The primary treatment of most ODs is arthroscopic debridement and microfracturing, with a success rate of 85% (van Dijk and van Bergen 2008, Zengerink et al. 2009). For failed primary treatment and large defects, there are various alternative treatment options, including cancellous bone grafting, osteochondral autograft transfer, and autologous chondrocyte implantation (van Bergen et al. 2008, Zengerink et al. 2009). However, each of these has disadvantages such as donor site morbidity, limited availability, and two-stage surgery (van Bergen et al. 2008, Paul et al. 2009).

A novel metal inlay implantation technique (HemiCAP) was developed in 2007 for medial defects after failed primary treatment (Figure 1). Its clinical goals are to reduce pain and prevent (further) cyst formation by resurfacing the OD. The talar implant set consists of 15 offset sizes to provide an implant that matches the geometry of the medial talar dome in a variety of talar specimens. A precise surgical technique is required in terms of implantation depth, position, and orientation because of the biomechanical properties of the ankle joint. A protruding implant may damage the opposite cartilage by causing excessive contact pressures during loading, which is thought to be due to “plowing” of the cartilage (Loening et al. 2000, Milentijevic and Torzilli 2005, Custers et al. 2007). On the other hand, a deep implant might result in collapse of the adjacent cartilage due to insufficient support (Jackson et al. 2001). Hence, an optimum middle course is aimed for. The talar dome cartilage thickness was found to be 1.42 (SD 0.18) mm in an in vivo study (Wan et al. 2008). Talar cartilage peak contact strain under weight-bearing conditions reached 35% (SD 7) (Wan et al. 2008). We therefore propose an optimal implantation depth of 0.5 mm (35% of 1.4 mm) below the level of the articular cartilage surface.

The mean implantation level was 0.45 (SD 0.18) mm (95% CI: 0.33–0.57) (Table 1). The lateral part of the implant, located on the talar dome, was measured to be recessed a mean of 0.57 mm (SD 0.16) mm. The medial part of the implant, located on the medial talar facet, was the least recessed (mean 0.03 (SD 0.40) mm). The central part of the implant, as assessed by visual inspection, was recessed in 10 specimens and protruding in 1 specimen (no. 11) (Figure 5).

During the post-implantation pressure measurements, one specimen (no. 2) fractured (comminutive intraarticular distal tibial fracture), and it was excluded from pressure analysis. In 2 specimens (nos. 3 and 11) the pre-implantation contact pressure was not measured, as the screw component had already been implanted during a pilot experiment.

Of the remaining 8 specimens, in all positions combined, the median pressure of the prosthetic area before implantation was 3.3% (0.02–18) relative to the total contact pressure. After implantation, the median prosthetic pressure decreased to 0.09% (0.02–13) in these 8 specimens. The highest median contact pressure was measured in the neutral position and the lowest in the plantarflexed position, both before and after implantation (Table 2).

There have been few studies dealing with the level of implantation of metallic implants and its effect on opposing cartilage. In a comparative rabbit study, Custers et al. (2007) showed that protruding implantation of metal plugs in the femoral condyle caused the most damage to the opposite tibial cartilage, and that flush implantation caused the least damage. Their deep implantations were below the level of the subchondral plate and therefore cannot be compared with the average implantation level in our study. Becher et al. (2008) compared implantation levels of a femoral condyle HemiCAP using human cadaver specimens. They concluded that protruding implantation led to substantially increased peak contact pressure, as also occurred in our single protruding implantation (specimen no. 11). However, the knee and ankle cannot be compared because the ankle joint is more congruent and its cartilage has different mechanical and biochemical properties (Shepherd and Seedhom 1999, Treppo et al. 2000).

The protruding central part of the implant in specimen no. 11 was due to an elevated screw rather than an inappropriate articular component. The level of the screw determines the central height, while the offset sizes of the articular component determine the peripheral height. It is thus crucial to implant the screw to a precise level, which is indicated by a mark on the screwdriver. The level of the implanted screw can be checked initially by placing a trial button, and adjusted if necessary. The availability of higher offset sizes allows for a recessed peripheral implantation even in the presence of a relatively high screw.

The implantation level of 0.5 mm below the cartilage surface aimed at in this study was based on the biomechanical properties of talar cartilage (Li et al. 2008, Wan et al. 2008). Li et al. (2008) measured a peak contact strain of 30% (SD 6.1) after 30 s of loading, using magnetic resonance and dual-orthogonal fluoroscopic imaging. Similarly, Wan et al. (2008) measured a peak contact strain of 35% (SD 7.3) in human subjects with a mean medial talar dome cartilage thickness of 1.42 (SD 0.31) mm. To compensate for the compressible property of cartilage, the incompressible metallic implant was aimed 0.5 mm (i.e. 35% of 1.42 mm) below the cartilage surface in our study. The average implantation level achieved, 0.45 mm below the level of the articular cartilage, corresponds to 31% (95% CI: 23–40) of 1.42 mm.

In the event that the implant would be non-weight bearing due to deep implantation, the remaining weight-bearing area of the talar dome has been shown to be capable of carrying the loads without any statistically significant alteration in contact pressure (Christensen et al. 1994). In particular, the periphery of the implant cap should be recessed because “edge-loading” may result in a poor outcome (Koh et al. 2006, Becher et al. 2008). The lateral part of the implant can be considered to be crucial because it is located on the weight-bearing talar dome articulating with the tibial plafond. Here, the implant was recessed in all cases (mean 0.57 mm) with little variability (SD 0.16 mm). In contrast, the medial part of the implant is located on an area of the talus that typically bears less weight, i.e. the medial talar facet. Hence, a protruding implantation on the medial side is possibly less harmful to the opposite cartilage of the medial malleolus.

Our study has limitations. We used cadaver specimens of elderly individuals (aged 66–89 years), whereas patients suffering from an OD are often young adults. The cartilage of our specimens may have undergone degenerative changes with aging. The ankle joint is not, however, prone to osteoarthritic changes during aging (Cole et al. 2003) and there is no correlation between the thickness of talar cartilage and age (Shepherd and Seedhom 1999). Furthermore, the specimens in our study were macroscopically graded: only those with intact cartilage were included.

Contact pressures were measured under static loading, while dynamic or cyclic loading may be advocated because of its supposedly better representation of real life. We applied static loading because more cartilage deformation can be obtained by one continuous load than by short-time loads with intervals (Herberhold et al. 1999). Because there is more cartilage deformation with static loading, any excessive contact pressure due to elevated implantation might be detected more adequately. This implies that, if tested under dynamic loading, a higher implantation level might be erroneously accepted. To imitate the stance phase of gait, the joint was loaded in three positions that correspond to the mean sagittal ankle motion with corresponding forces (Stauffer et al. 1977).

The loads were applied to the tibia, as was done in previous studies (Ramsey and Hamilton 1976, Macko et al. 1991, Lloyd et al. 2006). The real-life situation might have been better imitated by also loading the fibula, which transmits 7% of the total force through the lower leg (Goh et al. 1992). However, loading the fibula would probably not have had any notable effect on the results because the pressure distribution would have shifted slightly to the lateral side (Thordarson et al. 1997), with minor effects on the medially-located implant.

Variability between specimens is a common finding in cadaver experiments (Matricali et al. 2009). The contact pressure of the prosthetic area before implantation ranged from 0.02% to 18% in the cadavers we used (Table 2). This variability does not affect the conclusions, as each talus served as its own control.

Possible inaccuracies of Tekscan pressure-sensitive sensors are their sensitivity to moisture and temperature, wrinkling, and change of their position during loading (Svoboda et al. 2002, Becher et al. 2008). By testing in a single lab, any inaccuracy due to difference in moisture or temperature was minimized. To detect inaccuracy due to wrinkling, the sensor was inspected after a loading sequence and checked under the lever arm loading system. It was replaced in one specimen. Changes in sensor position were minimized by taking digital photographs at the first testing sequence, which we used for verification of positioning at the second sequence.

Our findings form a first basis for clinical use of the implant, but its clinical effectiveness remains to be investigated. Other designs of this prosthetic implant have been used clinically for different joint surfaces, including the first metatarsal head (Hasselman and Shields 2008), the femoral head (van Stralen et al. 2009), the humeral head (Uribe and Botto-van Bemden 2009), and the patella (Davidson and Rivenburgh 2008). Although the short-term clinical outcome of these designs is promising, it remains to be investigated whether the implant is effective for the treatment of ODs of the talus, in terms of reducing pain and preventing cyst formation.