Fractures of the clavicle are common and account for around 3% of all adult fractures [5]. The incidence ranges from 64 to 71/100,000 per year [10, 23]. Historically, midshaft clavicular fractures have been treated nonoperatively with good functional outcomes. A major complication of conservative therapy is the relatively high nonunion rate, which is reported to be 15–20% in some studies [32, 33].

In the past decade, randomized control trials have been conducted to compare operative and nonoperative therapy of midshaft clavicular fractures. A recent meta-study concludes that surgical treatment significantly reduces the nonunion rate, shortens the time to union, and, despite a slightly higher incidence of complications, leads to better shoulder functional scores in both short- and long-term follow-up [9]. Although other studies state that there is still no consensus on treatment, there has been an overall paradigm shift towards operative therapy with open reduction and plate fixation [18, 27, 29]. However, the problem of nonunion remains after both conservative and operative management.

Less is known about the results after revision of clavicular nonunions, regardless of whether the initial therapy was operative or nonoperative. Nonunion is either atrophic or hypertrophic. Most studies report a higher rate of atrophic nonunions (see Table 5). The surgical procedure usually includes a resection of the nonunion (decortication according to Judet) and (angular stable) plating combined with autologous bone grafting [20, 28]. Bony union can reliably be achieved with this technique [6, 25]. McKnight et al. compared the surgical management of midshaft clavicular nonunions with that of acute fractures and found an increased risk of short-term complications such as wound infections [21]. Long-term results are scarce. O’Conner et al. found a higher level of disability according to assessment with the Disabilities of Arm, Shoulder, and Hand (DASH) questionnaire in the operated group compared with the normal population after a mean follow-up of 42.1 months [24].

The aim of the present study was therefore to examine the clinical and radiographic results after operative revision of clavicular nonunions in a multicenter study. We hypothesized that the functional results would be similar to the noninjured side and that radiographic fracture union could be achieved in the majority of the cases.

Methods

Study design and patient cohort

Two level‑1 trauma centers in northwestern Germany conducted a retrospective analysis of patients with clavicular nonunion who needed revision surgery during the past 10 years. The study was approved by the local ethics committees (2019-159-f-S/8688_BO_S_2019). All patients who underwent operative revision for clavicular nonunion in these two hospitals were included. Patients with a record of pre-existing shoulder conditions (e.g., osteoarthritis), earlier operations at the injured side, or regular immunosuppressive medication were excluded. This was a two-center retrospective cohort study with a level III of evidence.

Outcome assessment

All patients underwent a structured clinical and radiological assessment. The clinical interview included a survey of the relevant epidemiological data, pre-existing illness, and individual course of treatment. Risk factors such as smoking, diabetes, and osteoporosis (when explicitly stated as a diagnosis in the patient reports) were also recorded. Pain was assessed using the Visual Analog Scale (VAS). The major outcome parameter was functional assessment using the age- and gender-adjusted Constant score according to Katolik [14].

For the radiographic follow-up, the major outcome parameter was fracture union on conventional X‑rays of the clavicular in two planes (anteroposterior [ap] and tangential). This was defined by three continuous cortices. Technical aspects, such as the plate type and length as well as usage of conventional and angular stable screws, were also recorded. Furthermore, the usage of autologous bone grafting was noted. This was harvested either from the iliac crest or from the resected hypertrophic nonunion itself (decorticated bone was cleaned from soft tissue, reduced to small chips, and used for bone grafting). Additionally, complications such as wound site infections with re-operation and the occurrence of paresthesia were examined.

Statistical analysis

Data were collected in a Microsoft Excel data sheet. Mean and standard deviations were calculated where applicable. Statistical analysis was performed with GraphPad (GraphPad Prism 8.3.0, San Diego, CA, USA). Welch’s t test and two-way ANOVA were used with Sidak’s multiple comparisons test and significance was set at p < 0.05.

Results

A total of 31 patients from the two centers were available for follow-up. No patients had to be excluded because of the exclusion criteria defined earlier.

Epidemiological data

The data of 31 patients were analyzed for this study. General epidemiologic data of the cohort are summarized in Table 1.

Table 1 Epidemiological data of the cohort

Initial treatment

The etiology was a mono-trauma of the clavicle in 28 (90.3%) patients, and only three (9.7%) patients suffered a polytrauma. In 18 (58.1%) patients the left side was injured; 14 of the 31 (45.1%) patients were treated in a primarily conservative manner, of whom four (28.6%) had to undergo surgery later than 42 days after trauma owing to secondary dislocation. For 17 patients, surgery was performed primarily within 6.6 ± 8.2 days (range: 0–35). The mean operation time was 53.3 ± 15.8 min (range: 37–80) but these data were only available for six of the operated patients. In summary, ten (32.6%) patients were treated conservatively prior to revision surgery while 21 (67.7%) patients underwent primary surgery or conversion to surgery. In seven (33.3%) surgically treated patients an angular stable plate was used, while an elastic stable intramedullary nail (ESIN) was inserted in three (13%) patients (no information on fixation material in 11 patients, 52.4%). The mean plate length was 7.6 ± 2.3 holes (range: 6–10).

Revision surgery

Data on revision surgery were available for all 31 patients. The mean time from trauma to revision was 19.9 ± 20 months (range: 4–107). The nonunion was classified as atrophic in 26 (83.9%) patients and hypertrophic in five (16.1%) patients. The mean radiographic distance of the nonunion gap was 10.9 ± 6.6 mm. A loss of reduction in the operated cohort was noted in 16 (76.2%) cases while an implant breakage was present in five (23.8%) patients. The number of previously conducted operations prior to the revision surgery in one of the two hospitals ranged from 1 to 3 with a mean of 1.38 ± 0.59 operations per patient. In nine (42.9%) cases the implants were already removed.

The mean operating time of the revision was 96.3 ± 35.9 min (range: 49–174). A straight angular stable plate was used in 19 cases (61.3%), while a pre-contoured plate was utilized in 8 (25.8%) cases. The mean plate length was 8.7 ± 1.4 holes with an average of 7.5 ± 1.2 screws per plate (4.3 ± 2.5 conventional screws/3.2 ± 3 locking screws). In 25 (80.6%) patients, autologous bone grafting was used, either from the iliac crest (54.8%) or from the resected nonunion (25.8%). Representative X‑rays and intraoperative images from a study patient are shown in Fig. 1.

Fig. 1
figure 1

Case of a 35-year-old male patient who had a mountain bike accident. Presentation 6 months after trauma to our department with two previous operations after medial implant loosening (a). Complete removal of the material was necessary after verification of a septic nonunion (b). After antibiotic therapy, surgical revision was performed (c). Intraoperatively, a 2-cm bony defect was seen (e). Accordingly, an autologous bone graft from the iliac crest (f) was combined with double plating (g). X‑rays 1 year postoperatively show complete union of the defect (d)

Clinical outcome

The clinical survey and assessment were conducted with a mean follow-up of 69.7 ± 31.2 months. Eight (25.8%) of the patients reported active smoking, while 17 (54.8%) stated no pre-existing illnesses. None of the patients had a history of osteoporosis or diabetes. The active range of motion is summarized in Table 2. In total, 29 (93.5%) patients could return to their previous work and 28 (90.3%) could return to sport. A total of 26 (83.9%) patients were able to conduct overhead work after revision surgery. Moreover, 27 (87.1%) of the patients answered the question: “Were you satisfied with revision surgery?” with “Yes” and the same number of patients would opt for revision surgery again.

Table 2 Mean values of active range of motion

Pain at rest measured with the VAS for the injured side was 0.5 ± 1.5 (range: 0–7) while stress-induced pain was 1.7 ± 2.7 (range: 0–8). Eight (25.8%) patients had tenderness on palpation, three (9.7%) patients reported taking regular pain medication, and five (16.1%) patients regularly received physiotherapy. Two patients had to be re-operated because of a postoperative complication (infection), yielding a complication rate of 6.5%: one patient was from the primarily conservative and the other from the primarily operated group. At the time of follow-up, 19 (61.3%) patients had their implants removed. A total of 20 (64.5%) patients reported paresthesia above or ventral of the clavicle. The total number of operations for all patients was 2.54 ± 0.87 (range: 1–4).

Constant score

The analysis of the age-adjusted Constant score according to Katolik et al. [14] showed a difference between the injured and noninjured side (Fig. 2). The breakdown of the different categories within the Constant score revealed a significant difference in the range of motion. The other categories also showed a tendency toward lower scores for the injured side, although these differences were not significant (Fig. 3).

Fig. 2
figure 2

Age-and gender-adjusted Constant score according to Katolik for the injured and noninjured shoulder

Fig. 3
figure 3

The different outcome categories for the Constant score. ROM range of motion

Radiological outcome

The mean radiological follow-up after revision surgery was 49 ± 30 months (range: 2–120). Fracture union was seen in 30 (96.8%) patients while no implant breakage, loosening, or loss of reposition were detected. Ossifications were present in 23 (74.2%) patients.

Initial treatment: nonoperative vs. operative

A total of ten patients underwent nonoperative treatment, while 21 were operated on (initial treatment before revision surgery). We compared these groups in terms of the major outcome criteria. Epidemiological data are summarized in Table 3. The conservative group included significantly more smokers.

Table 3 Epidemiological data of conservative vs. operative treatment groups (initial treatment before revision surgery)

The main clinical and functional outcome parameters are summarized in Table 4. Apart from the number of operations, no significant differences were detected. There were also no differences in all of the other aforementioned parameters (data not shown).

Table 4 Differences in clinical and functional outcome between conservative and operative group

Discussion

The most important finding of our study is that revision surgery of clavicular nonunions can lead to satisfying mid-term functional outcomes. The great majority of patients were satisfied with the revision surgery and were able to return to work and sport. Radiographic union was reached in 30 out of 31 patients and the overall complication rate was low.

Longer-term follow up data on the functional outcome after surgical treatment of clavicular nonunions in large cohorts are rarely reported. We identified a total of 15 studies over the past 25 years that included the assessment of functional results through either the Constant or DASH score (Table 5).

Table 5 Summary of available studies with outcome assessment based on the Constant or DASH score

In summary, the majority of the available studies are retrospective analyses with limited cohort sizes. Rollo et al. presented a larger study population with 57 consecutive cases [26]. Although we were not able to significantly increase the cohort size compared with the available literature, we report the results of two different study centers. Interestingly, our cohort is older on average than any of the aforementioned studies (52.3 ± 15.2 years). The clinical follow-up and the rate between atrophic and hypertrophic nonunions are comparable to those in the literature. Furthermore, similar to the available functional outcome assessment data, the Constant score is significantly decreased compared with the noninjured side (95.6 vs. 82.2). The difference is greater than the accepted threshold for the minimal clinically important difference for the Constant score [17]. Additionally, the range of motion for abduction was decreased in our study. Nevertheless, bony union was achieved in 96.8% of the cases and only two complications requiring revision surgery were recorded.

Additionally, our study provides detailed information on the technical aspects of the surgical revision. Huang et al. reported a mean operation time of 96 min, which is comparable to our data and longer than the average operation time for the initial fixation of a clavicular fracture in our cohort (96.3 ± 35.9 vs. 53.3 ± 15.8 min; p = 0.0002; [11]). Also, the plate length seems to be longer in the revision case compared with the initial surgery (8.1 ± 1.2 vs. 8.7 ± 1.4).

The comparison between initially nonoperative and operative treatment revealed no difference in the outcome measures. Interestingly, significantly more smokers were seen in the conservative cohort. This is in line with studies that identified smoking as an independent risk factor for the development of a nonunion [12].

The case presented in Fig. 1 is a good example of a typical failure of primary clavicular fracture fixation and secondary development of complications like infection and nonunion. In these cases, in which stability of the fixation seems to be the major issue, double plating is a good option to increase the overall fixation stability [15, 22].

Limitations

There are several inherent limitations to our study. Firstly, this is a retrospective cohort analysis solely reporting functional and radiographic outcomes. Like other comparable studies published previously, we were not able to recruit a larger cohort, although we did manage (for the first time) to report results from two centers. In addition, the indication for the index therapy (conservative vs. operative) remained unclear and was based on the decision of the treating surgeon. Similarly, the surgeon decided on the need for a bone graft in the revision situation. Therefore, the revision procedures were not standardized.

Practical conclusion

  • Operative treatment for clavicular nonunion with iliac bone grafting and plate fixation is a well-established and safe procedure.

  • Radiographic union and good functional outcomes can regularly be achieved.

  • Our study showed that this also applies to an older patient cohort.

  • In order to compare outcome measures between initial conservative and operative treatment, significantly larger cohorts are needed.