Comparison between 1.5T and 3.0T MRI: both field strengths sensitively detect subclinical inflammation of hand and forefoot in patients with arthralgia

Abstract

Objectives

Magnetic resonance imaging (MRI) of small-joints sensitively detects inflammation. This inflammation, and tenosynovitis in particular, has been shown to predict RA-development in arthralgia-patients. These data have predominantly been acquired on 1.0T-1.5T MRI. However, nowadays 3.0T is commonly used in practice. Evidence on the comparability of these field strengths is scarce and never included subtle inflammation in arthralgia-patients or tenosynovitis. Therefore, we assessed the comparability of 1.5T and 3.0T in detecting subclinical inflammation in arthralgiapatients.

Methods

2968 locations (joints/bones/tendon-sheaths) in hands and forefeet of 28 patients with small-joint arthralgia, at-risk for RA, were imaged on both 1.5T and 3.0T MRI. Two, blinded readers independently scored erosions, osteitis, synovitis and tenosynovitis (in line with RAMRIS). Features were summed into inflammation (osteitis, synovitis, tenosynovitis) and RAMRIS (inflammation and erosions). Agreement was assessed with intraclass correlation coefficients (ICCs) for continuous scores and after dichotomization into presence or absence of inflammation, on patient- and location-level.

Results

Interreader-ICCs were excellent (>0.90). Comparing 1.5 and 3.0T revealed an ICC of 0.90 for inflammation and RAMRIS. ICCs for individual inflammation features were: tenosynovitis 0.87 (95%CI 0.74-0.94), synovitis 0.65 (0.24-0.84) and osteitis 0.96 (0.91-0.98). Agreement was 83% for inflammation and 89% for RAMRIS. Analyses on location-level showed similar results.

Conclusion

Agreement of subclinical inflammation on 1.5T and 3.0T was excellent. Although synovitis-scores were slightly different, synovitis often occurs simultaneously with other inflammatory signs, suggesting that scientific results on the predictive value MRI-detected inflammation for RA, obtained on 1.5T-MRI, can be generalized to 3.0T-MRI.

Introduction

Magnetic resonance imaging (MRI) plays a prominent role in risk prediction for development of rheumatoid arthritis (RA), especially in the arthralgia-phase when clinical arthritis cannot yet be detected and joint inflammation is subclinical.(1) To standardize measurement of MRI inflammation, the Outcomes in Rheumatology Clinical Trials (OMERACT) working group on RA developed the RA MRI Score(RAMRIS).(2) From the different RAMRIS-inflammatory features (synovitis, osteitis and tenosynovitis), tenosynovitis has the highest accuracy for clinical arthritis development and most strongly underlies characteristic symptoms.(3, 4) In contrast, evaluating MRI for erosions in addition to subclinical inflammation does not provide added predictive value for inflammatory arthritis development.(5)

Research on predictive ability of MRI in arthralgia-patients is predominantly performed on dedicated extremity 1.0T-1.5T MRI. However, these types of MRI systems are gradually replaced, mostly by large bore 3.0T MRI. As a result, in future RA research and clinical practice, 3.0T MRI will often be used. It has been previously shown that dedicated extremity 1.5T MRI and large bore 1.5T MRI are equivalent.(6) Paradoxically, the scientific evidence on the comparability of 1.5T dedicated extremity and 3.0T large bore MRI is scarce.

Until this date, most studies comparing field strengths have focused on evaluating dedicated extremity low-field (0.2T) to mid-field MRI (1.0-3.0T).(7–9) MRI inflammation on large bore 1.5T was compared to large bore 3.0T in three studies, two of which studied only one RAMRIS feature (osteitis and synovitis respectively) without contrast-enhancement on both field strength, hence limiting the ability to depict synovitis.(10, 11) The third study compared 1.5T and 3.0T MRI (both contrast-enhanced) in 17 established RA-patients, but did not include tenosynovitis.(12)

Within this limited evidence MRI of arthralgia-patients were never studied. This is important as MRI is increasingly used in early recognition of RA. Moreover, inflammation in arthralgia-patients is more subtle than in established RA and diagnostic properties can vary between populations with different prevalence’s of inflammation. Finally, tenosynovitis, which seems to be the most predictive MRI-detected feature(3, 4), has not been compared in previous literature.

This study therefore aims to determine if there is a difference between 1.5T and 3.0T MRI in the assessment of all relevant MRI features (synovitis, tenosynovitis, osteitis and erosions) in arthralgia-patients. This study does not address either 1.5T or 3.0T as the golden standard. Instead we aspire to increase feasibility of the use of MRI in arthralgia patients in general, by examining if 3.0T can be used if 1.5T MRI is not available.

Methods

Patient population

Between August-2019 and March-2020, participants of the TREAT EARLIER trial were asked to also participate in this comparative MRI-study. The TREAT EARLIER is a randomized placebo-controlled proof-of-concept trial that evaluates the efficacy of a single dose of IM-corticosteroids and a 1-year course of methotrexate in preventing the development of clinical arthritis and RA. At trial-inclusion patients had arthralgia that was clinically suspect for progression to RA and subclinical inflammation in metacarpophalangeal (MCP)2-5, wrist and/or metatarsophalangeal (MTP)1-5 joints MCP on contrast-enhanced 1.5T MRI. During the trial follow-up period of 2 years, MRIs were repeated at different time points (4, 12 and 24 months). The trial is still ongoing and has been described in detail previously.(13) In this study, 28 trial participants underwent a contrast-enhanced 3.0T MRI of along with one of the 1.5T MRIs that were performed as part of the trial follow-up (on 4, 12 or 24 months), permitting comparison between the two field strengths. Both scans were performed with at least 2 days in between to account for sufficient clearance of IV contrast and within 7 days to prevent incomparability due to biological variation. The study was approved by the local medical ethical committee. All patients gave written informed consent.

MRI protocol

Unilateral MRIs were made of the wrist, 2nd–5th MCP and 1st–5th MTP joints of the most painful side, or the dominant side in case of symmetric symptoms. Patients were asked not to use any nonsteroidal anti-inflammatory drugs(NSAIDs) during the 24-hours before MRI.

The 1.5T MRI was performed on an MSK Extreme extremity MRI system (GE, Wisconsin, USA), using a 100mm coil. Patients were positioned in a chair beside the scanner, with the hand fixed in the coil with cushions.

The 3.0T MRI was performed on a full body MRI system (Philips, Best, NL), using dedicated extremities coils for hand and foot. Patients were positioned ‘feet first’ into the scanner, with the hand and foot fixed in the coil with cushions.

The sequences that were obtained after gadolinium-chelate-enhancement (Dotarem, dose of 0.1mmol/kg) on the 1.5T were: T1 FSE with selective fat saturation of MCP-, wrist- and MTP-, region in the coronal and axial plane. On the 3.0T these sequences were: T1-weighted SPAIR of MCP-, wrist- and MTP-region in the coronal and axial plane. Field of view (FOV) was 140mm for the forefoot on both field strengths. FOV in MCP region and wrist was 150 mm on 3.0T and 100 mm on 1.5T respectively.

Further details on the MRI protocols are provided in the supplementary methods.

Image analysis

Each MRI-scan was independently scored by two readers, blinded to clinical data. Field strengths were also scored independently. Both readers were PhD-students and had at least two years of prior experience in scoring RAMRIS-inflammation. ICCs on a 1.5T test set before the start this study were: 0.98 (intrareader 1), 0.94 (intrareader 2), 0.96 (reader 1 vs reader 2). Erosions, osteitis and synovitis were scored according to the RAMRIS-method(2), tenosynovitis according to Haavardsholm.(14) Features were considered present if scored as ≥1 by both readers. Features were summed into inflammation (osteitis, synovitis, tenosynovitis) and RAMRIS (inflammation and erosions).

Further details on image analysis are provided in the supplementary methods.

Statistical analysis

Intraclass correlation coefficients (ICCs; 2-way-mixed-effects-model) of continuous scores assessed interreader-reliability (comparing reader 1 & reader 2) and field strength agreement (comparing 1.5T&3.0T).

ICC-values were interpreted as follows: <0.50 poor reliability; 0.50-0.75 moderate; 0.75-0.90 good; and >0.90 excellent.(15) Bland-Altman(BA)- and correlation plots were drawn. Next, field strength agreement was determined after dichotomization of inflammation (present/absent) and calculated as the proportion of concordant scores. Analyses were performed on patient- and location level (joint/bone/tendon sheath). SPSS v25 was used.

Results

Patient characteristics of the 28 included patients are given in supplemental table 1. Mean age was 52 years (sd14), 71% was female, 21% was ACPA-positive. Median inflammation and RAMRIS were 3 and 5 on 1.5T, respectively.(Table 1) Mean scanning interval was 4 days. Scan time without positioning and survey scans were 25 minutes on 1.5T and 20 minutes on 3.0T. We kept an extended report of any adverse events, as these patients also participated in an RCT.(13) No side-effects of any MRI, either 1.5T or 3.0T, were reported.

Table 1. Agreement of different field strengths, scored according to RAMRIS in arthralgia-patients.

	Median	Median	ICC	ICC	ICC
	score	score	3.0T-1.5T	3.0T-1.5T	3.0T-1.5T
	1.5T	3.0T	Reader 1	Reader 2	Mean of the two readers
	(IQR; max)	(IQR; max)	(95% CI)	(95% CI)	(95% CI)
RAMRIS	5 (2-8; 35)	7 (4-9; 31)	0.81(0.62-0.91)	0.92(0.82-0.96)	0.90(0.78-0.95)
Inflammation	3 (2-4; 29)	4 (3-6; 25)	0.79 (0.60-0.90)	0.93 (0.83-0.97)	0.90 (0.78-0.95)
Erosions	2 (1-4; 7)	3 (2-4; 8)	0.78 (0.56-0.89)	0.77 (0.56-0.89)	0.81 (0.63-0.91)
Osteitis	1 (0-2;11)	1 (0-1; 11)	0.81 (0.63-0.91)	0.89 (0.71-0.95)	0.96 (0.91-0.98)
Synovitis	1 (0-3; 10)	3 (1-4; 8)	0.59 (0.26-0.79)	0.76 (0.28-0.90)	0.65 (0.24-0.84)
Tenosynovitis	0 (0-1; 8)	0 (0-1; 7)	0.66 (0.39-0.82)	0.81 (0.63-0.91)	0.87 (0.74-0.94)

Median of the mean score of the two readers are given in the first column. ICCs are based on single measure in a two-way mixed effect model. RAMRIS is the summed score on erosions, osteitis, synovitis and tenosynovitis. Inflammation is the summed score on osteitis, synovitis and tenosynovitis.

Abbreviations: ICC= intraclass correlation coefficient (ICC); IQR interquartile range; max= maximum; CI=confidence interval

Interreader-reliability

Interreader-reliability on both 1.5T and 3.0T was excellent for RAMRIS- and inflammation scores (ICCs≥0.95) and good to excellent for separate features (ICC≥0.82)(Supplemental Table 2).

Field strength agreement; continuous scores

Median scores on both field strengths were similar (Table 1). Field strength ICCs were excellent for inflammation (0.90; 0.78-0.95) and RAMRIS (0.90;0.78-0.95). ICCs per feature were good to excellent for erosions, osteitis and tenosynovitis (0.81 (0.63-0.91), 0.96 (0.91-0.98), and 0.89 (0.77-0.95), respectively) and moderate for synovitis (0.65 (0.24-0.84)).

Bland-Altman and correlation plots showed that systematic bias was low and mainly caused by higher synovitis scores on the 3.0T.(Figure 1)

Figure 1. Bland-Altman plots (A) and correlation plots (B) comparing 1.5T and 3.0T.

A: Bland-Altman plots depicting agreement between two field strengths for summed inflammation and RAMRIS scores and per feature separately. The Y-axes demonstrate the absolute difference between 1.5T minus 3.0T. The x-axes denote the average score of two field strengths ((1.5 T + 3.0 T) /2). The middle grey line depicts the mean difference between the field strengths, the upper and lower line depict the ±95% limits of agreement.

B: Pearson’s correlation coefficients on log transformed scores are 0.86(p<0.01) for RAMRIS score, 0.75(p<0.01) for inflammation score 0.85(p<0.01) for erosions, 0.86(p<0.01) for osteitis, 0.60(p<0.01) for synovitis, 0.77(p<0.01) for tenosynovitis. The middles dotted reference line depicts absolute agreement between the two systems.

RAMRIS is the summed score on erosions, osteitis, synovitis and tenosynovitis. Inflammation is the summed score on osteitis, synovitis and tenosynovitis.

Field strength agreement; dichotomized scores

Agreement between dichotomized 1.5T and 3.0T scores was 89% for RAMRIS and 83% for inflammation.(Table 2) Evaluating the RAMRIS-features separately, less agreement was found for synovitis (68%) compared to the other inflammatory features (≥82%). Discordance was primarily caused by positive scores on 3.0T with accompanying negative scores on 1.5T.

Table 2. Agreement of the two field strengths on dichotomized scores on patient-level (A) and location-level (B).

	Concordance			Discordance
	1.5T+/3.0T+	1.5T-/3.0T-	Agreement	1.5T-/3.0T+	1.5T+/3.0T-
Patient-level (A)
RAMRIS(%)	89	0	89	11	0
Inflammation(%)	79	4	83	18	0
Erosions(%)	75	11	86	14	0
Osteitis (%)	43	39	82	11	7
Synovitis (%)	57	11	68	32	0
Tenosynovitis(%)	32	57	89	7	4
Location-level (B)
RAMRIS(%)	3	94	97	2	1
Inflammation(%)	2	96	98	2	1
Erosions(%)	5	91	96	2	2
Osteitis(%)	1	97	98	1	1
Synovitis(%)	5	83	88	9	3
Tenosynovitis(%)	1	97	98	1	1

⁺presence of a feature or summed score ≥1, -: absence of a feature or summed score <1, Field strength-agreement was determined after dichotomization into presence or absence of inflammation and calculated as the proportion of concordant dichotomized scores on both field strengths. A feature or summed score was considered present if scored as ≥1 by two readers. RAMRIS is the summed score on erosions, osteitis, synovitis and tenosynovitis. Inflammation is the summed score on osteitis, synovitis and tenosynovitis. Results are given as the percentage of the total number of patients (n=28)(A) resp. bones/joints/tendon sheaths(n=2968)(B): 924 bones are scored for erosions and osteitis (224 in hands, 420 in wrists and 280 in feet), 336 synovial linings are scored for synovitis (112 in hands, 84 in wrists and 140 in feet) and 784 tendons are scored for tenosynovitis (224 in hands, 280 in wrists, 280 in feet).

On location-level, agreement was 98% and 97% for inflammation and RAMRIS, respectively. For synovitis the agreement was 88% and for the other features >96%.

For illustrative purposes, 1.5T and 3.0T MRI-images of the same patient are presented in figure 2.

Figure 2. Examples of a contrast enhanced sequence of the MCP region with tenosynovitis and MTP region with synovitis of the same patients scanned at 3T and 1.5T.

Image A and B reveal tenosynovitis of the third flexor tendon sheath in the MCP region on 3.0T (A) and 1.5T (B). Image C and D show synovitis op the first MTP joint on 3.0T (C) and 1.5T (D).

Discussion

Predictive values of MRI in (pre-)RA have been mostly acquired on 1.0T-1.5T. However, 3.0T is nowadays increasingly used in early recognition of RA. Paradoxically, scientific evidence on the comparability of assessment of subclinical inflammation with these field strengths is scarce and unavailable in arthralgia-patients without clinical arthritis. We showed that agreement of 1.5T and 3.0T was predominantly good to excellent, but moderate for synovitis. Our research suggests scientific results on the predictive value of subclinical inflammation and tenosynovitis, obtained on dedicated extremity 1.5T can be generalized to large bore 3.0T when used in daily practice. However, when assessing synovitis as a separate feature on 3.0T, it should be kept in mind that synovitis scores on this field strength could be somewhat increased compared to 1.5T.

3.0T was slightly more sensitive for synovitis, leading to moderate agreement for this feature on patient-level. The most important explanation for this minor difference is the dependence on field strength of tissue contrast enhancement and the ensuing contrast between enhancing and nonenhancing tissue. The higher the field strength, the higher the signal intensity on Gd-chelate enhanced T1-weighted images.(16, 17) Reassuringly, on location-level, dichotomized agreement on synovitis was high. Moreover, for tenosynovitis, inflammation and RAMRIS, agreement was excellent. From the different inflammatory features, tenosynovitis and inflammation have been shown to be considerably more predictive for the development of clinical detectable arthritis and osteitis is the inflammatory feature that is most predictive for erosion development. (1, 18, 19) In contrast, synovitis is not an independent predictor and subtle synovitis (grade 1) is common in older healthy individuals. Furthermore, the different inflammatory features often occur concomitantly in patients. Altogether, although the agreement for synovitis was moderate compared to 1.5T and we have no gold standard, the reliability between 3.0 and 1.5T of most features was very high. We therefore assume that the predictive accuracy of MRI on 3.0T on patient-level, is not considerably different compared to 1.5T MRI.

Future longitudinal research should determine if any added predictive value exists of the 3.0T MRI as a more sensitive modality for detection of synovitis. Furthermore, the benefit of automatic or semi-automatic AI-based segmentation methods could be investigated on both field strengths. In addition, a potential advantage of large bore 3.0T is the larger field of view compared to 1.5T extremity MRI, allowing a reduction in MRI scan time. This potential advantage could be further explored.

The current study is the first study comparing field strengths in arthralgia-patients, which is a meaningful advance, as MRI plays a prominent role in risk-prediction in this early stage of the disease. Only one small study has compared MRI-detected RAMRIS-features on contrast-enhanced 1.5T and 3.0T in established RA-patients.(12) This previous study is concordant in their finding that no significant differences on the number and extent of erosions, osteitis and synovitis exist between 1.5T and 3.0T. In addition, our data suggest that tenosynovitis is also reliably assessed on both field strengths. This is important in the phase of arthralgia, as tenosynovitis has the highest predictive value for RA-development.(3, 4)

Since this study was performed in arthralgia patients, it important to consider that the current results cannot be directly extrapolated to other populations, such as established RA-patients. The choice to study arthralgia-patients could have posed a statistical hurdle in assessing agreement measures, since ICCs are sensitive to a lack of variability among sampled subjects and subclinical inflammation is per definition reflected by low RAMRIS-scores with low variability.(15) Nevertheless, reliability- and agreement scores of RAMRIS-features were largely good to excellent.

Secondly, another possible limitation in this study is that readers only had previous experience with scoring 1.5T images and not with 3.0T images. Reassuringly, interreader-ICCs were good to excellent on both MRIs.

Thirdly, arthralgia-patients in this study might have been treated with methotrexate in the context of the Treat Earlier trial. This possibly influenced the amount and presence of subclinical MRI inflammation in these patients. However, as this study has a cross-sectional design, this cannot have influenced the comparison between field strengths.

Finally, a limitation could be that we applied the scoring method developed by Haavardsholm et al. for tenosynovitis.(14) The RAMRIS was recently updated and now includes a slightly modified tenosynovitis-score(21) However, as both field strength were scored according to the same protocol in this study, comparability could be adequately assessed. Based on current results, differences in agreement of 1.5T and 3.0T when using updated protocols is not expected.

A strength of this study is scoring of all images by two independent readers, blinded for the results on the other field strength, ascertaining a reliable comparison between the two field strengths.

In conclusion, although the moderate agreement on synovitis between the two field strengths should be taken into account, the current data imply that scientific results on the predictive value MRI-detected inflammation, obtained on 1.5T-MRI, may be generalized to 3.0T-MRI when used in daily practice.

Data availability statement

Data are available from DK upon reasonable request.