Due to the scarcity of valid data and the difficulties in accomplishing a prospective randomized trial due to the strong opinions held by treating experts [21], evidence-based recommendations for treatment are lacking [21]. However, during the last decade various expert-based algorithms of treatment have been proposed [24, 26, 29]. Traditionally, the Eichenholtz stage relying on clinical and standard radiographic examination [11] and the site of the disease [32] provide the key determinants for selecting the primary therapeutic strategy [7, 21]. Nonoperative immobilization techniques using specific orthoses or total contact casts still represent the mainstay for the initial phases of Charcot neuroarthropathy, while surgery is reserved for patients with infection, recurrent ulceration, and substantial deformity or joint instability not manageable by casting or orthotic devices [3, 20, 29]. As most Charcot feet manifest in the midfoot, the majority of studies describe lesions in the midfoot area [8, 10, 16, 18, 33, 34, 36]. However, the success of total contact casting even in the hands of a single experienced physician may result in subsequent ulcerations in 30% of the patients during the treatment [13]. Recurrence of ulceration means prolonged disability and the requirement for recasting, long-term use of ankle-foot orthoses, or possible secondary surgery [29]. In particular, unstable deformities in these frequently obese patients are difficult to brace with a high inherent risk of ulceration in the insensate feet [19, 21]. Late institution of immobilization and unloading runs a high risk of permanent deformity after healing [6, 14]. An underlying fixed deformity or osseous prominence is one of the main reasons for recurrence of ulceration [35]. As such, more than 40% of the patients with midfoot disease who had been managed primarily nonoperatively may require surgery in the further clinical course [21, 23]. While formerly corrective arthrodesis of nonplantigrade neuroarthropathic feet has been regarded as a salvage procedure and an alternative to amputation [20], more recent reports relying on biomechanically sound stabilization techniques have demonstrated maintenance of walking independence using commercially available therapeutic footwear and an infection-free foot [26] despite a high complication rate varying from 10% to more than 30% [3, 19]. One study of 14 patients with Stage I Charcot arthropathy at the midfoot level reported successful healing of the primary corrective arthrodesis without immediate or long-term complications or any recurrent ulceration during a mean followup period of 41 months [35].

Rearfoot and ankle Charcot deformities, though less common than midfoot disease, tend to be more unstable than midfoot disease and more likely to undergo subsequent surgery [5, 36]. Identification of patients at-risk who already present with manifest ulceration or who are more likely to have skin ulceration over plantar bony prominences due to a nonplantigrade deformity or severe instability appears beneficial for timely decision-making in favor of surgical reconstruction [24, 26]. Early arthrodesis in the very unstable cases with a reduced period of total contact casting reportedly provides comparable results to patients treated nonoperatively [5, 36].

The vascular status of each patient was assessed clinically by palpation of peripheral pulses. In patients with major soft tissue swelling, clinical examination was supplemented by color-coded duplex sonography. We routinely obtained weight-bearing radiographs of both ankles and feet including Saltzman views [28] of the hindfoot. CT scanning with multiplanar reconstruction was obtained in two cases where complex deformity or major osseous defects were visible on standard radiographs and the surgeons wanted better preoperative spatial imaging for planning of the reconstruction.

The indications for surgery included (1) a nonplantigrade foot position; (2) a high degree of hindfoot or midfoot instability not amenable to bracing or total contact casting based on expert opinion (KK, TM); and (3) manifest or impending ulcerations at bony prominences due to noncompensated deformity with nonlinear talar-first metatarsal axes on lateral (Fig. 1C) and anterior-posterior weight-bearing radiographs (Fig. 1D). The mean duration from the first presentation of the patient to the surgical procedure was 1.7 months (range 0.5–5 months).

We attempted to limit the extent of corrective arthrodesis and the number of surgical incisions to the minimum needed to achieve a plantigrade and stable foot position. With isolated midfoot neuroarthropathy (n = 9), we maintained the talonavicular joint and with hindfoot involvement we performed triple arthrodesis (n = 5) (Table 1). When we performed arthrodesis of the ankle and the subtalar joint (n = 2) or tibiocalcaneal arthrodesis (n = 3), the talonavicular and the calcaneocuboid joints were preserved in four of five feet to avoid the generation of a completely stiff foot. A plantigrade foot position was achieved either by osteotomy or excision of bone at the apex of the deformity or by resection of necrotic and destroyed bone and reorientation of the foot axes. Residual osseous defects after obtaining a plantigrade foot position were filled with autologous iliac crest grafts. The corrected foot position was preliminarily maintained with K-wires. We simulated partial weight bearing intraoperatively by manually pressing a plate against the plantar surface and obtained fluoroscopy prior to definitive stabilization to verify an adequate reduction. We stabilized the midfoot or the Chopart joint either by noncannulated 7-mm fully threaded screws or by 7.3-mm cannulated screws supplemented by 3.5-mm plates. If the ankle was fused, we inserted a retrograde interlocking nail (n = 1) or a 4.5-mm cannulated blade plate (n = 3). Lengthening of the gastrocnemius-soleus complex either by open gastrocnemius recession or by percutaneous Achilles tendon lengthening was an integral procedure in the 11 cases with an equinus deformity. We locally débrided ulcers in eight feet and covered them with dry postoperative dressings for secondary healing.

Postoperatively, we applied a nonweight-bearing total contact cast. We reviewed patients and obtained radiographs at 6 weeks and every 3 to 4 weeks until osseous healing. Casts were routinely replaced every second or third week until signs of osseous consolidation (bony bridges at the correction site, no signs of bone resorption around implants) with maintenance of the corrected foot position became visible on all three radiographic views. The mean postoperative casting period in a nonweight-bearing cast was 10.6 weeks (range 6–16 weeks). Partial weight bearing was allowed when the first signs of callus formation became visible on radiographs. Weight-bearing radiographs were performed prior to allowing full weight bearing in usually commercially available therapeutic footwear or customized orthopaedic shoes. The mean period from surgery to full weight bearing was 3.5 months (range 3–5 months).

Mean postoperative AOFAS scores were increased compared with preoperative values both in the hindfoot and the midfoot subgroup, respectively (Tables 2–5). Activity level and maximum walking distance rose in 19 patients and remained unchanged in three patients comparing the pre- and postoperative level.

We note several limitations of our study. First, knowing which degree of instability and deformity will be at high risk of failure with nonsurgical treatment relies on judgment [21, 23] and the risk factors are not well understood. Expert assessment of these key parameters represents the basis for most treatment protocols from centers with wide experience in Charcot feet [18, 19, 21, 23, 24, 29]. On the other hand, we believe decision making for surgery in Charcot feet on the basis of any given algorithm requires assessment by someone with considerable experience with Charcot feet and it should not be left to the inexperienced [21]. Second, our single cohort was heterogeneous, making generalizations more difficult. All major studies on the topic published during the last 15 years (Tables 6, ​,7)7) are retrospective with limited numbers of patients [27]. The authors are not aware of any prospective comparative study or any high-quality randomized controlled trial of primary nonoperative versus primary surgical treatment of Charcot feet. The need for such studies has been repeatedly mentioned [21, 27, 29] as has the difficulty in conducting such a trial owing to the heterogeneity of the deformities and instability [21]. Nevertheless, our data suggest a subset of patients with advanced instability and deformity benefit from primary surgical reconstruction comparable to patients who have secondary surgical reconstruction.

As such, in the present study with two-thirds of the feet having hindfoot manifestation of neuroosteoarthropathy and a high risk for a potential failure of nonoperative treatment none of the patients was assigned to primary nonoperative treatment. All our patients, whether they achieved bony or fibrous union, achieved a plantigrade and stable foot that remained ulcer-free during the observation period (Table 7). This compares quite favorably with the results after secondary surgery, where the recurrence rate of ulceration varied between 0% and 20% (Tables 6, ​,77).

Our patients accomplished full weight bearing within a mean of 3.5 months after surgery, which is longer than in nondiabetic patients but compares well with the only series of primary arthrodesis in Eichenholtz Stage I patients [35] and shorter than reported following secondary arthrodesis (4–7 months, Table 7). The high rate of patients being mobilized in accommodative shoewear after primary arthrodesis favorably compared to that of other series after secondary reconstruction where a substantial number of patients permanently use ankle-foot orthoses (Table 7). Functional outcome has rarely been analyzed in patients after surgical reconstruction of Charcot feet employing the AOFAS scores while most authors have expressed the functional result in a more descriptive manner (Table 7). A substantial increase in AOFAS scores occurred in our cohort within a relatively short time mainly due to functional improvement and better realignment both in our midfoot and hindfoot groups. The results exceed the AOFAS scores given for a limited series of 10 patients with mostly hindfoot involvement [36].

The deep infection rate after primary arthrodesis was also low compared with secondary arthrodesis where the infection rate ranged from 0% to more than 30% (Table 7). While almost one-third of our patients had preoperative ulcers in combination with instability or deformity (eight of 26 feet, Table 6), none of our patients had progressive deformity or subsequent amputation. All of these ulcers in our study healed uneventfully following corrective arthrodesis. In contrast, a number of authors advocate postponing surgery until a concomitant ulcer has definitely healed [8, 30, 31] or rely on alternative techniques of fixation as external fixators [24]. Despite the fact that an open wound may increase the risk for complications [4], we observed no major infectious complications. Other complications in our series were high when compared to studies with secondary intervention strategy [3, 10, 16, 17]. Hardware failure often coincides with the nonunion rate which was substantially high in our (23%) as in other series (0–30%, Table 7). Despite this, a stable fibrous nonunion in good position does not necessarily require surgery. Choice of implants or implant combinations was adapted to the underlying biomechanical requirements preferably using large-size internal implants. We believed none of the observed implant failures could directly be related to improper selection of implants. This even holds true for the single patient with chronic fistula who denied revision as she was satisfied with the functional level achieved.