Biomechanics of the natural, arthritic, and replaced human ankle joint

Experimental in-vitro work was performed by the present authors explicitly to investigate whether or not a preferred path of joint motion is prescribed by the passive joint structures alone during dorsi- plantar-flexion in virtually unloaded conditions [17]. This is fundamental knowledge for any design of relevant surgical treatments. A rig was built to move the ankle complex through its full range of flexion while applying a minimum load. Joint motion was constrained therefore only by the articular surfaces and the ligaments.

The movements of the calcaneus, talus and fibula relative to the stationary tibia in lower-leg preparations were tracked with a stereophotogrammetric system. It was shown that for each individual specimen, the calcaneus follows a unique path of unresisted coupled motion relative to the tibia and that most of this motion occurred at the ankle, with little motion at the subtalar level. The CaFi and the TiCa ligaments showed near-isometric pattern of rotations about their origins and insertions, whereas posterior ligaments slackened during plantarflexion and anterior ligaments slackened during dorsiflexion. In other words, during virtually unloaded joint movement, there are ligament fibres that maintain constant length throughout movement, and this must guide joint mobility, and others that tighten to define only the extremes of this movement.

All specimens showed motion of the axis of rotation relative to the bones. Perturbations from this unique path of passive motion induced by the application of load involved mostly subtalar joint motion and were resisted. The perturbations were completely recovered when the loads were removed, and the joint returned to its unique path of passive motion. Therefore, the ankle complex exhibits one degree of unresisted freedom. The subtalar joint complex behaves as a flexible structure which moves only because of soft tissue deformation when loaded [18]. Further experiments with higher resolution techniques, i.e. combination of roentgen-stereophotogrammetry and 3D digitisation, showed that the most anterior fibres within the CaFi and TiCa rotate most isometrically [20], i.e. experience the smallest strain, and that an anterior translation of the articular contact on the tibial mortise occurs during dorsiflexion [19]. It was deduced that the ankle is a single degree-of-freedom mechanism where passive mobility is allowed by rolling as well as sliding of the articular surfaces upon each other and by the isometric rotation of the two ligament fibres about their origins and insertions, therefore without major deformation of these tissue. In other words, in the absence of tissue deformation, passive motion is unresisted.

Computer-based geometrical models [21] elucidated this mechanism, initially in the sagittal plane (Figure 2), where most of the passive motion was shown to occur. Fibres within the CaFi and TiCa ligaments rotate isometrically about their origins and insertions (a four-bar-linkage when projected in the sagittal plane), while all other ligament fibres located more anteriorly slacken during dorsiflexion, those located posteriorly slacken during plantarflexion (Figures 2 and 3), becoming tight just at the corresponding extremes.

The instantaneous centre of rotation (IC), the point at which the two isometric ligament fibres cross in the plane, moves from a postero-distal to an antero-proximal position during dorsiflexion. The articular contact point, depicted in Figure 2 by the common normal (CN), moves from the posterior part of the tibial mortise in maximal plantarflexion to the anterior part in maximal dorsiflexion so that the talus rolls forwards while sliding backwards on the tibial mortise during dorsiflexion, and it rolls backwards while sliding forwards on the mortise during plantarflexion.

The shapes of the articular surfaces must be compatible with this ligament rotation, i.e. articular surfaces must move in contact with one another while maintaining these fibres just tight at constant length. The deduced shape of the complementary surface of the talus, compatible with a mortise shape taken as an arc of a circle, is a polycentric and polyradial curve as in the intact talus.

The model was then extended by including arrays of fibres for each ligament (Figure 3). The mechanical effect of the extensor retinaculae was included to predict the changing lever arm lengths of the main flexor and extensor muscles [22], calculated as the perpendicular distances from the IC to each tendon. Figure 3 shows that the changing positions of both muscle lines of action and of the instantaneous centre of rotation produce a lever arm of the flexor muscles maximised in dorsiflexion, and that of the extensor muscle maximised in plantarflexion. The joint positions in which these two muscle groups fire during gait are exactly those in which they were predicted to be mechanically advantaged.

Three-dimensional (3D) computer-based geometrical kinematic models of the tibiotalar articulation were subsequently developed to explain the multi-axial coupled motion observed experimentally during passive motion [23, 24]. Two one-degree-of-freedom spatial equivalent mechanisms for the tibiotalar joint passive motion simulation were initially proposed [23]. The mechanisms were based on the assumption of the guiding role of the joint passive structures, such as ligaments and articular surfaces, and on their geometric dimensions. These assumed isometricity of fibres within the calcaneofibular and tibiocalcaneal ligaments and rigidity of the articulating surfaces, taken as three sphere-plane contacts in one model (Figure 4), and as a single spherical pair in the other one.

Although motion predicted by the models was reasonably compatible with that measured in corresponding specimens, considerable differences were observed. This was accounted for by the oversimplifications adopted to represent the anatomical structures, particularly the complex articular surfaces in spheres and planes. Later, a surgical navigation system with cluster of active markers was used to collect more precise skeletal kinematics and anatomical geometry of the passive structures, i.e. articular surfaces and attachment areas of the ligaments, by digitisation with a pointer [24]. An equivalent spatial mechanism for the passive motion simulation was defined by three sphere-to-sphere contact points and two rigid links. These contact points were identified at the lateral talofibular articulation and two at the articulation between tibial mortise and trochlea tali. The two rigid links were identified by the isometric fibres at the calcaneofibular and tibiocalcaneal ligaments. An optimisation algorithm was developed for the identification of the final geometrical parameters resulting from an iterative refining process, which targets best matching between model predictions and corresponding experimental measurements of the spatial motion.

The specimen-specific equivalent spatial mechanisms replicated the original passive motion from corresponding specimens very well. The study demonstrates further that the articular surfaces and the ligaments, acting together as a mechanism, control the passive kinematics of the ankle joint in a complex 3D path of motion. In particular, it was demonstrated that the passive structures of the ankle joint alone are able to guide the complex triplanar motion, where the about 45° flexion in the sagittal plane is coupled with about 4-5° and 7-8° rotations respectively in the frontal and transverse planes.

During activities of daily living, the ankle functions under load. In response, ligaments stretch or slacken and articular surfaces in contact indent. The passive motion models here above define the initial configuration of these joint structures at each flexion angle, from which the final configuration under load can be calculated incrementally, as described in the following paragraphs.

Experimental and modelling work on fibre recruitment, in both 2D and 3D, provides information to interpret elementary mechanical tests routinely used for clinical assessments, in particular the anterior drawer test. At each joint position within the flexion arc, the ligament structures which resist the external force change not only orientation, but also the thickness, because of the progressive recruitment and tightening of fibres (Figure 3; see also [20]. This is one of the possible explanations of the observation that the resistance to load of the natural ankle varies along its flexion arc [26‐28].

Mathematical models of the ankle joint were developed to study ligament fibre recruitment and to calculate relevant load/displacement curves at different flexion angles within the passive flexion arc [29, 30]. Ligaments were modelled as 3D arrays of fibres, though their orientations at different flexion angles were taken from the four-bar-linkage model in the sagittal plane [21]. A non-linear stress/strain relationship was assumed for ligament fibres and relevant mechanical parameters were taken from the literature. Talus and calcaneus were assumed to move as a single rigid body. The antero/distal translational motion of the talus relative to the tibia was calculated.

The ankle joint was found to be stiffer at the two extremes of the flexion range, and the highest laxity was found around the neutral position, confirming previous experimental works. In a first paper [29] the quantitative comparison between model predictions and experimental measurements was not fully encouraging, because of the elementary nature of the datasets used for the mechanical parameters of the ligaments. In a second paper [30], the anterior drawer test was assessed also considering the effect of ligament viscoelasticity on the force response of the ankle joint, and a third data set [31]. The stiffness of the model ankle joint increased only modestly with velocity. The response force found for a 6 mm displacement at plantarflexion increased by only 13% for a one hundred-fold increase in velocity from 0.1 to 10 mm/s. The model predictions agreed well with the same experimental results cited above. The flexion angle was confirmed as the most influential parameter in the mechanical response of the ankle to anterior drawer test, supporting further the view that the comprehension of joint mobility is a necessary prerequisite for the comprehension of joint stability.

Specific clinical applications of multi-segmental foot models have, in particular, included pathologic gait characterization in rheumatoid arthritis (RA), posterior tibial tendon dysfunction, and hallux rigidus [40]. Woodburn et al. [47] showed that, in these patients, painful valgus deformity of the rearfoot is associated with excessive eversion at the ankle complex and internal rotation of the shank, when walking barefoot and also in shoes. The effect of RA at the forefoot was described by Khazzam et al. [60] by a multi-segmental foot model, supported by anterior-posterior, lateral, and modified coronal radiographs to relate marker position to underlying bony anatomy. As compared to a control population, the RA group showed prolonged stance time, shortened stride length, increased cadence, and a slower walking speed; at the rearfoot, they found delayed and decreased plantarflexion, increased external rotation and increased inversion, in contrast with the previous observations. Turner and Woodburn [61] described RA patients with severe forefoot, rearfoot or combined deformities, and reported decreased plantarflexion in terminal stance and increased eversion at the rearfoot. In particular, they identified different characteristic gait patterns among those groups of patients.

A few papers have compared gait before and after ankle arthroplasty using standard gait analysis, i.e. where the foot is limited to a single rigid segment. Little information is available about gait in ankle osteoarthritis (OA). Two recent studies have described ankle kinematics and kinetics before and after total ankle replacements (TAR) [57, 62], and have implied therefore quantitative assessment of gait in arthritic ankles. In the former, with respect to 15 age-/gender-matched control subjects, 15 unilateral post-traumatic ankle OA patients showed a decrease of the second active maximal vertical and the maximal medial ground reaction force, and, at the ankle joint, a decrease of the tri-planar movement, a reduction of the sagittal and transverse moments, a reduction of the power. In the latter paper, 9 patients treated for post-traumatic and 1 for psoriatic arthritis were assessed pre-operatively and at 6 and 12 months follow-up. With respect to the control group as reported at the beginning of 2.2, range of rotation of the foot with respect to the shank in the sagittal, frontal and transverse planes were respectively 11 (30 in normal ankles), 10 (14) and 12 (22) degrees. In both studies, the extent to which the low performance in ankle OA is affected by pain and difficulty in progression is not known, but is demonstrated by low clinical scores and considerable deficiency in most of the spatiotemporal parameters.

Gait analysis after ankle arthrodesis has been reported only by using a single foot segment model, thus describing, according to the specific marker set, the overall foot-to-shank motion, which includes the confounding effect of foot deformation and the undesirable skin motion artefact [63, 64]. Very different motion patterns are expected when either the tibio-talocalcanear arthrodesis (or triple, with intra-medullary nail for a combined ankle and subtalar arthritis) or the isolated tibio-talar fusion (with surgical screws and plates) are performed.

Only two studies were able to distinguish between rear- and fore-foot motion [65, 66], although the former paper reported from ten patients only, with no information as to surgical technique and with a large spectrum of follow-up (0.5 – 4 years). In general, the reduced motion at the ankle complex was found to be compensated for by increased motion at the knee and at the more distal foot joints. Significant increase of motion was found radiographically at the subtalar joints in one study [67], stiffness and loss of motion in another [66]. The compensatory hypermobility at the subtalar and midfoot joints is deemed responsible for increased stress at these joints [63, 66, 67].

Since the early 1970s, TAR has been considered a possible alternative for the treatment of severe erosions of the articular surfaces of the ankle, mainly because arthrodesis can result in high incidence of non-union, secondary degenerative changes at neighbouring joints, high incidence of postoperative infection, and total loss of motion [68]. The improving survivorship of ankle replacements and the potential benefits of restoring movement, improving gait and protecting adjacent joints are recent persuasive arguments in favour of arthroplasty [69].

The effect of arthrodesis and arthroplasty, with three different current designs, on the arthritic ankle was analysed preliminary in-vitro [70‐72], showing that total replacements changed the natural motion at the ankle joint complex less than arthrodesis, which reduced considerably the range in all three planes as expected. A two-part prosthesis restricted talar motion within the ankle mortise much more than the two three-part designs, likely resulting in an increase of stress forces within and around the prosthesis, potentially leading to polyethylene wear and loosening at the bone-implant interfaces.

In-vivo, gait analysis was performed in a few recent studies. Piriou et al. [69] analysed 12 patients before and after ankle arthroplasty, and compared these to 12 patients after successful ankle arthrodesis and to a healthy control group of 12 subjects. Although neither arthroplasty nor arthrodesis restored normal walking speed or lower limb movements, the former group after arthroplasty had greater motion at the ankle, a symmetrical timing of gait and restored ground reaction force patterns, whereas ankle arthrodesis resulted only in a faster gait with a longer step length compared to arthroplasty.

The two cited studies on kinematics and kinetics analyses in arthritic ankles [57, 62], reported on these patients also after TAR. The former described a worsening of gait at three months follow-up, but spatiotemporal variables not different from the normal subjects at 12 months follow-up; however, in six of the eleven kinematic and kinetic variables analysed there was only a partial restoration. In the latter, gait analysis, together with the AOFAS clinical scoring system, was performed at 6 and 12 months from surgery. The function sub-score and the spatio-temporal parameters improved considerably already at 6 months. More normal patterns and ranges of rotations and moments were observed in all the three anatomical planes of the replaced ankle joint at 6 months, and maintained at 12 months. Electromyography revealed also a good recovery of physiological muscle activity. These studies demonstrated that ankle prosthesis can produce an early functional recovery.

Compared to the pre-surgery condition, increased motion at the hip and knee joints, and in ankle flexion moment and power were also observed, at a mean follow-up of four years [73]. Compared to the contralateral non-operated ankle at one year follow-up, several differences were still noted, but nearly physiological motion and loading were observed in the replaced ankle though limited to the stair climbing task [74]. However deterioration of the spatial-temporal parameters and abnormal muscular activation have also been noted at longer follow-up [75].

In explicit comparisons between the two ankle treatments, significantly larger improvements in foot mobility were found after arthroplasty, as expected, with in addition several significant impairments remaining after arthrodesis [48, 69, 76].

In summary, in-vivo gait analysis showed that although neither arthroplasty nor arthrodesis restored fully normal walking speed or lower limb joint movements, the former allows larger motion at the ankle complex, a symmetrical gait and normal ground reaction force patterns [48, 69, 73, 76], though patients with arthrodesis had faster gait and longer step length [69].

Total joint replacement must address an original dilemma [3]. When the main target of the designers is the restoration of normal mobility, in terms of patterns and ranges of 3D motion, unconforming, semi-constrained designs (Figure 10) are sought because these allow for the necessary freedom of joint motion; however, this requires incongruent contact with attendant inadequate load-bearing capacity, high contact stresses and eventually high wear rates. On the other hand, when the main target is congruency of the artificial surfaces, full conforming designs are sought, which produce large contact areas which minimises the risk of polyethylene wear, but tend to constrain motion and overload the fixation system. The current generation of three-part TAR designs are the only apparent solution to this dilemma, because the articulating surfaces have conforming shapes, but how the relative motion is guided by the remaining passive and active structures is unknown [78]. Solutions must be sought to guarantee full congruity at the artificial surfaces through the arc of physiological passive flexion.

The implants least susceptible to wear can be completely congruent (or nearly so) two-part devices or three-part designs [81, 90] with a meniscal bearing in between the two metal bone-anchored components (Figure 11). The two-part devices require a thick layer of polyethylene typically attached to the tibial component.

The three-part designs employs fully-congruent meniscal bearings free to slide on both the articular surfaces of the component fixed to the bones. A meniscal bearing prosthesis can allow translational movement and yet maintain congruence of all the articular surfaces throughout the range of movements. One of the two bone-anchored components must have constant radius to allow fully-congruent contact with the meniscal bearing in all joint positions. To also allow translational movement, the other component should have a flattened surface, although slightly concave or convex fully-congruent surfaces can also be used. Flat tibial components can experience only compressive force assuming no friction, and therefore these would transmit only compressive stress to the bone-implant interface, and would not need a robust intramedullary stem for fixation to the bone. A polyethylene meniscal bearing component is inserted in between, with the articulating surfaces fully-congruent to those of the metal bony-anchored components. It is free to translate in any direction to accomplish the relative movements of the two components as guided and constrained by the passive and active structures at the joint (Figure 12). Dislocation of the meniscal bearing is resisted by the interpenetration of the convex bony-anchored component into the concavity of the bearing, in which it is held by the tension in the joint’s ligaments. A spherical interface between the convex bony-anchored component and the meniscal bearing has also the advantage of maintaining congruence also in transverse and frontal plane rotations. Unlike a cylindrical interface, the spherical interface can also accommodate for slight inaccuracies of implantation. Intact retained ligaments can be restored to their original normal tensioning pattern by the choice of an appropriate thickness of the bearing component. As in the natural joint, as shown above, where articular surfaces alone do not guide the reciprocal movements of the bones but merely allow them, in meniscal bearing replacement the unconstrained components perform in the same way as guided by the ligamentous mechanism.

All these considerations emphasise the importance of the intact status of the ligamentous structures in any ankle meniscal bearing replacement. The implantation of freely mobile bearings into joints which lack an intact and functioning ligamentous apparatus is theoretically mistaken, and has proven to be unsatisfactory in practice also for knee replacements [91]. It is irrational to build into a prosthesis the freedom to translate in the absence of the mechanism which controls that freedom.

Currently most of the TAR designs in clinical use have fully conforming mobile bearings [89], and only apparently these represent correct compromises between mobility and conformity. These are claimed to be anatomical, but all feature a flat shape of the tibial component, very unnatural, in addition to the natural anatomical talus. These must rely fully on ligaments for final joint stability, but unfortunately the functioning of the ligaments was not considered explicitly in the design.

In addition to replication of original joint function, i.e. mobility and stability, it is also necessary to achieve implantability and durability in TAR. The reliability and repeatability of the operative technique is considered by the surgeons as a fundamental characteristic for a TAR design. Relevant instrumentation must be robust and accurate enough for guaranteeing the correct position of the components with the minimum bone stock removal. Durability is also dependent on good fixation of the components, which would involve an appropriate load transfer to the bone and a minimum risk of loosening. The current designs show a large variety of fixation elements. Pegs, long or short stems and cylindrical or rectangular bars have been used [92]. More recent designs use bone screws [93, 94].

As far as the materials are discussed, moving from the original tibial components made in polyethylene, most of the recent two-part designs include a metal-backed tibial component. The design of the elements used to limit the floating of the bearing core is then an additional critical issue. Entrapment of the meniscal bearing in some prostheses is enforced by sharp limiting interfaces, to prevent dislocation and separation. Ribs and grooves, lugs and cutouts, and even systems of interlockable flanged grooves have been devised for this purpose [83, 84, 94]. These latter prostheses may be at high risk of polyethylene wear through contact at these interfaces.