjects with progression were detected than were detected by radiography (P < 0.05).

Conclusion

MRI, regardless of whether it covers unilateral wrist and MCP joints or bilateral wrist and MCP joints plus unilateral MTP joints, is significantly superior to radiography of the hands, wrists, and forefeet with respect to detection of progressive joint destruction in RA.

Evaluation of structural joint damage on repeated conventional radiography is a central outcome measure in rheumatoid arthritis (RA), in clinical trials as well as routine clinical management (1, 2). In clinical trials, the most widely used radiographic scoring systems are the Larsen and the Sharp methods and modifications of these (3–7). Due to the superior ability of the van der Heijde modification of the Sharp method (6) to detect change over time, this method is most often preferred in clinical trials (8).

Magnetic resonance imaging (MRI) may challenge conventional radiography as the most competent imaging modality for detection of temporal changes in structural joint damage in RA. In several studies of early RA, more bone erosions have been detected by MRI than by radiography (9, 10), and a higher frequency of erosive progression in individual joints has been observed (11–13). All MRI studies of RA patients have focused on 1 or 2 joint regions—most frequently the wrist and/or the second through fifth metacarpophalangeal (MCP) joints. Consequently, MRI may reduce the necessary sample size and followup time in RA trials, due to greater sensitivity in distinguishing responders from nonresponders (14).

However, there is no definite evidence that MRI is more sensitive than radiography for the detection of temporal variations in destructive joint changes in RA, as illustrated in the following scenarios: 1) radiography normally is performed on both hands, all fingers, and both forefeet (7) and, due to the greater anatomic coverage, may be more sensitive to change than MRI of the number of joints that it is feasible to image in 1 session (e.g., 1 wrist and/or the second through fifth MCP joints); and 2) the reproducibility of radiographic readings may be higher than that of MRI readings, potentially making radiography more sensitive to change than MRI when only changes beyond measurement error (the smallest detectable difference [SDD]) are considered.

These issues prompted us to investigate whether MRI of unilateral wrist and MCP joints (hereinafter designated the “few-joints approach”) (Figure 1) and/or of bilateral wrists and MCP joints and unilateral metatarsophalangeal (MTP) joints (the maximal number of joints feasible to image in 1 MRI session) (hereinafter designated the “many-joints approach”) is more sensitive in detecting changes in joint destruction than assessment by the optimal clinically used radiographic approach (the Sharp/van der Heijde method) (6) when only changes beyond the SDD are considered. In addition, we explored which joint areas or combinations of joint areas most frequently display detectable temporal changes in joint destruction as evidenced by MRI, and thereby may be the most advantageous to image in clinical studies of RA patients.

Figure 1.

Illustration of the joints that were assessed by the 2 magnetic resonance imaging (MRI) examination approaches (“few-joints approach” and “many-joints approach”) and by the Sharp/van der Heijde radiographic method.

PATIENTS AND METHODS

Subjects

Thirty-five patients with RA according to the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 revised criteria (15) and 9 healthy subjects were enrolled in a 12-month observational study. All patients were recruited from the Department of Rheumatology, Copenhagen University Hospital at Hvidovre. Clinical, biochemical, radiographic, and MRI data were available on all subjects at baseline and after 1 year of followup. Local ethics committee approval was obtained before the study protocol was initiated.

Clinical and biochemical examination

All clinical examinations were performed by the same rheumatologist and included assessment for joint swelling and tenderness as recommended by the European League Against Rheumatism (16). Composite Disease Activity Scores based on assessment of 28 joints (DAS28) (17) were computed. Laboratory studies included measurement of serum C-reactive protein and IgM rheumatoid factor levels.

MRI protocol

MRI of the wrists and second through fifth MCP joints bilaterally and the first through fifth MTP joints unilaterally was performed. Two major scanning approaches were compared, i.e., the “few-joints approach” (MRI of unilateral wrist and second through fifth MCP joints) and the “many-joints approach” (MRI of bilateral wrists and second through fifth MCP joints and unilateral first through fifth MTP joints). However, other joint areas and combinations were also explored (Table 1).

Table 1. Numbers of subjects with increased and decreased erosion scores, by MRI and radiography, from baseline to 1-year followup*
Joints/joint regions examinedBaseline erosion scores, mean (IQR)Erosive disease
No. of “progressors”No. of “regressors”SDDNo. of “definite progressors” (change greater than the SDD)No. of “definite regressors” (change greater than the SDD)
  • *

    MRI = magnetic resonance imaging; IQR = interquartile range; SDD = smallest detectable difference; D = dominant; non-D = nondominant; MCPs = metacarpophalangeal joints; MTPs = metatarsophalangeal joints.

  • “Few-joints approach” and “many-joints approach” (see text).

  • P < 0.001 versus the number detected by radiography.

  • §

    P < 0.05 versus the number detected by radiography.

MRI      
 D wrist and second through fifth MCPs17.7 (1–12.3)2502.1313§0
 D second through fifth MCPs5.2 (0–3.3)1401.2470
 D wrist12.5 (1–9.5)19§02.1380
 D first through fifth MTPs5.6 (0–4)1421.3642
 D and non-D wrists25.6 (1.75–2)2703.5213§0
 D and non-D MCPs9.5 (0–6.3)1701.9190
 D and non-D wrist + MCPs35.1 (3.3–25)2903.8316§0
 D wrist and D MTPs18.0 (1–13.3)2422.1812§1
 D and non-D wrist + D MTPs31.2 (1.75–25.3)2813.6913§0
 D MCPs and D MTPs10.8 (0.8–7)19§22.1350
 D and non-D MCPs + D MTPs15.0 (1–9.3)20§22.6890
 D and non-D wrist + MCPs + D MTPs40.6 (3.3–29.3)3014.1915§0
Radiography (Sharp/van der Heijde)35.8 (0–28.3)946.2350

MRI was performed with a 0.2T dedicated-extremity MRI unit (Artoscan; Esaote Biomedica, Genoa, Italy) equipped with a dual phased array wrist coil for wrist and MCP joint examinations and a dual phased array knee coil for MTP joint examinations. All MRI examinations were carried out using a T1-weighted 3-dimensional gradient echo sequence with subsequent multiplanar reconstruction. The scanning parameters were as follows: echo time 12 msec, repetition time 30 msec, field of view 140 × 140 × 80 mm, matrix 192 × 160 × 80, flip angle 65°, interslice gap 0 mm. Multiplanar reconstruction was done with a slice thickness of 1 mm and an in-plane resolution of 0.55 × 0.55 mm. The scanning time was 6 minutes.

MRI evaluation

All MR image sets were evaluated for bone erosions by the same observer, using the OMERACT (Outcome Measures in Rheumatology Clinical Trials) MRI scoring system (RAMRIS) (18). Images were evaluated in pairs but were masked with regard to chronology, and the assessor was unaware of the result of the radiographic readings. The RAMRIS entails separate assessment of each metacarpal/metatarsal head and the phalangeal base of the bones of each second through fifth MCP joint, first through fifth MTP joint, and wrist joint, for erosions. Bone erosions are scored on a scale of 0–10, based on the proportion of eroded bone compared with the assessed bone volume judged on all available images, as follows: 0 = no erosion; 1 = 1–10% of bone eroded; 2 = 11–20% of bone eroded, etc. The score for 1 wrist can range from 0 to a maximum of 150; the corresponding range is 0–80 for the second through fifth MCP joints and 0–100 for the first through fifth MTP joints. Sixteen subjects (approximately one-third of the study population) were reevaluated in order to provide data for calculating the SDD.

Conventional radiography

Radiographs of the wrists, hands, and forefeet were obtained using a posteroanterior projection and evaluated by an experienced radiologist (AV) for joint space narrowing and bone erosions, according to the method described by van der Heijde (8). The sets of radiographic images were evaluated in pairs but were masked with regard to chronology, and the radiologist was unaware of the results of the MRI assessments. The same subjects who underwent reevaluation by MRI were reevaluated by radiography as well.

Statistical analysis

The SDD (19, 20) is a statistical method for defining measurement error based on the 95% limits of agreement, as described by Bland and Altman (21).The SDD is reader and sample specific and represents the smallest change in score that can be discriminated from the measurement error of the scoring method. Using the SDD as the threshold level for a definite change in score ensures that the changes observed are not due to reading variability. We computed the SDD using the following formula:

equation image

where SD = standard deviation, obsa = score at first reading, and obsb = score at second reading. The numbers of patients with progressive joint destruction on MRI and radiography were compared using the chi-square test. P values less than 0.05 were considered significant.

RESULTS

The median age of the patients was 55 years (range 24–78), and the median disease duration was 5 years (range 0.2–27). Median clinical and laboratory values for measures of disease activity were as follows: number of swollen joints 5 (range 0–10), number of tender joints 7 (range 0–35), DAS28 4.2 (range 1.75–7.24), C-reactive protein level 8 mg/dl (range 8–103). Sixty-eight percent of the patients were positive for serum IgM rheumatoid factor. None of the control subjects had signs or symptoms of joint disease or IgM rheumatoid factor positivity.

The study allowed comparison of MRI findings in many different combinations of joint regions (Table 1). We focused primarily on the “few-joints approach” and the “many-joints approach” detailed above. At the end of the observation period, MRI by the “few-joints approach” revealed increased RAMRIS bone erosion scores in 25 subjects, while MRI by the “many-joints approach” revealed increased bone erosion scores in 30 patients and a decreased score in 1. Conventional radiography revealed increased Sharp/van der Heijde scores in 9 subjects and decreased scores in 4.

The SDDs with the Sharp/van der Heijde radiographic scoring method, the RAMRIS erosion score by the “few-joints approach,” and the RAMRIS erosion score by the “many-joints approach” were 6.2, 2.1, and 4.2, respectively (Table 1). When only subjects with a change in erosion score equal to or above the SDD were considered, radiography, MRI by the “few-joints approach,” and MRI by the “many-joints approach” revealed 5, 13, and 15 subjects with “definite” progressive joint disease, respectively, while no subjects exhibited a decreased score by any of the methods (Figure 2). In only 1 patient was “definite” deterioration found by radiography (based on increased joint space narrowing) but not by MRI with the “few-joints approach.” However, “definite” deterioration was revealed in this subject by MRI with the “many-joints approach.”

Figure 2.

Number of subjects with changed magnetic resonance imaging (MRI) and radiographic scores during the 1-year followup. P values were determined by chi-square test. NS = not significant; SDD = smallest detectable difference.

The number of subjects with progressive joint destruction seen on radiography was significantly lower than the numbers observed using the 2 MRI scanning approaches, irrespective of whether “raw” data (P < 0.001 by chi-square test) or SDD-adjusted data (P < 0.05 by chi-square test) were considered. The 2 MRI protocols did not differ significantly with respect to the numbers of subjects in whom erosive progression was detected.

Even though the present study focused on MRI using specific joint combinations (the “few-joints approach” and “many-joints approach”), other combinations were also evaluated. For example, 29 subjects were identified as having disease progression by MRI of the wrists and the second through fifth MCP joints bilaterally, in the “raw” data analysis. This number decreased to 16 after adjustment for the SDD (3.8). Further data on MRI findings obtained using different joint combinations are shown in Table 1.

No control subject had destructive joint changes seen on radiography. At baseline, low-grade erosion-like bone changes were seen on MRI in 2 controls by the “few-joints approach” and in 5 by the “many-joints approach.” No control subject exhibited a change in MRI scores above the SDD (“definite progressors”) within the observation period.

DISCUSSION

The present study compared the sensitivity to change in structural joint damage as determined using the OMERACT RA MRI score applied to different joint combinations, and the Sharp/van der Heijde radiographic score. Temporal changes in scores of joint destruction can be assessed by analyzing scores extracted from single measurements at each time point. However, this does not take into account the variation caused by measurement error. The smallest detectable difference of a scoring method, e.g., of the Sharp/van der Heijde radiographic score, can be used as a threshold level for definite change (19, 22); this approach was adopted in the present study of MRI and radiographic scores. MRI of unilateral wrist and MCP joints, as well as MRI of bilateral wrists and MCP joints plus unilateral MTP joints, revealed progression of joint damage in a significantly greater number of patients than did radiographic assessment of both hands, wrists, and forefeet by the Sharp/van der Heijde method. This was the case both when all observed changes were considered and when only changes above the possible measurement error (the SDD) were considered (Figure 2).

While radiography is most often performed on the hands, wrists, and feet bilaterally (7), previous MRI studies of RA patients have focused on 1 or 2 joint regions, most frequently the wrist and/or the second through fifth MCP joints (10, 23, 24). No studies have addressed the question of whether examination of additional joint regions by MRI increases the ability to identify RA patients with progressive destructive joint changes. To our knowledge, this is the first study to investigate the use of different joints/joint regions for MRI and compare their sensitivity to change. Prior to initiation of the study it was decided to focus mainly on MRI of the dominant wrist and second through fifth MCP joints (the joint areas most frequently imaged in MRI) (“few-joints approach”) versus MRI of bilateral wrists and MCP joints plus unilateral MTP joints (the maximal number of joints feasible to image in 1 session) (“many-joints approach”). We found no statistically significant difference in the number of patients with erosive progression detected by these 2 approaches, either in “raw” scores or in scores adjusted for the SDD. However, the number of patients with SDD-adjusted erosive progression was slightly higher by the “many-joints approach” (15 patients) than by the few-joints approach (13 patients), so the possibility of a Type II error cannot be ruled out. However, feasibility issues strongly favor the “few-joints approach,” which requires less than half of the scanning and evaluation time of the “many-joints approach.”

Other joint combinations were also explored. The performance of MRI of unilateral MCP joints in combination with unilateral MTP joints was comparable with that of radiography. However, MRI of only the first through fifth MTP joints did not perform as well as radiography (Table 1). The latter finding is in accordance with the lower sensitivity of MRI compared with radiography for detection of bone erosion in the fifth MTP joint, as suggested by Forslind et al (25). Furthermore, Bird et al have reported that MRI of the second through fifth MCP joints is less sensitive to progression compared with radiography (scored by the Larsen method) of both hands and wrists in RA (26). The latter contrasts with our data, since we found a slightly higher number of “progressors” by MCP joint MRI than by the Sharp/van der Heijde score (which is generally considered more sensitive to change than the Larsen method [7]), but the discrepancy between these findings may be explained by a higher sensitivity in our study due to smaller slice thickness of the MR images. However, MRI of only the second through fifth MCP joints did not provide statistically significantly higher sensitivity to change than radiography.

Thus, it appears advisable that MRI evaluation not be restricted to the MCP joints. The highest number of subjects with progressive joint destruction above the SDD was reached if MRI included bilateral wrists and second through fifth MCP joints. The fact that this approach revealed numerically more “definite progressors” than the “many-joints approach” may be explained by a higher degree of observer variation when evaluating toe joints than when assessing wrists and MCP joints. However, examination of both wrist and hand joints may be considered infeasible in the clinical setting, particularly if inflammatory changes, e.g., synovitis, are being assessed with the use of gadolinium contrast agent injection.

Consistent with previously published data (27), small bone changes resembling erosions were observed in some healthy controls. However, none of them had score changes above the SDD and no control subjects had destructive joint changes seen on radiography.

The intraobserver variation, i.e., the SDD, of the radiographic readings in this study is in accordance with previously published data (20). Raw data analysis of radiographic findings revealed a changed score in 13 patients, but this quantity was markedly reduced (to 5) if only “definite” changes (greater than the SDD) were considered. Only 1 patient exhibited a “definite” deterioration as assessed on radiography (due to an increased joint space narrowing score) that was not “caught” on MRI by the “few-joints approach,” demonstrating that radiography may, in individual cases, detect progressive joint damage not detected by MRI. However, progression was seen in this patient by the “many-joints approach.”

Low-field dedicated-extremity MRI has recently been shown to be highly sensitive and specific for detection of bone erosions in wrist and finger joints when conventional high-field whole-body MRI is used as the standard reference method, and is more sensitive and accurate than radiography (28, 29). The present results provide further evidence that low-field MRI is more sensitive than radiography in the detection of destructive joint changes.

In conclusion, MRI of unilateral wrist and MCP joints (the joint areas most frequently assessed by MRI) (“few-joints approach”) as well as MRI of bilateral wrists and MCP joints plus unilateral MTP joints (the maximum number of joints that can feasibly be evaluated by MRI in 1 session) (“many-joints approach”) displayed statistically significantly higher sensitivity to changes in structural joint damage than radiographic assessment of both hands, wrists, and forefeet by the Sharp/van der Heijde method. This was the case when comparing both changes in raw data and changes beyond the possible measurement error (SDD). Although MRI by the “many-joints approach” revealed erosive progression in a larger number of patients, no statistically significant advantage compared with the “few-joints approach” was demonstrated. Consequently, the “few-joints approach” may be preferable, due to feasibility issues. Low-field dedicated-extremity MRI is significantly superior to radiography with respect to detection of progressive joint destruction in RA.

Ancillary