KeywordsSubtalar jointInstabilityWeight bearing CTImaging
Chronic hindfoot instability is a frequent problem that is evident in up to 33% of patients with a history of ankle sprains . Hindfoot instability often includes the ankle joint but can also affect the subtalar joint [2–4]. While ankle joint instability can be diagnosed clinically, an accurate assessment of the subtalar joint remains elusive [2, 3, 5]. To provide an adequate treatment for patients with post-traumatic hindfoot instability, a meaningful assessment of the subtalar joint is desirable .
Two ligaments are the primary stabilizers of the subtalar joint: the interosseous talocalcaneal ligament (ITCL) and the calcaneofibular ligament (CFL) [6–9]. Other ligaments providing subtalar joint stability include the cervical ligament, lateral talocalcaneal ligament, and deltoid ligament [6–8, 10, 11]. Recent cadaver studies showed that the CFL is potentially the most important stabilizer of the subtalar joint when subjected to inversion and external rotation stress [7, 8]. As the CFL crosses both the ankle and subtalar joint, the stability of both joints is affected after injury. In contrast, the ITCL only provides stability to the subtalar joint .
Variation in injury patterns and its complex anatomy make diagnosing subtalar joint instability particularly challenging [2, 12]. Long-lasting instability of the lateral ligament complex results in degenerative changes and chronic hindfoot pain [2, 12–14]. The impact of subtalar joint instability on this development remains unclear . This emphasizes the relevance of a radiographic diagnosis, currently performed by several two-dimensional (2D) measurements including tibiotalar tilt (TT), anterior talar translation (ATT), and subtalar tilt (STT) [2, 4, 16–21]. However, these measurements are limited in their ability to identify subtalar joint instability when using stress radiographs [2, 4].
While conventional radiographs are limited in assessing the subtalar joint, weight bearing computed tomography (CT) scans have demonstrated emerging diagnostic applications as they offer an accurate representation of hindfoot joint alignment under weight bearing conditions [22–25]. However, the clinical use of this imaging modality to diagnose subtalar joint instability has yet to be investigated. We hypothesized that isolated subtalar joint instability can accurately be diagnosed when using weight bearing CT scans in a cadaver model.
Data Source and Specimens
Seven pairs of fresh frozen male cadavers (tibial plateau to toe-tip) were included (mean age 63 ± 5 [range 54–69] years; mean weight 77.2 ± 6.9 [range 68.7–90.7] kg; mean BMI 24.1 ± 1.3 [range 22.2–25.7] kg/m2). Inclusion criteria were 20 to 70 years of age and a body mass index (BMI) of less than 35 kg/m2. Only male cadavers were included to ensure a homogeneous cohort. Exclusion criteria were a history of foot and ankle injuries or previous foot and ankle surgery.
Non-weight bearing and weight bearing (85 kg; determined from the average of specimen donor anthropometrics) CT scans with and without application of 10 Newton meter (Nm) internal torque applied at the Ilizarov apparatus (corresponding to external torque of the foot and ankle) were collected . Ten Nm torque was chosen for consistency with cadaver studies testing the stability of the distal tibial syndesmosis [27, 28]. Preconditioning of the specimen was performed by consistent loading of the frame with 42.5 kg and 85 kg for 2 minutes each before experimentation.
Imaging and Measurements
Digitally reconstructed radiographs (DRRs) were automatically created using the CT scan dataset (CurveBeam LLC, Warrington, USA). The anteroposterior (AP) view of the ankle joint was generated perpendicular to a line connecting the center of the calcaneus (midway between the medial and lateral process of the tuber calcanei) and the second metatarsal base on a dorsoplantar (DP) view . By virtually externally rotating the foot 90 degrees, a lateral view was generated. Additionally, a 30/40-degree Broden view was reconstructed (30 degrees internal rotation and 40 degrees upward tilt of the foot) . The TT and ATT were measured on the AP and the lateral view, while the STT was assessed on the 30/40-degree Broden view [2, 4, 19–21]. TT and ATT measurements were performed to additionally evaluate the effect of ligament transection (CFL and deltoid ligament) on ankle joint congruency.
Intraclass correlation (ICC) was used to quantify the agreement of measurements between and within observers. Estimates and 95% confidence intervals (CI) were calculated for each type of measurement within each view. Interobserver agreement was modeled with a two-way random effect model of absolute agreement with a single measurement per observation. Intraobserver agreement was modeled with a two-way mixed effect model of consistency with a single measurement per observation. Agreement was rated as very good with an ICC > 0.80; good with an ICC = 0.61–0.80; moderate with an ICC = 0.41–0.60; fair with an ICC = 0.21–0.40; and poor with an ICC < 0.20 . Measurements for interobserver agreement calculation were done by a fellowship trained orthopedic surgeon and a research analyst. For calculation of the intraobserver agreement, measurements were performed two times with an interval of 3 weeks by a fellowship trained orthopedic surgeon.
Linear mixed effect models were fit for responses. Within DRR and CT measurements, separate models were fit for each measurement (TT, ATT, and STT) and, for DRR measurements, within each view. Cadaver, a random effect, foot (left or right), and a fixed effect were included in all models in addition to the variables presented. Models were fit for subsets of the data, and estimates and 95% CI were given for differences in measurements in different levels of a specific variable. For each model, only the differences in response that were associated with either different load application, different torques, or different conditions were calculated; the data was subset by the other two variables, and they remained constant within each model. The first set of models compared the differences in response for full versus non-weight bearing with no torque applied (condition constant within each model). The second set compared the differences in response to 10 Nm versus 0 Nm of torque applied with full weight bearing load (condition constant within each model). The last set compared the differences between Conditions 1 through 3 and the native ankles, with full weight bearing load (torque constant within each model). Coefficients and 95% confidence intervals were reported, and statistical significance (marked by an asterisk in all tables and graphs) was determined based on a P value less than 0.05. All calculations were done in R 3.4.1, specifically using packages psych and lmerTest.
Digitally Reconstructed Radiographs
Reliability of digitally reconstructed radiographs measurements assessed by intraclass correlation (ICC) 
Interobserver: ICC(2,1) estimate (95% CI)
Intraobserver: ICC(3,1) estimate (95% CI)
Mean difference for each condition with and without load application (no torque) 
Estimate (95% CI)