Conventionally, midshaft clavicular fractures have been treated nonoperatively, with the arm immobilized in a sling. In 1960, Neer reported an astonishingly low rate of nonunion in conservatively treated middle-third clavicular fractures1. Recent studies, however, have emphasized the risk of nonunion or symptomatic malunion following nonoperative treatment. In 1997, Hill et al. reported a 15% nonunion rate in nonoperatively treated clavicular fractures and a relationship between shortening and the risk of nonunion2. In 2004, Nowak et al. reported a 7% nonunion rate in nonoperatively treated clavicular fractures after six months as well as a risk of sequelae at the nine to ten-year follow-up. They also defined predictable risk factors, including lack of osseous contact at the fracture site, a transverse fracture, and increased age, that may cause complications in fracture-healing and overall recovery and were considered to be indications for operative treatment3. A systematic review of 2144 midshaft clavicular fractures indicated an overall nonunion rate of 5.9% after nonoperative treatment and a rate of 15.1% for fractures that were displaced. Fracture displacement, fracture comminution, female sex, and advanced age were associated with nonunion following conservative treatment4.
The goal in the treatment of clavicular fractures is to restore function of the upper extremity and prevent disability. A randomized clinical trial by the Canadian Orthopaedic Trauma Society in 2007 concluded that operative treatment of displaced midshaft clavicular fractures improved functional outcomes and diminished the rate of malunion and nonunion compared with nonoperative treatment at the one-year follow-up evaluation5. Similarly, in a randomized clinical trial comparing elastic stable intramedullary nailing with nonoperative treatment in 2009, Smekal et al. noted improved functional outcomes in the operatively treated group6. In contrast, in another randomized clinical trial in 2009, Judd et al. found no advantages of treatment with a Hagie pin compared with sling treatment at the one-year follow-up evaluation7.
The aim of our study was to compare functional outcome and disability following nonoperative or operative treatment of completely displaced midshaft clavicular fractures. Our hypothesis was that there would be no difference in functional outcome (as measured with the Constant score) or disability (as measured with the Disabilities of the Arm, Shoulder and Hand [DASH] score) between the nonoperative and operative treatment groups at the one-year follow-up evaluation.
This clinical trial was performed in a Level-I trauma center that provides trauma care to an area with 1.4 million inhabitants (Fig. 1). Institutional approval was obtained from the local Ethics Committee before initiation of the trial. The trial was registered at ClinicalTrials.gov (NCT01199653). Fractures were classified according to the AO/OTA (Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association) classification.
Inclusion Criteria
Inclusion criteria were (1) a middle-third clavicular fracture that was completely displaced (i.e., vertical displacement on an anteroposterior radiograph was at least equal to the width of the clavicle, such that there was no cortical contact between the main diaphyseal fragments) (Fig. 2), (2) treatment within seven days of injury, (3) patient age between eighteen and seventy years, and (4) patient willingness to provide informed consent.
Exclusion Criteria
Exclusion criteria included (1) a fracture without displacement, (2) a multiply injured patient, (3) an associated neurovascular injury, (4) insufficient patient cooperation (because of mental illness or drug addiction), (5) cancer or any other severe illness, (6) a pathological fracture, (7) treatment beginning later than seven days after injury, (8) an open fracture, (9) use of corticosteroid or immunosuppressive medication, (10) a concomitant upper-extremity fracture, (11) a previous fracture of the clavicle or elsewhere in the shoulder region, (12) pregnancy, or (13) lack of informed consent.
Nonoperative Treatment
Nonoperative treatment was performed with use of a sling (Polysling; Mölnlycke Health Care, Sweden) for three weeks. Pendulum motion was permitted during the first three weeks, followed by active abduction and flexion up to the horizontal plane from three to six weeks. The full range of active motion was permitted after six weeks, and return to full activities was permitted after three months.
Operative Treatment
Surgery was performed within seven days after injury in all patients. An orthopaedic surgeon, or a resident on duty under the supervision of a senior orthopaedic surgeon, performed the procedure. A prophylactic antibiotic (intravenous cefuroxime, 1.5 g) accompanied induction of general anesthesia. A beach-chair position was utilized. The involved upper extremity was free-draped. The incision was made in the line with the diaphysis of the clavicle. The deltoid and pectoralis major muscle attachments were raised and preserved to make a two-layered closure achievable. An AO/ASIF stainless steel reconstruction plate (thickness, 2.8 mm; width, 10.0 mm) and 3.5-mm stainless steel cortical screws were used (Synthes, Oberdorf, Switzerland). The plate was contoured to the curvature of the clavicle. The length of the plate was determined according to the amount of comminution of the fracture. The aim was to place at least three bicortical screws in the medial and lateral main fragments. The plate was placed on the anterior aspect of the clavicle (Fig. 3). No bone-grafting was performed. The deltoid and pectoralis major attachments were closed with multifilament absorbable sutures. The skin was closed with staples. Postoperatively, the arm was immobilized in a sling (Polysling) for three weeks. The postoperative exercise protocol was similar to that in the nonoperative group. No implant removal was scheduled.
Outcome Measures
The patients were examined clinically after three and six weeks, three months, and one year. The primary outcomes, which were the Constant and DASH scores, were evaluated at three months and at one year. At the three-month appointment, the Constant score was obtained by an independent physiotherapist instructed in standardized testing. At the one-year appointment, the Constant score was obtained by the main author (K.J.V.). These outcome assessors were not blinded to the patient treatment group. The physical examination was performed in a standardized manner. Secondary outcome measures were pain, fracture-healing, and complications. Pain was measured with use of a visual analog scale (VAS, 0 to 100 mm) after three and six weeks, three months, and one year.
Fracture-healing was evaluated on anteroposterior and 15° vertical (caudal to cranial) radiographs by an independent radiologist. Radiographs of both clavicles were made at the time of randomization and used to measure the shortening (the difference in length between the injured and the uninjured clavicle). Vertical displacement was assessed in proportion to the diaphyseal width of the clavicle. Union was defined as complete periosteal and endosteal bridging visible between the medial and lateral fragments on radiographs and the absence of pain and instability in the fracture region.
A complication or adverse effect was considered to exist if a subsequent surgical procedure was performed for any reason, if primary reduction was lost, if fixation failed or produced irritation, or if antibiotics were necessary to treat a wound infection. Symptomatic malunion was defined as shortening of >20 mm, angulation, or displacement of the clavicle combined with sequelae such as pain, weakness, or a tendency to fatigue easily. Delayed union was defined as the absence of bridging callus and endosteal healing on radiographs at the three-month assessment combined with pain and/or instability in the fracture region. Nonunion was defined as the absence of periosteal and endosteal healing on radiographs at the one-year follow-up evaluation. In addition, loosening or breakage of the implant in the operative group was counted as a nonunion. Symptoms of thoracic outlet syndrome due to brachial plexus compression (pain, paresthesia, numbness, or weakness in the hand, arm, or shoulder) were also assessed.
Sample Size Calculation
A power calculation was performed prior to the study to determine the sample size that would be needed to detect a ten-point difference in the Constant score between the groups with a power of 0.80 at a significance level of 0.05 if the means of the groups were compared with use of a t test. The required sample size was twenty-six patients per group. We therefore enrolled a total of sixty subjects to allow for an anticipated dropout rate of 15%.
Randomization
A research assistant generated the random allocation sequence, which was concealed from the authors. Block randomization was used, and the block size was varied randomly between four and ten. Patients were randomly assigned to two parallel groups, initially at a 1:1 ratio, to receive either nonoperative or operative treatment. Randomization was carried out with use of sequentially numbered, opaque, sealed envelopes. In the emergency room, the attending orthopaedic resident identified each potential patient and gave him or her introductory information on the study procedure. A member of the research group then gave the patient a through explanation, obtained informed consent, and enrolled the patient in the study. The health care providers involved with subsequent patient care were not blinded to the treatment.
Statistical Analysis
Statistical analysis was performed with use of PASW Statistics software (version 17.0; SPSS, Chicago, Illinois). A check for normality was performed with use of the Kolmogorov-Smirnov test. Nominal variables were analyzed with use of the chi-square test or the Fisher exact test. The latter was used if the calculated frequency was less than five. Scale, ordinal, and continuous variables were analyzed with use of the Mann-Whitney U test. A p value of <0.05 (two-sided) was considered significant. An intention-to-treat analysis was used to compare the groups. We used “last observation carried forward” as the imputation method for the patients who were lost to follow-up prior to the one-year appointment (i.e., the three-month outcome values were used in place of any missing one-year values).
Source of Funding
Our study was supported by Helsinki University Central Hospital research funds. The funder had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Sixty patients were included between August 2004 and October 2007, and an additional thirteen patients declined to participate. The nonparticipants did not differ from the enrolled patients in sex distribution, mean age, fracture classification, fracture shortening, or displacement. Twenty-eight of the included patients were randomized to the operative group and thirty-two to the nonoperative group (Tables I and II). One patient in the nonoperative group underwent successful surgery after four months because of symptoms of brachial plexus irritation (pain, a feeling of weakness, and paresthesia in the arm) that were noted clinically at the three-month appointment. On the basis of the intention-to-treat principle, she continued to be analyzed in the nonoperative group. Nine (15%) of the sixty patients were lost to follow-up between three months and one year; seven were in the nonoperative group and two in the operative group. Six of these patients did not appear at the one-year follow-up appointment despite several calls, one had emigrated, one had moved to another locality, and one could not participate because of pregnancy. Thus, twenty-six patients in the operative group and twenty-five in the nonoperative group completed the one-year assessment. Participants with missing outcomes were included in the one-year analysis with use of imputation; the missing one-year values were replaced by the three-month values (“last observation carried forward”).
In our primary hypothesis, we assumed that no difference in function (Constant score) or in disability (DASH) between the groups would be evident at the one-year follow-up assessment. The results verified our hypothesis; no significant difference in the Constant score between the nonoperative and the operative group was found at the three-month (p = 0.77) or one-year (p = 0.75) follow-up assessments. Likewise, we found no difference in the DASH score at three months (p = 0.81) or one year (p = 0.89) (Table III).
Our secondary outcome measures were pain, fracture-healing, and complications. At the three-week assessment, the operative group had less pain compared with the nonoperative group (p = 0.049), but no difference existed between the groups at six weeks (p = 0.76), at three months (p = 0.65), or at one year (p = 0.98).
All fractures in the operative group healed, but only nineteen (76%) of the twenty-five fractures in the nonoperative group were seen to be healed on the radiographs (p = 0.01). Five of the six patients whose radiographs indicated nonunion had the injury on the nondominant side. During follow-up, we discussed the option for operative intervention with the patients with nonunion, but because of the minor disability that resulted, none elected to undergo surgical treatment.
A total of nineteen complications were recorded, twelve in the nonoperative group and seven in the operative group (p = 0.15). None of the complications were major. As mentioned, all six of the nonunions occurred in the nonoperative group (p = 0.01) (Fig. 4). In addition, delayed healing occurred in three fractures in the operative group and one in the nonoperative group (p = 0.61); all of these fractures were consolidated at the one-year follow-up appointment. Two patients had a symptomatic malunion (shortening of >20 mm, angulation, or displacement of the clavicle, combined with sequelae); both were in the nonoperative group (p = 0.24), and both of these patients experienced only mild disability and declined surgery. No wound infections occurred. In one patient in the surgical group, the plate became slightly bent, and reduction was lost between the six-week and three-month visits. In another patient, the one-year radiograph showed the plate to have broken, but the fracture had healed without loss of reduction. In another patient, the plate caused mild irritation. One patient in the nonoperative group underwent surgery after four months because of brachial plexus irritation (pain, feeling of weakness, and paresthesia in the arm) (p = 0.49). Three refractures occurred, two in the nonoperative group and one in the operative group (p = 0.61). The two patients in the nonoperative group both fell six months after the initial injury and sustained a new fracture at the bridging callus, and both of these fractures healed. The patient in the operative group fell from a motorcycle two weeks before the one-year appointment and sustained a new fracture adjacent to the medial end of the plate. Because of this recent refracture, his Constant score (39.0 points) at the one-year assessment was poor compared with the mean value in the operative group (87.3 points).
Analysis of the nonunions in the nonoperative group revealed a significant difference in displacement at baseline between united and nonunited fractures (p = 0.009). All of the fractures in this group that were displaced by <1.5 clavicle widths united, whereas one-half of the fractures with displacement of >1.5 clavicle widths failed to unite. No differences in union rate according to fracture classification (p = 0.06) or shortening (p = 0.77) were identified. At the one-year assessment, the Constant score (p = 0.50) and pain (p = 0.57) did not differ between patients with united and nonunited fractures. However, there was a 16-point difference, in favor of united fractures, in disability assessed with use of the DASH score (p = 0.047) (Table IV).
This randomized clinical trial comparing sling treatment to plate osteosynthesis for displaced midshaft clavicular fractures revealed no differences in the Constant score, DASH score, or pain at the one-year follow-up assessment. All fractures in the operative group healed, but the prevalence of nonunion in the nonoperative group was high (24%). The patients with nonunion experienced greater disability compared with patients with union, but they declined any proposed reconstructive surgery.
Another recent randomized clinical trial of displaced midshaft clavicular fractures indicated that the operative group had better Constant and DASH scores at the one-year follow-up assessment, as well as lower rates of malunion and nonunion, compared with the nonoperative treatment group5. Another trial by Smekal et al. in 2009, which compared stable intramedullary nailing to nonoperative treatment, revealed no difference in the DASH score at twenty-five weeks of follow-up and a statistically significant but not clinically important difference in the Constant score at two years of follow-up6. A randomized clinical trial comparing intramedullary fixation to plating for displaced midshaft clavicular fractures indicated no difference in functional shoulder scores or in the union rate8. A prospective study with ten years of follow-up revealed that 46% of nonoperatively treated patients did not fully recover and 27% had cosmetic defects resulting from the fracture3.
Our study indicated that nonoperative and operative treatment of displaced midshaft clavicular fractures resulted in equivalent results with regard to the Constant score, DASH score, and pain, and that nonunion resulted in only mild impairment of shoulder function. In contrast, the 2007 study by the Canadian Orthopaedic Trauma Society indicated that operative treatment yielded statistically better results with regard to the Constant and DASH scores compared with nonoperative treatment5. However, the Canadian study population differed from ours in several ways: adolescents were also included (the age range was sixteen to sixty years), some patients had concomitant fractures and injuries, and surgical treatment was required to take place within twenty-eight days after injury.
To properly evaluate functional recovery, we need to know the minimal clinically important difference (MCID) or minimal detectable change (MDC) in a score—i.e., the smallest change that the patient may notice. For the Constant score, no MDC or MCID has been reported, which limits use of this score in follow-up studies9. In a systematic review involving the DASH score, the MDC was considered to be 10.5 points and the MCID, 10.2 points10. It should be noted that MDC and MCID values may depend on the disease or injury, and some cultural variation in these values probably also exists. However, in our study, the DASH score differed between the groups by only 0.1 point at the three-month assessment and 2.8 points at one year, representing clinically imperceptible changes.
The mean shortening of the injured clavicle on the initial radiographs was 10 mm in the nonoperative group and 11 mm in the operative group compared with 14.3 and 15.7 mm in the Canadian study5. This difference in fracture shortening between the studies might be explained by dissimilar measuring methods. We compared the length of injured clavicle with that of the contralateral one, whereas the Canadian group estimated shortening of the clavicle on the radiograph of the fractured side only. Fractures of the diaphyseal aspect of the clavicle are commonly oblique or comminuted. Assessing only the radiographs of the injured side is challenging because true fracture shortening is difficult to measure. Shoulder position during radiography also has a substantial effect on the length of the fractured clavicle. In our study, the patient was facing toward the radiography film, and thus the fracture position might have improved momentarily. The association between clavicular shortening after a healed midshaft fracture and the clinical outcome is still controversial. Some studies have indicated that shortening of the clavicle after fracture causes disability of the shoulder2,5,11, whereas others have found no association with disability3,12,13.
The difference in the DASH score between the united and nonunited fractures in the nonoperative group was both clinically important (15.8 points) and significant (p = 0.047). According to the DASH score, patients with nonunion experienced mild disability. However, when function was assessed with use of the Constant score, the difference between patients with union and nonunion was only 1.8 points, which was not significant (p = 0.5). On the basis of the Constant scores in our study, nonunion does not appear to diminish shoulder function or strength of the upper limb to an appreciable extent.
When pain was evaluated with use of a VAS ranging from zero to 100 mm at three and six weeks, three months, and one year, the greatest difference between the groups (6 mm) was observed at the three-week follow-up assessment, and this was the only difference that was significant (p = 0.049). A total of twelve complications were noted in the nonoperative group and seven in the operative group. On the basis of previous studies, we had expected the nonoperative group to have the higher nonunion rate, but the 24% nonunion rate was surprisingly high. A connection appeared to exist between nonunion and fracture displacement. Displacement of >1.5 clavicle widths was associated with a considerable risk of nonunion in nonoperatively treated patients.
Our study was performed at a Level-I trauma center to which patients with minor injuries must have a referral. The two other hospitals providing emergency health-care services in our district were informed about our ongoing study and asked to refer all patients with a clavicular fracture to our hospital. Despite this effort, some patients may not have been referred, and this may have produced selection bias in the study population. Because it was difficult to blind the investigators to the selected treatment, some detection bias may also have occurred. Regardless of minor biases, however, we believe that our results can be generalized to any individual with an isolated, displaced midshaft clavicular fracture.
Our study has some limitations. A significant difference in mean age existed between the randomized groups. In theory, the lower mean age in the nonoperatively treated patient group could have given them some advantage with regard to fracture-healing. Even so, we concluded that the difference between the mean ages of thirty-three and forty-one years did not bias the results. Nine patients did not complete the one-year follow-up assessment, and the patients who were lost to follow-up were not distributed equally because seven were in the nonoperative group. Thus, on the basis of the nonunion rate in our study population, it could be speculated that other nonunions occurred among these lost patients. We did not assess function and disability at early visits (three and six weeks), and for this reason we lack knowledge of the difference in recovery between the groups in the very early phase. Prior studies have suggested that an earlier return to activity might be a major advantage of primary stabilization. Distinguishing nonunion from union on radiographs may be challenging. The nonunion can be subtle, and it can be missed because of the plate. Performance of the final evaluation by the main author may have produced detection bias. In the analysis phase of the study, we used imputation for missing one-year data. Ideally, the use of imputation should have been specified prior to starting the study. Finally, because we had no high-demand patients, such as professional athletes, we are unable to address the difficulties or deficiencies that malunion or nonunion may cause for a person with very high physical demands.
Although it may seem that our study has revealed findings different from those reported in prior publications, there are many similarities. We found primary operative stabilization of displaced midshaft clavicular fractures to be a reliable surgical technique with a low complication rate and a high union rate. We also found that nonoperative treatment resulted in a significantly higher nonunion rate (24%) and that nonunion was associated with greater initial fracture displacement (p = 0.009). Moreover, patients with nonunion had worse DASH scores (mean difference, 16 points; p = 0.047) compared with patients whose fractures healed. The substantial difference between the findings of our study and those of prior trials was the relatively high function and minimal symptoms of the patients in our study whose radiographs indicated a nonunion. Cultural and psychosocial disparities may be responsible for this discrepancy. Future research should focus on prognostic factors that may predict nonunion or symptomatic malunion. We found no difference in the Constant score between patients with united and nonunited fractures. One explanation may be that the Constant score is unable to distinguish minor differences in upper-extremity function. In future trials, comparison of nonoperative with operative treatment in high-demand patients such as professional athletes and individuals doing physically demanding work is important. Plate osteosynthesis, elastic stable intramedullary nailing, and nonoperative treatment should also be compared in a future clinical trial.