To compare X-ray and MRI as diagnostic tests of active Charcot neuro-osteoarthropathy (CNO) in diabetes.
X-rays and MRI scans of 48 participants were rated for severity of fracture (0 = no fracture, 1 = fracture, 2 = collapse/fragmentation), and for absence/presence of bone marrow edema (BME) on MRI and absence/presence of bone injury on X-ray. The agreement between modalities was assessed with tests for symmetry, marginal homogeneity, and κ-coefficients.
X-ray underscored MRI in grading fractures in the metatarsals (P = 0.05) and tarsals (P < 0.001) and reported as normal 79% of the bones with BME. The agreement between X-ray and MRI for grading severity of fracture was moderate to substantial (κ = 0.53; P < 0.001) and for detecting bone injury, slight to fair (κ = 0.17; P < 0.001).
The significant underperformance of X-ray in the assessment of the hot, swollen foot in diabetes should be considered when confirming or refuting the diagnosis of active CNO.
Introduction
Charcot neuro-osteoarthropathy (CNO) is often diagnosed late, with disastrous consequences for people living with diabetes (1,2). Delayed recognition leads to bone collapse/fragmentation and treacherous foot deformity with an attendant risk of ulceration, infection, and lower-limb amputation (3). Foot and ankle X-rays and MRI scans are increasingly used in the assessment of the suspected Charcot foot, although their agreement in grading fractures is unknown. We aimed to compare X-ray and MRI as diagnostic tests of active CNO in diabetes.
Research Design and Methods
This study is a post hoc analysis of existing cross-sectional data from a clinical trial conducted in 48 participants with active CNO (clinical trial reg. no. 2009-016873-13, eudract.ema.europa.eu) (4). Trial inclusion and exclusion criteria have been reported (4). Briefly, participants were age ≥18 years and had type 1 or type 2 diabetes and intact feet. They presented with unilateral hot, swollen foot with skin temperature ≥2°C than the contralateral foot and had either typical radiographic features of active CNO (bone fracture, fragmentation, with or without joint subluxation), or abnormal MRI (bone marrow edema [BME] with or without fracture). Twenty-two bones (proximal phalanges, metatarsals, medial and lateral sesamoids, tarsals (cuneiforms, navicular, cuboid, talus, calcaneum and ankle [tibial plafond, medial and lateral malleoli]) were rated at separate sessions by a blinded musculoskeletal radiologist for severity of fracture (0 = no fracture, 1 = fracture, 2 = collapse/fragmentation) on X-ray and MRI (5). The same bones were binary graded for absence/presence of BME on MRI and for absence/presence of any bone injury (fracture or collapse/fragmentation) on X-ray.
The agreement between 961 pairs of scores (X-ray and MRI) was assessed with McNemar-Bowker test for symmetry, Stuart-Maxwell test for marginal homogeneity, and weighted (three-level score) and unweighted (binary score) κ-coefficients (6–9). Statistical analyses were performed using R version 4.1.1 software.
Results
Demographic, clinical, and imaging data were acquired before initiation of study intervention (Table 1). Of 961 pairs of scores (X-ray and MRI), a total of 201 bone injuries (rated as 1 = fracture or 2 = collapse/fragmentation) were detected by at least one modality (Table 2). Of these, only 65 (32.3%) bones were equally scored by X-ray and MRI. The remaining 136 (67.7%) bone injuries were rated by one modality (as fracture or collapse/fragmentation) but missed or underscored by the other. The proportion detecting bone injuries (fracture or collapse/fragmentation) was 0.18 by MRI and 0.13 by X-ray, with a difference of 0.05 (95% CI 0.03, 0.07) (McNemar-Bowker χ2 = 19.9; P < 0.0001).
Demographic, clinical, and imaging characteristics of study participants
Characteristic . | Mean (95% CI) or N (%) . |
---|---|
Participants, N | 48 |
Demographic variables | |
Age, years | 55 (51.8, 57.8) |
Sex, n (%) | |
Men | 37 (77) |
Women | 11 (23) |
Duration of diabetes, years | 20 (16.8, 23.8) |
Clinical variables | |
Foot skin temperature difference between the involved and the opposite foot, °C | 3.4 (3.0, 3.9) |
Vibration perception threshold of the Charcot foot, V | 35 (31.1, 39.1) |
Diagnosis of active CNO, n (%) | |
Normal X-ray and abnormal MRI | 11 (23.9) |
Abnormal X-ray and abnormal MRI | 35 (76.1) |
Imaging scores | |
X-ray (fracture or collapse/fragmentation) | 2.7 (2, 3.4) |
MRI (fracture or collapse/fragmentation) | 3.7 (3, 4.4) |
MRI (BME) | 12.5 (11.3, 13.6) |
Characteristic . | Mean (95% CI) or N (%) . |
---|---|
Participants, N | 48 |
Demographic variables | |
Age, years | 55 (51.8, 57.8) |
Sex, n (%) | |
Men | 37 (77) |
Women | 11 (23) |
Duration of diabetes, years | 20 (16.8, 23.8) |
Clinical variables | |
Foot skin temperature difference between the involved and the opposite foot, °C | 3.4 (3.0, 3.9) |
Vibration perception threshold of the Charcot foot, V | 35 (31.1, 39.1) |
Diagnosis of active CNO, n (%) | |
Normal X-ray and abnormal MRI | 11 (23.9) |
Abnormal X-ray and abnormal MRI | 35 (76.1) |
Imaging scores | |
X-ray (fracture or collapse/fragmentation) | 2.7 (2, 3.4) |
MRI (fracture or collapse/fragmentation) | 3.7 (3, 4.4) |
MRI (BME) | 12.5 (11.3, 13.6) |
All participants underwent assessment with both imaging modalities (X-ray and MRI), as the original study aimed to evaluate the full extent of bone and joint damage at presentation as well as the effect of treatment on bone healing on follow-up (4). In 35 participants, the diagnosis of active CNO was based on typical radiographic features of active CNO (bone fracture, fragmentation, with or without joint subluxation). In the remaining 11 participants in whom foot and ankle X-rays were normal, the diagnosis of active CNO was confirmed by abnormal MRI (BME with or without fracture). X-rays and MRI scans were semiquantitatively assessed at the end of study at separate sessions by a blinded musculoskeletal radiologist (5). There were no MRI scores reported in two participants. Imaging scores represent the mean number of bone injuries identified on X-rays (n = 48) and MRI (n = 46). With each modality, bones were binary scored for absence (score = 0)/presence (score = 1) of bone injury. Demographic, clinical, and imaging data (X-ray and MRI) were measured before initiation of study therapy (daily subcutaneous recombinant human parathyroid hormone or placebo) (4).
Number of bone injuries (fracture and collapse/fragmentation) reported on X-ray and MRI in participants with diabetes and active CNO
Anatomic zone . | Pairs of scores (X-ray and MRI) . | Bone injuries (fracture or collapse/ fragmentation) detected on X-ray or MRI (rated as 1 or 2 by at least one modality) . | Equal rating of bone injuries . | Bone injuries detected by MRI but missed or underscored by X-ray . | Bone injuries detected by X-ray but missed or underscored by MRI . | ||||
---|---|---|---|---|---|---|---|---|---|
Fracture or collapse/ fragmentation equally detected by X-ray and MRI (rated as 1 or 2 on both modalities) . | Fracture detected by MRI (rated 1) but missed by X-ray (rated 0) . | Collapse/ fragmentation detected by MRI (rated 2) but missed by X-ray (rated 0) . | Collapse/ fragmentation detected by MRI (rated 2) but scored as fracture by X-ray (rated 1) . | Fracture detected by X-ray (rated 1) but missed by MRI (rated 0) . | Collapse/ fragmentation detected by X-ray (rated 2) but missed by MRI (rated 0) . | Collapse/ fragmentation detected by X-ray (rated 2) but scored as fracture by MRI (rated 1) . | |||
All zones | 961 | 201 | 65 | 40 | 39 | 15 | 11 | 21 | 10 |
Phalanges | 207 | 11 | 4 | 1 | 0 | 0 | 4 | 1 | 1 |
Sesamoids | 82 | 7 | 0 | 2 | 5 | 0 | 0 | 0 | 0 |
Metatarsals | 222 | 80 | 28 | 13 | 17 | 6 | 4 | 9 | 3 |
Tarsals | 320 | 100 | 32 | 24 | 17 | 9 | 1 | 11 | 6 |
Ankle | 130 | 3 | 1 | 0 | 0 | 0 | 2 | 0 | 0 |
Anatomic zone . | Pairs of scores (X-ray and MRI) . | Bone injuries (fracture or collapse/ fragmentation) detected on X-ray or MRI (rated as 1 or 2 by at least one modality) . | Equal rating of bone injuries . | Bone injuries detected by MRI but missed or underscored by X-ray . | Bone injuries detected by X-ray but missed or underscored by MRI . | ||||
---|---|---|---|---|---|---|---|---|---|
Fracture or collapse/ fragmentation equally detected by X-ray and MRI (rated as 1 or 2 on both modalities) . | Fracture detected by MRI (rated 1) but missed by X-ray (rated 0) . | Collapse/ fragmentation detected by MRI (rated 2) but missed by X-ray (rated 0) . | Collapse/ fragmentation detected by MRI (rated 2) but scored as fracture by X-ray (rated 1) . | Fracture detected by X-ray (rated 1) but missed by MRI (rated 0) . | Collapse/ fragmentation detected by X-ray (rated 2) but missed by MRI (rated 0) . | Collapse/ fragmentation detected by X-ray (rated 2) but scored as fracture by MRI (rated 1) . | |||
All zones | 961 | 201 | 65 | 40 | 39 | 15 | 11 | 21 | 10 |
Phalanges | 207 | 11 | 4 | 1 | 0 | 0 | 4 | 1 | 1 |
Sesamoids | 82 | 7 | 0 | 2 | 5 | 0 | 0 | 0 | 0 |
Metatarsals | 222 | 80 | 28 | 13 | 17 | 6 | 4 | 9 | 3 |
Tarsals | 320 | 100 | 32 | 24 | 17 | 9 | 1 | 11 | 6 |
Ankle | 130 | 3 | 1 | 0 | 0 | 0 | 2 | 0 | 0 |
All study participants underwent assessment with both X-ray and MRI. Twenty-two bones were scored on each modality (X-ray and MRI) for severity of bone fracture (three-level ordinal scale: 0 = no fracture, 1 = fracture, and 2 = collapse/fragmentation). There were no MRI scores reported in two participants, and they were excluded from the analysis. A total of 961 pairs of scores (X-ray and MRI) were available, with an average of 20 pairs of bones per patient. Data are presented for the overall sample (all zones) and per anatomic zone, according to Sanders and Frykberg’s (11) classification.
The analysis of symmetry and homogeneity showed that the MRI-derived fracture scores were significantly higher than the X-ray–derived fracture scores (P < 0.0001) (Table 3). In addition, there was a significant difference between the two modalities in the scoring of collapse/fragmentation versus no fracture (P = 0.02). Regarding zones of involvement, the MRI-derived fracture scores were significantly higher than the X-ray–derived fracture scores in the metatarsals (P = 0.05) and tarsals (P < 0.001). No significant differences were found between X-ray and MRI in the scoring of collapse/fragmentation versus no fracture (metatarsals P = 0.12; tarsals P = 0.26) or in the scoring of collapse/fragmentation versus fracture (metatarsals P = 0.32; tarsals P = 0.61).
McNemar-Bowker and Stuart-Maxwell tests of agreement between X-ray and MRI in active CNO for the overall sample (all zones) and per anatomic zone
. | Symmetry . | Fracture vs. no fracture . | Collapse/fragmentation vs. no fracture . | Collapse/fragmentation vs. fracture . | Marginal homogeneity . | |||||
---|---|---|---|---|---|---|---|---|---|---|
Anatomic zones . | χ2 test (3 df) . | P . | χ2 test (1 df) . | P . | χ2 test (1 df) . | P . | χ2 (1 df) . | P . | χ2 (2 df) . | P . |
All zones | 22.9 | <0.0001 | 16.5 | <0.0001 | 5.4 | 0.02 | 1.0 | 0.32 | 20 | <0.0001 |
Phalanges | 3.8 | 0.38 | 1.8 | 0.38* | 1.0 | 1.0* | 1.0 | 1.0* | 3.6 | 0.16 |
Sesamoids | 7.0 (2 df) | 0.03 | 2.0 | 0.16* | 5.0 | 0.06 | No data | NA | 7.0 | 0.03 |
Metatarsals | 8.2 | 0.04 | 4.8 | 0.05* | 2.5 | 0.12 | 1.0 | 0.32 | 6.75 | 0.03 |
Tarsals | 23.1 | <0.0001 | 21.2 | <0.0001 | 1.3 | 0.26 | 0.60 | 0.61 | 17.3 | 0.0002 |
Ankle | 2.0 | 0.50 | 2.0 | 0.50 | No data | NA | No data | NA | 2.0 | 0.03 |
. | Symmetry . | Fracture vs. no fracture . | Collapse/fragmentation vs. no fracture . | Collapse/fragmentation vs. fracture . | Marginal homogeneity . | |||||
---|---|---|---|---|---|---|---|---|---|---|
Anatomic zones . | χ2 test (3 df) . | P . | χ2 test (1 df) . | P . | χ2 test (1 df) . | P . | χ2 (1 df) . | P . | χ2 (2 df) . | P . |
All zones | 22.9 | <0.0001 | 16.5 | <0.0001 | 5.4 | 0.02 | 1.0 | 0.32 | 20 | <0.0001 |
Phalanges | 3.8 | 0.38 | 1.8 | 0.38* | 1.0 | 1.0* | 1.0 | 1.0* | 3.6 | 0.16 |
Sesamoids | 7.0 (2 df) | 0.03 | 2.0 | 0.16* | 5.0 | 0.06 | No data | NA | 7.0 | 0.03 |
Metatarsals | 8.2 | 0.04 | 4.8 | 0.05* | 2.5 | 0.12 | 1.0 | 0.32 | 6.75 | 0.03 |
Tarsals | 23.1 | <0.0001 | 21.2 | <0.0001 | 1.3 | 0.26 | 0.60 | 0.61 | 17.3 | 0.0002 |
Ankle | 2.0 | 0.50 | 2.0 | 0.50 | No data | NA | No data | NA | 2.0 | 0.03 |
NA, not applicable.
Exact.
Finally, we investigated the agreement between the modalities in detecting any bone injury (BME on MRI vs. bone damage on X-ray) binary rated as 0 = absent or 1 = present. Of 961 pairs of scores, a total of 574 (59.7%) bone injuries were detected by X-ray or MRI. Of these, 452 (78.8%) bones showed BME on MRI but were reported as normal on X-ray, 2 (0.3%) bones were reported as injured by X-ray but missed by MRI, and only 120 (20.9%) bones were rated as injured by both modalities. The proportion detecting bone injury by MRI was 0.59 and by X-ray, 0.12, with a difference of 0.47 (95% CI 0.43, 0.51) (McNemar-Bowker χ2 = 444; P < 0.001). The agreement between the modalities for grading severity of fracture was moderate to substantial (κ = 0.53; 95% CI 0.46, 0.6; P < 0.001) and for detecting any bone injury, slight to fair (κ = 0.17; 95% CI 0.14, 0.2; P < 0.001) (Table 4).
Measures of agreement between X-ray and MRI in grading severity of fracture and detecting bone injuries in active CNO
. | Grading severity of fracture on X-ray and MRI . | Detection of bone injury on X-ray and MRI . | ||||
---|---|---|---|---|---|---|
Anatomic zone . | Weighted κ-coefficient (95% CI) . | P . | Agreement level between X-ray and MRI . | Unweighted κ-coefficient (95% CI) . | P . | Agreement level between X-ray and MRI . |
All zones | 0.53 (0.46, 0.6) | <0.001 | Moderate to substantial | 0.17 (0.14, 0.2) | <0.001 | Slight to fair |
Phalanges | 0.66 (0.4, 0.92) | <0.001 | Fair to almost perfect | 0 (0, 0) | NA | Slight |
Sesamoids | 0 (0, 0) | NA | Slight | 0.19 (0.07, 0.3) | <0.001 | Slight to fair |
Metatarsals | 0.47 (0.35, 0.59) | <0.001 | Fair to moderate | 0.13 (0.08, 0.19) | <0.001 | Slight |
Tarsals | 0.53 (0.43, 0.63) | <0.001 | Moderate to substantial | 0.1 (0.07, 0.13) | <0.001 | Slight |
Ankle | 0.66 (0.13, 1) | <0.001 | Slight to almost perfect | 0.15 (0, 0.31) | 0.001 | Slight to fair |
. | Grading severity of fracture on X-ray and MRI . | Detection of bone injury on X-ray and MRI . | ||||
---|---|---|---|---|---|---|
Anatomic zone . | Weighted κ-coefficient (95% CI) . | P . | Agreement level between X-ray and MRI . | Unweighted κ-coefficient (95% CI) . | P . | Agreement level between X-ray and MRI . |
All zones | 0.53 (0.46, 0.6) | <0.001 | Moderate to substantial | 0.17 (0.14, 0.2) | <0.001 | Slight to fair |
Phalanges | 0.66 (0.4, 0.92) | <0.001 | Fair to almost perfect | 0 (0, 0) | NA | Slight |
Sesamoids | 0 (0, 0) | NA | Slight | 0.19 (0.07, 0.3) | <0.001 | Slight to fair |
Metatarsals | 0.47 (0.35, 0.59) | <0.001 | Fair to moderate | 0.13 (0.08, 0.19) | <0.001 | Slight |
Tarsals | 0.53 (0.43, 0.63) | <0.001 | Moderate to substantial | 0.1 (0.07, 0.13) | <0.001 | Slight |
Ankle | 0.66 (0.13, 1) | <0.001 | Slight to almost perfect | 0.15 (0, 0.31) | 0.001 | Slight to fair |
The limits for agreement were based on the Landis-Koch benchmark scale classification (poor agreement, κ <0; slight agreement, κ = 0–0.20; fair agreement, κ = 0.21–0.40; moderate agreement, κ = 0.41–0.60; substantial agreement, κ = 0.61–0.80; and almost perfect agreement, κ = 0.81–1.0) (8). The weighted κ-coefficients graded disagreement of collapse/fragmentation (rated 2 in the ordinal scale) vs. no fracture (rated 0) as more serious (twice the severity) than a disagreement of either, collapse/fragmentation vs. fracture (rated 1) or disagreement of fracture vs. no fracture (rated 0) (9). The unweighted κ-coefficients were used to assess level of agreement between the two modalities to identify any bone injury (BME with or without fracture) on MRI vs. (fracture or collapse/fragmentation) on X-ray. For this comparison, the three-level ordinal X-ray fracture score was transformed into a binary score to differentiate bones with an injury (score 1) from bones without an injury (score 0). Data are presented for the overall study sample (all zones) and per anatomic zone according to Sanders and Frykberg’s (11) classification. NA, not applicable.
Conclusions
This study is the first to investigate the agreement between X-ray and MRI as diagnostic tests in active CNO in diabetes. X-ray significantly underscored MRI in detecting fractures in the metatarsals and tarsals, and the agreement in grading severity of fracture was moderate to substantial. When comparing the ability of each modality to detect any bone injury associated with an active CNO, X-ray reported as normal 79% of the bones that showed evidence of BME on MRI, and the agreement was slight to fair.
The failure to recognize the clinical presentation of the active Charcot foot and the paucity of assessment methods have led to misdiagnosis and poor management and prognosis (10). In the absence of disease markers, imaging is central to the investigation of the suspected CNO. Foot and ankle X-rays have been traditionally used to detect skeletal damage (11). Recently, these have been enhanced with advanced modalities, among which, MRI has been most used (12,13).
To our knowledge, no other study has examined the strength of agreement between X-ray and MRI in active CNO. Methodological grading of 22 foot bones with an established proforma with good intra- and interobserver reliability (5) offered structured assessment for the entire foot and for each of the five patterns of CNO, as classified by Sanders and Frykberg (11). The three-level scoring system was originally designed to assess treatment response and graded collapse/fragmentation as more serious than fracture alone (4,5). Equal ratings between X-ray and MRI were observed only in one-third of the injured bones, and the agreement ranged from fair to moderate in the metatarsals and from substantial to moderate in the remaining zones. On further inspection, only classifications of fracture versus no fracture in the metatarsals and tarsals were significantly different, and in both zones, MRI significantly outperformed X-ray. Patterns II and III (metatarsals and tarsals) not only are most frequently observed in CNO (11), but assumingly, most frequently unrecognized (1,2). Nearly one-half of individuals experience misdiagnosis of CNO with an average diagnostic delay of almost 85 days (2). Failure to detect fractures in the metatarsals and tarsals can precipitate foot disruption and deformity and trigger untoward adverse events (3).
Historically, foot fracture has been identified as the harbinger of the Charcot foot (14). However, research has suggested that the earliest pathological lesion is BME, which can be detected only on MRI (12,13). The latter is a key diagnostic criterion for active CNO in people with diabetes presenting with a unilateral hot, swollen, intact foot. X-ray scored as normal 79% of the bones that showed BME on MRI, confirming its poor performance in the early diagnosis of CNO. Based on concordance between MRI and X-ray, the active stage of CNO has been delineated into grade 0 with diffuse BME and soft tissue edema on MRI but with negative X-ray and grade 1 associated with BME and soft tissue edema on MRI and fractures with cortical disruption on X-ray (15). These severity grades are included in the modified Eichenholtz classification that combines clinical, radiological, and MRI features of active CNO (15).
Our study has limitations. First, the comparison between X-ray and MRI was performed as two diagnostic tests in a cohort with high clinical suspicion for active CNO (4,5). In the absence of a gold standard test, discordance was assumed to result from false-negative results in one modality rather than from false-positive results in the other. The rate of false-negative results in grading fractures was significantly higher for X-rays, although it was not 0 for MRI. Sixteen percent of the bones rated as fractured or collapsed/fragmented by X-ray were misclassified by MRI. Nevertheless, all bones were scored as positive for BME. Possibly, simplification of coding (binary rating of bones as injured or not instead of a three-level rating) or reporting by anatomic regions, such as the Balgrist Score (16), may improve reliability of scoring for nonexperts. Alternatively, composite scores (bone, tendon, muscle, ligament, and joint) may facilitate early identification of structural damage (17), as the significance of confined BME in patients with a low or low/moderate clinical index of suspicion for active CNO remains unknown (18,19).
Second, we used baseline data from a cohort assessed for participation in a randomized clinical trial in which clinical suspicion was high (4). The original study was powered to compare different treatment strategies, not imaging modalities. Thus, a larger, nonselected study sample that includes participants in whom clinical suspicion is low/moderate is needed to confirm our findings and ascertain the diagnostic accuracy of X-ray and MRI in active CNO in diabetes.
A strength of this study is an emphasis on the high rate of failure of X-ray to detect fractures of metatarsals and tarsals compared with MRI. Another strength is that the study has decisively confirmed the discrepancy between the modalities in identifying bone injury and the high proportion of bones with BME on MRI being rated as normal on X-ray.
In conclusion, the reported underperformance of X-ray in the assessment of the hot, swollen foot in diabetes should be considered when confirming or refuting the diagnosis of active CNO. Offloading while awaiting further diagnostic imaging as indicated in the recent international guideline is mandatory for improved recognition of the Charcot foot (20).
Article Information
Acknowledgments. The authors are grateful to all the study participants for their commitment to this study. The authors thank Prof. Ana Donaldson from the Biostatistics Unit, King’s College Hospital NHS Trust and Dental Institute, King’s College London, for assistance during the study.
Funding. This study was supported by Diabetes UK (project grant RD08/0003729).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. O.A. performed the statistical analysis. L.M. and D.A.E. were involved in scoring foot and ankle X-rays and MRI scans. O.A., M.E.E., and N.L.P. were involved in editing and revision of the manuscript. O.A. and N.L.P. conceived the study and wrote the initial draft of the manuscript. All authors approved the final manuscript. N.L.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented in abstract form at the 18th Diabetic Foot Study Group Meeting, Bratislava, Slovakia, 16–18 September 2022.