The geometric architecture of the subtalar and midtarsal joints in rheumatoid arthritis based on magnetic resonance imaging




To compare in vivo the 3-dimensional (3-D) geometric architecture of the subtalar and midtarsal joints in normal and rheumatoid arthritic (RA) feet, using magnetic resonance imaging (MRI) analysis.


MRI was performed on 23 patients with RA, all of whom had disease activity in the subtalar and/or midtarsal joints. Image processing techniques were used to create 3-D reconstructions of the calcaneus (C), cuboid (c), navicular (N), and talus (T) bones. Twenty-four standard architectural parameters were measured from the reconstructions and were compared with data from 10 normal subjects. These parameters defined both 3-D distance and angular relationships among the 4 bones studied. Pattern classification techniques were used to establish a geometric architecture foot profile for the RA patients. The degree of individual patient fit to the new RA foot profile and to profiles for normal, pes planus, and pes cavus foot types was derived. Logistic regression was used to examine the relationship of foot architecture to inflammatory disease characteristics and physical examination variables.


Subtalar or midtarsal pain was reported by all 23 patients, and 22 of the 23 patients presented with ≥1 clinical feature of pes planovalgus deformity. In 21 patients, ultrasonography revealed synovitis at ≥1 tarsal joint or surrounding tendon. In the RA group, the normalized distances between the geometric centroids were significantly closer for bone pairs Cc and cT and significantly distracted for bone pair CN compared with the distances in normal subjects. In RA patients (versus normal subjects), the angles subtended at the bone centroids were significantly decreased in 3 bone groups (CNc, TCN, and TNc) and significantly increased in 3 bone groups (CcN, CcT, NTc). The angles formed between the major principal axes of bone pairs CT and cT were significantly increased in RA patients compared with those in normal subjects. Pattern classification defined 11 RA feet as having normal structure and 12 as having abnormal structure. However, the abnormal feet did not fit consistently with structures defined for RA, pes planus, or pes cavus foot types. Logistic regression demonstrated that subtalar joint synovitis was the only predictive factor for abnormal subtalar and midtarsal architecture (odds ratio 19.2, 95% confidence interval 1.77–200.0).


This unique 3-D MRI-based technique successfully quantified the effects of RA on the geometric architecture of the foot and the patient-specific nature of these changes. This technique can be used to provide logical therapy for correction.

The reported prevalence of adult-acquired pes planovalgus deformity in patients with rheumatoid arthritis (RA) is 46–64% (1–4). This condition is associated with significant localized pain and gait dysfunction and is difficult to treat, both conservatively and surgically (5–7). Pes planovalgus is a clinical description of 1 or more geometric features, including loss of arch height, abduction of the forefoot, and valgus of the rearfoot (1, 2, 5). Radiologically, pes planovalgus has been defined when standardized angles, such as the talocalcaneal angle of Kite, are above normal limits (3–5). The tarsal joints are relatively small and form complex structures, which have been shown to be vulnerable to the pathologic changes associated with persistent synovitis in RA (6, 8–10). Patient-specific interactions of localized inflammatory and biomechanical factors may contribute to the development of different patterns of midfoot and rearfoot deformity (loosely referred to clinically as flatfoot).

Plain film 2-dimensional (2-D) radiographic techniques have been shown to lack precision and accuracy, and such images are also difficult to obtain and measure in some patients with RA (5, 11). Given the complex nature of tarsal joint disease in RA, the 2-D technique may be insensitive to the complex changes in joint geometry anticipated in the RA foot (5, 11). However, using 3-D imaging techniques, complex foot deformity in conditions such as diabetes, symptomatic flatfoot, and infant clubfoot has been successfully quantified (12–14). A complete magnetic resonance image (MRI)–based technique for in vivo quantification of peritalar morphology, architectural geometry, and joint kinematics has been previously described (15–18). This well-validated technique quantifies important architectural parameters for all of the bones in the tarsal complex in 3 dimensions with a high degree of precision and accuracy, and it may reveal patterns or relationships that are not otherwise apparent (15, 16). Furthermore, it avoids the irradiation issues associated with radiography and computerized tomography. Therefore, we conducted a study to compare the 3-D geometric architecture of the tarsus in normal subjects with that in RA patients who had subtalar and/or midtarsal joint disease, using an MRI-based approach.


Patient selection.

During a 3-month period, 23 consecutive patients with arthritis of the subtalar and/or the midtarsal joints were recruited from the outpatient rheumatology departments of the Leeds General Infirmary and St. Luke's Hospital, Bradford, UK. Patients were eligible for inclusion if they met the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for the diagnosis of RA (19). Approval for this project was obtained from the local research ethics committee.

Clinical assessment.

Patients self-reported the sites of pain in the right foot, using 4 standard photographs (anterior, posterior, medial, and lateral ankle region from the malleoli to the midfoot). Localized symptoms of pain, stiffness, and instability that had occurred in the last month were recorded. Independent clinical examinations were conducted by an orthopedic surgeon (AB) and a podiatrist (JW), and a consensus was reached regarding the presence or absence of right subtalar and midtarsal joint swelling and tenderness, medial longitudinal arch depression, valgus deformity of the heel, and medial bulging of the talonavicular joint. An overall grade for the severity of the combined foot deformities was assigned (grade I = mild, grade II = moderate, grade III = severe).

Ultrasound examination was performed by an experienced sonographer (RJW) using an HDI 3000 unit (Advanced Technologies Laboratories, Bothell, WA) with a 10-5 MHz “hockey stick” transducer. Examinations were performed at the subtalar, talonavicular, and calcaneocuboid joints to determine the presence of synovitis and bone damage. The tendons of the medial ankle (tibialis posterior, flexor digitorum, and flexor hallucis longus) and those of the lateral ankle (peroneus longus and peroneus brevis) were also examined for the presence of tenosynovitis. Synovitis was defined as an abnormal hypoechogenic area within the joint, and bone erosion was defined as a cortical defect detected in at least 2 planes. Tenosynovitis was defined as an abnormal hypoechogenic area around the tendon within the tendon sheath.

MRI protocol.

MRI of the ankle to the midtarsal joints of the right foot was performed. Images were acquired on a Gyroscan ACS-NT 1.5-Tesla MR scanner (Philips, Best, The Netherlands). MR parameters were as follows: a 3-D T1-weighted gradient-echo sagittal pulse sequence with repetition time 22 msec, echo time 9.2 msec, and flip angle 55°; a 256 × 256 acquisition matrix; 150–190-mm field of view; 50–60 1.5-mm thick contiguous slices with a pixel size of 0.78 mm; and an acquisition time of 212 seconds. The orientation of the right foot in the scanner was standardized by using a custom nonmetallic pronation–supination jig (Figure 1) (15). The jig held the foot against a vertical plate designed to rotate on a pivot oriented to the mean subtalar joint axis (15). With the lower leg secured, the foot was rotated until neutral alignment of the subtalar joint was achieved, and the jig was locked and secured. To maximize the MRI signal, 2 elements of a commercial body coil were placed medially and laterally over the ankle regions inside the jig (15).

Figure 1.

Nonmetallic pronation–supination magnetic resonance imaging foot jig, with foot positioned in neutral alignment.

Image processing.

3DVIEWNIX software (Medical Image Processing Group, University of Pennsylvania, Philadelphia) was used for 3-D reconstruction of the calcaneus, cuboid, navicular, and talus bones (15). Briefly, this process involved slice-by-slice “live-wire” segmentation of the bone surfaces to produce binary volume images containing the 4 bones. The binary images were interpolated with a shape-based algorithm to minimize artifacts in the final rendering. The images were filtered with a smoothing Gaussian filter, and the surfaces were created and displayed by a surface-rendering algorithm (15). Figure 2 depicts 3-D renditions of the 4 tarsal bones in a normal subject and in a patient with RA. The original volume image has a reference system called the scene coordinate system. Within that system, each 3-D reconstructed bone has an associated reference system called the bone axes system. For a bone b, the bone axes system is defined by the geometric centroids Gb and by 3 unit vectors representing the principal axes of the bone, αb, βb, γb, expressed in the scene coordinate system. Once the surface of b is determined, Gb,αb, βb, and γb are computed automatically.

Figure 2.

Three-dimensional rendition of tarsal reconstruction (slight posteromedial view) in a normal subject and a patient with rheumatoid arthritis (RA; patient 4). A, In the neutral position, the medial longitudinal and transverse arch of the normal subject is well developed. B, In the RA patient, the calcaneus and cuboid bones are more planar and the calcaneus everted, while the talus is displaced in an anterior, plantar, and medial direction. The navicular is also directed plantarward.

Geometric architecture definition.

The subtalar and midtarsal joints are formed by the articulations of the calcaneus (C), cuboid (c), navicular (N), and talus (T) bones. We adopted published classification criteria that utilize 24 parameters of geometric architecture for the subtalar and midtarsal joints by MRI (15–17): a) the 3-D Euclidean distance between the geometric centroids, normalized to the length of the principal axis of the calcaneus, giving 6 distance parameters (denoted by dCc, dCN, dCT, dcN, dcT, and dTN) for the 4 bones studied; b) the angles subtended at the centroids (for each of the 4 possible triplets of bones b, b′, b′′, we define 3 angles [in degrees] subtended at Gb, Gb′, Gb′′ [Figure 3]. There are 4 possible groups of 3 bones and 12 angles, which are referred to as aCcN, aCNc, aNCc, aCcT, aCTc, aTCc, aCNT, aCTN, aTCN, aTNc, aNTc, and aTcN. Figure 3 shows these angles for the triplet CTc); and c) the angles between the major principal axes.

Figure 3.

Three-dimensional rendition of the calcaneus, cuboid, talus, and navicular bones. The architectural parameters are represented using the bone triplet calcaneus/talus/cuboid (CTc). A, The geometric centroids for each bone are represented by GC, Gc, and GT. The distances between the bone pairs (d) are represented by dCT, dTc, and dCc; the angles subtended at the centroids (a) are represented for the bone triplet CTc by aTCc, aCTc, and aCcT. B, The angles between the principal axes of the calcaneus (PAC) and cuboid (PAc) are represented by ACc. Color figure can be viewed in the online issue, which is available at

Given each bone pair b,b′, the angles between their major principal axes, as defined by the vectors αb and αb′, are calculated (Figure 3). There are 6 possible angles between the 4 bones, which are designated ACc, ACN, ACT, AcN, AcT, and ANT. These parameters were calculated automatically within the 3DVIEWNIX software. Variability, most likely from motion artifact, user interaction during segmentation, or foot position in the jig in the MR scanner, was <0.5% for the distance and subtended angle parameters and <2% for angles formed between the principal axes for repeat 3-D reconstructions. We compared these parameters with published values derived from 10 normal subjects who underwent imaging at the University of Pennsylvania Medical Center in a similar jig with their feet in a comparable position (17). This comparison was justified, because preliminary studies revealed that the technique could be replicated in the collaborating institutions with a high degree of precision.

Algorithm profile and classification of feet.

The set of 24 parameters was used to establish the characterization and classification of the geometric architecture of the RA feet. For this, a pattern classification technique was used to create a profile that specifies which parameters are crucial in discriminating a pathologic group from other groups, and to determine the extent of deviation from these parameters (20). This technique has been used successfully to distinguish normal feet from pes planus and pes cavus foot types (18). According to this technique, for each parameter pi (1 ≤ i ≤ 24) we derive a counter value Ki. Starting from 0, the counter is incremented or decremented in a weighted manner for each parameter as follows: if value of pi < minimum value then decrement Ki by 2; if minimum value ≤ value of pi < mean value then decrement Ki by 1; if mean value < value of pi ≤ maximum value then increment Ki by 1; if maximum value < value of pithen increment Ki by 2. The minimum and maximum values of the parameters were defined as the normal mean minus 1 SD and the normal mean plus 1 SD, respectively (18). Finally, the counter values were normalized to lie within a range of −1 to +1. A threshold of Ki = 0.50 was used to include pi in the profile. For the RA patient group g, the fraction Fg of parameters that matched the profile for the normal, pes planus, pes cavus, and RA foot types was determined. When the Fg was ≥0.5 against each foot type profile, the foot was classified in this group. When >1 foot type classification was indicated, a hierarchy was established based on the magnitude of the fractions. Normal, pes cavus, and pes planus profiles were derived from published data (18).

Statistical analysis.

Means and SD are given as descriptive statistics. Statistical comparisons between the study groups for the 24 architectural parameters were performed by unpaired t-test. Because multiple comparisons were performed, P values were adjusted based on Holm's method. When P values were less than 0.05, the decisions made in the corresponding tests were considered significant.

Using the algorithm profile, the geometric architecture of the foot was classified dichotomously (normal or abnormal). Logistic regression was used to examine the relationship between architecture and inflammatory disease characteristics and physical examination variables. Only variables with P values less than 0.05 by chi-square univariate testing were selected for entry into a backward, stepwise, multivariate logistic regression analysis. Results are presented as the summary odds ratio (OR) with 95% confidence interval (95% CI).


Demographic and clinical characteristics.

Fourteen female and 9 male patients with a mean disease duration of 6.6 years were recruited into the study (Table 1). Of the 23 patients, 19 were taking disease-modifying antirheumatic drugs (DMARDs). All patients reported symptoms of pain in the subtalar/midtarsal region, with varying combinations of stiffness and instability, and 21 of 23 had clinical evidence of localized swelling and tenderness. Twenty-two patients presented with ≥1 feature of rearfoot/midfoot deformity, and the clinical impression was that 9 patients had mild deformity, 11 had moderate deformity, and 3 had severe deformity.

Table 1. Demographic and clinical data, and clinical features and high-resolution ultrasound changes at the right subtalar joint (ST) and right midtarsal joint (MT) in 23 patients with rheumatoid arthritis*
Patient/age/sexDisease duration, yearsDMARD useST/MT symptomsST/MT clinical featuresST deformityObserved foot deformityChanges identified by ultrasonography
PainStiffInstSwellingTendernessDep MLAValg heelTN bulgeTTSTTNCCOther
  • *

    DMARD = disease-modifying antirheumatic drug; Stiff = stiffness; Inst = instability; Dep MLA = depression of the medial longitudinal arch on weight bearing; Valg heel = valgus heel deformity on weight bearing; TN bulge = medial bulging of the talonavicular joint (TN) on weight bearing; TT = tibiotalar joint; CC = calcaneocuboid joint; S = synovitis; TNFα = tumor necrosis factor α blockade; MTX = methotrexate; E = erosion; SSZ = sulfasalazine; TP = tibialis posterior tendon; FDL = flexor digitorum longus tendon; TS = tenosynovitis; PL = peroneus longus tendon; PB = peroneus brevis tendon; oAU = oral gold; HCQ = hydroxychloroquine.

6/68/F1SSZ+++TN, CCST,TN,CC+++ModerateS,ESTP,FDL,TS


Ultrasonography detected synovitis at the tibiotalar joint in 10 of the 23 patients, at the subtalar joint in 15 patients, at the talonavicular joint in 12 patients, and at the calcaneocuboid joints in 12 patients (Table 1). Erosive arthropathy was detected at the tibiotalar joint in 3 patients, at the subtalar joint in 4 patients, at the talonavicular joint in 5 patients, and at the calcaneocuboid joint in 4 patients. Tenosynovitis of the tibialis posterior was present in 9 patients, but none, based on the criteria described by Jernberg et al, had complete tendon rupture (21). Tenosynovitis of the flexor digitorum longus was present in 4 patients, of the peroneus longus in 3 patients, and of the peroneus brevis in 4 patients.

Normalized distance between the centroids.

As shown in Table 2, in RA patients the distance between the geometric centroids was significantly decreased between bone pairs Cc (P = 0.006) and cT (P = 0.005), whereas it was significantly increased between bone pair CN (P = 0.036), compared with the distances in normal subjects.

Table 2. Geometric architectural parameters for the RA patients and normal subjects*
ParameterRA (n = 23)Normal (n = 10)PMean difference (95% CI)
  • *

    RA = rheumatoid arthritis; 95% CI = 95% confidence interval; d = normalized distance between the geometric centroids; C = calcaneus; c = cuboid; N = navicular; T = talus; a = angles subtended at the centroids; A = angles between the major principal axes.

  • See ref. 17.

  • By unpaired t-test.

dCc62.362.7164.571.610.006−2.21 (−3.39, −1.04)
dCN71.372.3769.951.990.0361.42 (0.39, 2.44)
dCT46.291.6646.031.200.4690.26 (−0.46, 0.97)
dcN39.192.1238.742.240.6380.45 (−0.47, 1.37)
dcT58.293.0960.633.290.005−2.34 (−3.68, −1.01)
dTN43.642.0744.351.800.345−0.71 (−1.60, 0.19)
aCcN86.213.1881.312.68<0.00014.90 (3.53, 6.28)
aCNc60.602.0965.682.61<0.0001−5.08 (−5.98, −4.18)
aNCc33.191.6533.012.120.6060.18 (−0.53, 0.89)
aCcT44.901.3842.811.62<0.00012.09 (1.49, 2.68)
aCTc72.253.6273.684.070.216−1.43 (−3.00, 0.14)
aTCc62.853.5563.504.590.776−0.65 (−2.19, 0.88)
aCNT38.692.3639.681.900.224−0.99 (−2.01, 0.03)
aCTN105.184.61102.773.930.1202.41 (0.41, 4.40)
aTCN36.142.6639.041.95<0.0001−2.90 (−4.06, −1.75)
aTNc89.303.7792.703.300.0002−3.40 (−5.04, −1.77)
aNTc42.222.2539.491.69<0.00012.73 (1.76, 3.70)
aTcN48.482.6547.282.380.2051.20 (0.05, 2.35)
aCc129.3113.49122.455.510.0926.86 (1.02, 12.69)
ACN64.9511.4165.352.400.867−0.40 (−5.23, 4.53)
ACT148.145.09144.482.530.0123.66 (1.46, 5.86)
AcN69.029.8365.846.850.4053.18 (−1.07, 7.43)
AcT136.0210.10129.415.790.0256.61 (2.24, 10.98)
ANT88.769.6786.204.330.4362.56 (−1.62, 6.74)

Angles subtended at the centroids.

Compared with normal subjects, in RA patients the angles subtended at the centroids were significantly larger for bone triplets CcN (P < 0.0001), CcT (P < 0.0001), and NTc (P < 0.0001) (see Table 2). For bone triplets CNc (P < 0.0001), TCN (P < 0.0001), and TNc (P = 0.0002), the subtended angles were significantly smaller than normal.

Angles between the major principal axes.

Compared with normal subjects, in RA patients the angles between the major principal axes were significantly increased for bone pairs CT (P = 0.012) and cT (P = 0.025) (Table 2).

RA foot profile.

Nine parameters were included in the RA foot profile; 7 were from the subtended angle parameters, of which 4 (aCcN, aCcT, aCTN, and aNTc) were above superior limits and 3 (aCNc, aTCN, and aTNc) were below the inferior limits, and 2 were from the principal axis parameters (ACT and AcT), both of which were above superior limits (Table 3). The RA foot profile was not characterized by any of the parameters related to the distance between the geometric centroids.

Table 3. Degree of deviation from normal range for 24 architectural parameters in 23 patients with rheumatoid arthritis*
  • *

    Ki = counter value for individual parameters (range −1 to +1); d = normalized distance between the geometric centroids; C = calcaneus; c = cuboid; N = navicular; T = talus; a = angles subtended at the centroids; A = angles between the major principal axes.

Subtended angle 
Principal axis angle 

Algorithm profile.

In the RA group g, 11 patients had an Fg <0.5 in comparison with 24 parameters from the normal group (average fraction 12/24) (Table 4). Of these patients, only 1 (patient 15), had an abnormal foot type classification (pes cavus). Of the 12 RA patients with an Fg ≥0.5 in comparison with normal, 4 (patients 10, 11, 16, and 17) had foot types that were unclassified according to the parameters set for pes planus, pes cavus, and RA. Two patients (patients 6 and 20) had abnormal, pes planus, and RA foot classification, where Fg ≥0.5, although the hierarchy for the classification was different in each case. Five patients (patients 2, 4, 8, 12, and 14) had abnormal and pes cavus foot classifications, with the abnormal classification being superior in the hierarchy in 1 patient, and the pes cavus classification being superior in 4 patients. One patient (patient 3) had both abnormal and pes planus classifications, with the abnormal classification being superior in the hierarchy.

Table 4. Fraction of parameters that fulfill the normal, pes planus, pes cavus, and rheumatoid arthritis (RA) foot profiles for each foot in the RA group*
PatientNormal (n = 10)Pes planus (n = 3)Pes cavus (n = 2)RA (n = 23)Classification hierarchyClinical impression
  • *

    The clinically normal (N) foot type is defined by 24 parameters; the pes planus (PP) foot type is defined by 15 parameters; the pes cavus (PC) foot type is defined by 11 parameters; the RA foot is defined by 9 parameters. A = abnormal; I = clinically observed mild rearfoot/midfoot deformity; II = clinically observed moderate rearfoot/midfoot deformity; III = clinically observed severe rearfoot/midfoot deformity (17).

118/243/153/110/9N   I
210/241/158/110/9PCA  II
38/249/155/113/9APP  II
410/244/157/113/9APC  III
517/246/153/112/9N   I
69/248/154/114/9APPRA II
713/244/152/111/9N   I
812/242/157/110/9PCA  II
913/244/153/110/9N   II
107/244/154/110/9A   I
1112/243/153/110/9A   II
1210/244/157/113/9PCA  III
1319/243/153/110/9N   II
1412/243/157/112/9PCA  II
1515/244/156/110/9NPC  I
168/244/154/112/9A   II
1711/245/155/113/9A   I
1813/241/155/112/9N   I
1913/246/154/110/9N   I
205/2412/155/114/9PPARA III
2114/246/154/112/9N   II
2213/245/155/110/9N   II
2313/243/151/110/9N   I

Predictors of geometric architecture.

From chi-square testing, 2 potential predictors of abnormal foot architecture were found: the presence of subtalar joint synovitis or erosion, and bulging of the talonavicular joint on physical examination. From the stepwise multivariate logistic regression analysis, only ultrasound-proven subtalar joint synovitis with or without erosion contributed significantly (OR 19.2, 95% CI 1.77–200.0) to predicting abnormal foot architecture (R2 = 0.409).


The aim of the present study was to compare the geometric architecture of the subtalar and midtarsal joints in normal and RA feet using MRI reconstruction techniques. The RA patients demonstrated significant tarsal derangement, characterized by multiple changes in the normal distance and orientation parameters between the bones forming the subtalar and midtarsal joints. Furthermore, these changes were accompanied by ultrasonographic evidence of joint inflammation and erosion and tenosynovitis of the ankle tendons, especially the tibialis posterior. However, only the presence of subtalar joint synovitis with or without erosion was predictive of abnormal architectural geometry.

Local symptoms of pain, stiffness, and instability were reported in all patients, and clinical examination revealed ≥1 sign of pes planovalgus in 22 of the 23 patients. Use of MR-based 3-D reconstruction of the tarsal joints to study the complex structure of the foot is an attractive prospect. The technique is unique in that it provides both distance and orientation descriptions between multiple bones in 3 dimensions. We previously reported the usefulness of this technique for determining normal peritalar architecture (15, 17), and the present study now extends its potential applicability to quantifying foot deformity in RA patients. Pes planovalgus in RA is characterized by loss of the medial longitudinal arch, medial bulging of the talonavicular joint, and valgus deformity of the heel (1, 3, 5).

When these complex 3-D abnormalities are projected onto radiographs, visualization may provide insufficient diagnostic information (6), and measurement is often limited to 2-D sagittal plane angles that quantify the medial longitudinal arch profile (3, 5). With 3-D reconstructions, renditions such as those in Figure 2 provide superior visualization and enhance the qualitative interpretation of gross bony anatomy. In addition, the technique generates 24 architectural parameters from the reconstructed images based on an accurate and reproducible bone axis system defined by the centroid and principal axes of the bones (15–18). In the present study, the technique was useful because it quantified structural abnormalities at multiple joint sites from one 3-D pose. This provided a more complete description of the widespread architectural changes seen in the tarsal region of the RA foot. Furthermore, the number of abnormal parameters and the extent to which they deviated from normal gave a composite measurement of the severity of deformity.

To the best of our knowledge, this study is the first to provide a detailed 3-D description of tarsal joint structural derangement in RA. The data derived from the 3-D approach support and add to the findings from 2 previous radiographic studies that defined pes planovalgus in RA by increased lateral talocalcaneal and internal arch angles (3, 5). Conceptually, the principal axis angle formed between CT as measured by the MR reconstruction is a 3-D approximation of these 2 radiographic parameters. In the present study, ACT was significantly increased in the RA patients in comparison with normal. This abnormality was evident in the 3-D renditions and showed anterior and plantar drift of the talus between the calcaneus and navicular. This drift could be further quantified by the significant increase in distance between the geometric centroids of the calcaneus and navicular (dCN) and the decrease in the subtended angles for the talus–calcaneus–navicular bone triad (aTCN). The origin of this most striking abnormality is combined subtalar and talonavicular joint instability.

Synovitis at the subtalar joint was observed in 15 patients, frequently in the sinus tarsi region, where it may have progressively weakened the cervical, interosseous talocalcaneal, and superomedial calcaneonavicular ligaments. These ligaments are known to provide resistance to eversion motion at the subtalar joint, and mechanical failure may partly explain the displacement and change in orientation of the talus relative to the calcaneus and other tarsal bones (22, 23). A stable talonavicular joint may have resisted the displacement of the talus. However, the distance between the geometric centroids of the calcaneus and navicular was significantly increased, suggesting that the inferior calcaneonavicular ligament, which supports the head of the talus, was structurally weakened. This is possible, because the joint was inflamed in 12 of the 23 patients. Loss of medial arch height and rearfoot deformity have been reported following release of the tibialis posterior (24, 25). Although no cases of tendon rupture were reported here, tenosynovitis was present in 9 patients, and muscle weakness may have further contributed to talonavicular and subtalar joint instability.

Geometric analysis of the 3-D reconstructions identified further orientation and distance abnormalities, highlighting the widespread changes throughout the tarsal complex. Of the 4 tarsal bones, the cuboid is the most plantar, and talar drift caused closer apposition of these 2 bones, as determined by a significant decrease in dcT coupled with an increased angle between the principal axes (AcT). The presence of synovitis at the calcaneocuboid joint in 12 of 23 patients, with subsequent weakening to the inferior calcaneocuboid and dorsolateral ligaments, may explain a flattening along the lateral aspect of the foot, because there was a significant increase in the angle formed between the principal axes of the calcaneus and cuboid (ACc) and closer apposition of the 2 bones. These features have not been previously reported.

Alternative 2-D techniques such as footprint analysis use increased weight-bearing in the midfoot as a proxy measure of transverse arch collapse associated with pes planovalgus (26). Using 3-D reconstruction, the planes formed between the lines connecting the geometric centroids of the tarsal bones in a medial-to-lateral direction are inclined more transversely and serve as a direct measure of transverse architecture. In the RA patients, the subtended angles for the bone triplets aCcN, aCNc, aCcT, aTNc, and aNTc were significantly different from normal. The significance of these abnormalities has yet to be fully understood, because the patterns were subtly different between cases and were not easy to differentiate on visualization, although the gross appearance on visualization of the 3-D renditions confirmed loss of transverse arch and midfoot widening.

Fitting individual RA foot profiles to predetermined foot types using pattern classification techniques was moderately successful. The technique classified ∼50% of the RA patients with abnormal foot architecture profiles. Based on our clinical understanding of the RA foot, the assumption was that these cases might fit well to an overall RA and/or pes planus profile. Although 8 of 12 patients had some of the parameters for the RA foot profile, no patient met the classification threshold of Fg ≥0.5. For pes planus, all RA patients met some parameters, but only 3 patients fulfilled the classification threshold. Surprisingly, 5 patients matched the abnormal and pes cavus profiles, and 2 of these were judged to have severe pes planovalgus deformity on clinical inspection. Classification to specific foot types lacks specificity, because the profiles for the pathologic pes cavus and pes planovalgus foot types were based on too few cases (17). In its current form, the technique should be limited to differentiating normal from abnormal foot architecture in RA patients.

The exact mechanisms leading to pes planovalgus in RA have yet to be elucidated, although several pathways linking inflammatory and biomechanical factors have been proposed (2, 5, 7). Reports suggest that the normal motion-guiding and stability properties of ligaments and tendons surrounding the tarsal joints are diminished when inflamed joints are subjected to deforming loads on weight-bearing (3, 5). Several in vitro studies have successfully modeled pes planovalgus by transecting tarsal ligaments and medial ankle tendons, yet there are no studies modeling capsule, ligament, or tendon stretching or degeneration (27, 28). Of note in the present study was the presence of tarsal inflammation in most patients despite aggressive DMARD therapy. However, in the regression model, only subtalar joint inflammation was predictive of abnormal tarsal architecture. Further work is required to elucidate the interactive role of inflammation and abnormal joint structure, and future regression models, with larger numbers of patients, should include important factors not entered here, particularly biomechanical variables. In addition, the concept that MR-based reconstruction techniques could be used to predict risk of advanced foot deformity in RA requires formal evaluation in controlled trials. Such trials would determine how deformity advances and the role of various inflammatory and biomechanical factors, and whether therapeutic interventions prevent the development of painful deformity. Preliminary reports suggest 3-D reconstructions are valuable for planning and evaluating pes planovalgus surgery in orthopedics, and this role could be expanded to the surgical management of complex foot deformity in RA (17, 18).

In conclusion, 3-D MRI reconstruction and analysis techniques have been used to describe and evaluate abnormal subtalar and midtarsal joint architecture in patients with RA. Reconstructing these joints in 3-D and quantifying the geometric relationship between the constituent bones may be useful diagnostically and for planning logical structurally based corrective treatment.


We would like to acknowledge the support of Brian Whitham and Mike Pullan for constructing the MRI foot jig and Dewey Odhner and Tad Iwanga for assisting with the 3DVIEWNIX operation.