Background: Nonunion is a rare complication of a fracture of the clavicle, but its occurrence can compromise shoulder function. The aim of this study was to evaluate the prevalence of and risk factors for nonunion in a cohort of patients who were treated nonoperatively after a clavicular fracture.
Methods: Over a fifty-one-month period, we performed a prospective, observational cohort study of a consecutive series of 868 patients (638 men and 230 women with a median age of 29.5 years; interquartile range, 19.25 to 46.75 years) with a radiographically confirmed fracture of the clavicle, which was treated nonoperatively. Eight patients were excluded from the study, as they received immediate surgery. Patients were evaluated clinically and radiographically at six, twelve, and twenty-four weeks after the injury. There were 581 fractures in the diaphysis, 263 fractures in the lateral fifth of the clavicle, and twenty-four fractures in the medial fifth.
Results: On survivorship analysis, the overall prevalence of nonunion at twenty-four weeks after the fracture was 6.2%, with 8.3% of the medial end fractures, 4.5% of the diaphyseal fractures, and 11.5% of the lateral end fractures remaining ununited. Following a diaphyseal fracture, the risk of nonunion was significantly increased by advancing age, female gender, displacement of the fracture, and the presence of comminution (p < 0.05 for all). On multivariate analysis, all of these factors remained independently predictive of nonunion, and, in the final model, the risk of nonunion was increased by lack of cortical apposition (relative risk = 0.43; 95% confidence interval = 0.34 to 0.54), female gender (relative risk = 0.70; 95% confidence interval = 0.55 to 0.89), the presence of comminution (relative risk = 0.69; 95% confidence interval = 0.52 to 0.91), and advancing age (relative risk = 0.99; 95% confidence interval = 0.99 to 1.00). Following a lateral end fracture, the risk of nonunion was significantly increased only by advancing age and displacement of the fracture (p < 0.05 for both). On multivariate analysis, both of these factors remained independently predictive of nonunion (p < 0.05), and, in the final model, the risk of nonunion was increased by a lack of cortical apposition (relative risk = 0.38; 95% confidence interval = 0.25 to 0.57) and advancing age (relative risk = 0.98; 95% confidence interval = 0.97 to 0.99).
Conclusions: Nonunion at twenty-four weeks after a clavicular fracture is an uncommon occurrence, although the prevalence is higher than previously reported. There are subgroups of individuals who appear to be predisposed to the development of this complication, either from intrinsic factors, such as age or gender, or from the type of injury sustained. The predictive models that we developed may be used clinically to counsel patients about the risk for the development of this complication immediately after the injury.
Level of Evidence: Prognostic study, Level I-1 (prospective study). See Instructions to Authors for a complete description of levels of evidence.
Nonunion is usually considered to be an uncommon complication following a clavicular fracture, although recent studies have suggested that the prevalence is higher than previously reported1-3. A wide range of factors have been hypothesized to contribute to the risk of nonunion after this injury; these include intrinsic factors, such as the age and gender of the patient, and extrinsic factors, such as the location and extent of displacement of the fracture1-8. The relative importance of these factors has not previously been estimated in an unselected, prospective series of patients with a clavicular fracture, to our knowledge.
Despite the risk of nonunion after these fractures, most are still treated nonoperatively. Although many forms of nonoperative treatment have been described, the most widely accepted treatment involves the provision of a simple sling for support during the initial phase of treatment, with early mobilization of the shoulder as the pain subsides. The aim of this prospective, observational cohort study was to determine the rate of nonunion after this form of nonoperative treatment for a consecutive series of unselected clavicular fractures and to examine in detail the effect of a wide variety of patient-related and injury-related risk factors on the risk of this complication.
Materials and Methods
Over the fifty-one-month period from January 1997 to March 2001, we performed a prospective, observational cohort study of a consecutive series of 918 patients referred to our service with a radiographically confirmed fracture of the clavicle. Our affiliated emergency department provides the only acute musculoskeletal trauma service for the local adult population, and, during the study period, all patients with radiographically proven fractures of the clavicle were referred for follow-up to a dedicated special review clinic.
Eight patients were excluded from the study, as they received immediate surgery because of skin compromise (an open fracture or severe skin tenting), a floating shoulder, or multiple injuries. We also excluded forty-two patients who were not locally resident in our catchment area and who were unable to return for follow-up examinations.
All of the remaining 868 patients (638 men and 230 women with a median age of 29.5 years; interquartile range, 19.25 to 46.75 years) included in the study were initially treated nonoperatively and were instructed to wear a collar and cuff sling for comfort during the first two weeks after the injury. Patients were encouraged to commence active shoulder movements, and formal physiotherapy supervision was not instituted except for the patients who had persistent shoulder stiffness at the six-week follow-up visit.
We aimed to evaluate all patients prospectively at six, twelve, and twenty-four weeks after the injury at the dedicated review clinic. A research physiotherapist coordinated the follow-up evaluations for all patients who were enrolled, and those who defaulted from follow-up were offered at least two additional appointments, either by mail or telephone. Standardized anteroposterior and 30° caudal tilt radiographs were made at each follow-up visit and were all reviewed by one of the authors (C.M.R.). Fractures in three regions of the clavicle, i.e., the diaphysis and the two ends, were recognized. The medial end was the fifth of the bone lying medial to a vertical line drawn upward from the center of the first rib. The lateral end was the fifth of the clavicle lateral to a vertical line drawn upward from the center of the base of the coracoid process, a point normally marked by the conoid tuberosity. The diaphysis was the intermediate three-fifths between these two areas. The measured severity of displacement, translation, angulation, and shortening at the fracture site was recorded from the initial radiographs made after the injury. The presence or absence of complete displacement of the bone ends was recorded as a binary categorical variable; the severity of translation and shortening was recorded in millimeters, and the degree of angulation at the fracture site was recorded in degrees (Table I).
We defined fracture union as the absence of mobility or pain on stressing the site of the fracture and evidence of bridging callus on radiographs. For the purposes of this study, a non-union was defined as a fracture that remained unhealed according to these criteria at twenty-four weeks after the injury. In twelve patients, there was uncertainty as to whether the fracture was united at that time because of equivocal signs of union, either on clinical examination or on radiographs. These patients underwent exploratory surgery, and four fractures were found to be united; the remaining eight had a definite nonunion, which was treated with plate fixation and bone-grafting. These eight fractures were considered to be nonunions at twenty-four weeks, but all united following surgery.
All information gathered in the study was recorded and analyzed with use of the SPSS software package (SPSS, Chicago, Illinois). Survival analysis was performed with use of life tables to establish the rate of union after clavicular fracture at six, twelve, and twenty-four weeks after the injury. In addition, we used a Cox proportional-hazards model of survivorship to estimate the effect of a wide variety of patient-related and injury-related putative factors (Table I) on fracture union. All variables that were significantly predictive of nonunion on univariate analysis were included in a multivariate model (with use of backward conditional stepwise methodology) to determine the factors that were independently predictive of union and to generate models to estimate the risk for the development of this complication. All models underwent standard tests for goodness-of-fit, including investigation of the need for interaction terms between covariates, an analysis of residuals, and tests of regression leverage and influence.
With use of the models, the summated products of the risk factor score and its regression coefficients (Bx) for an individual patient can be used to calculate the prognostic index (PI). The relative risk of nonunion is given by the expotential function of Bx (Exp(Bx)). The predicted probability (St) of a fracture having united at a given time (t) can be computed with use of the equation: where So(t) is the baseline cumulated survival function at the time t and exp(PI) is the exponential function of PI. Plotting St against PI at six, twelve, and twenty-four weeks allows an estimation of the probability of a fracture healing at each of these times for an individual with a given prognostic index. In all analyses, a p value of ≤0.05 was considered significant.
There were 581 fractures in the diaphysis, 263 fractures in the lateral fifth of the clavicle, and twenty-four fractures in the medial fifth. According to the commonly used classification systems2,9, fractures of the diaphysis occurred most commonly in younger males (median age, twenty-five years; interquartile range, seventeen to 39.5 years) (Table II), whereas fractures of the lateral fifth were more common in elderly women (median age, forty years; interquartile range, twenty-six to fifty-six years) (Table II). Fractures sustained during simple falls and falls from a height occurred more commonly in the middle-aged and the elderly, with an almost equal gender distribution, whereas fractures in the young and predominantly male population were more commonly sustained in sports injuries and traffic accidents (Table III). Injuries sustained during falls off mountain and road bicycles accounted for 42% of all injuries sustained in traffic accidents.
A total of 206 patients (23.7%) of the original cohort of 868 patients were lost to follow-up before the final follow-up assessment at twenty-four weeks. Twenty-one of them had died from unrelated causes, and forty-one were too frail or demented to return for follow-up. Of the remainder, forty-nine patients could not be traced and ninety-five did not return for follow-up, despite the fact that they lived at the same address and were reminded twice, either by mail or by telephone, to attend the clinic. The demographic data for these latter two groups in terms of the variables recorded in Table I were not significantly different from those for the remainder of the fracture population.
On survivorship analysis, the overall prevalence of non-union within our study population at twenty-four weeks after the fracture was 6.2% (95% confidence interval, 4.2% to 8.3%). At twenty-four weeks, 4.5% (95% confidence interval, 2.5% to 6.5%) of diaphyseal fractures, 11.5% (95% confidence interval, 5.9% to 17.1%) of lateral end fractures, and 8.3% of medial end fractures (95% confidence interval, 0% to 23.8%) had not united. The data on the prevalence of nonunion at six, twelve, and twenty-four weeks after the injury, according to age, gender, fracture location, fracture displacement, and comminution for diaphyseal and lateral end fractures, are shown in the Appendix.
The paucity of medial end fractures precluded a detailed statistical analysis of the factors affecting the rate of nonunion of such fractures. Although the rate of nonunion following displaced fractures (14.3%) was higher than that of nondisplaced fractures (6.7%) at twenty-four weeks, this difference was not significant because of the small number of patients who sustained a medial end fracture.
Univariate analysis revealed that, after a diaphyseal fracture, the risk of nonunion was significantly increased only by advancing age, female gender, complete displacement of the fracture (lack of any cortical apposition), and the presence of comminution (p < 0.05 for all). On multivariate analysis, all of these factors remained independently predictive of non-union (p < 0.05). Interactive product functions of injury mechanism with these variables were also predictive of non-union (p < 0.05), as was injury mechanism in isolation (p < 0.05), but none of these terms were independently predictive when adjusted for the other factors in the multivariate model. Further stratification to take into account the measured extent of fracture angulation, translation, or shortening did not significantly add to the predictive power for nonunion. In the final model, the prognostic index for an individual patient can be calculated as shown in Table IV; the estimated probability of a diaphyseal fracture remaining ununited at six, twelve, or twenty-four weeks can then be estimated with use of Figure 1.
On univariate analysis, after a lateral end fracture, the risk of nonunion was significantly increased only by advancing age and displacement of the fracture (p < 0.05 for both). As with diaphyseal fractures, further stratification of the extent of displacement did not significantly add to the predictive power for nonunion. On multivariate analysis, both of these factors remained independently predictive of nonunion (p < 0.05), and the final model is shown in Table V; the estimated probability of a lateral end fracture remaining ununited at a given time can then be estimated with use of Figure 2.
Our findings confirm that nonunion after a clavicular fracture is an uncommon occurrence, although the prevalence is higher than has been previously reported in retrospective series2,4,10. Our study cohort was a consecutive series of patients who presented to our unit following a clavicular fracture and who were followed prospectively. Other than a small minority of individuals who underwent emergency surgery because of skin compromise or floating shoulder, all patients were initially managed nonoperatively. The broad demographic features of the study group were similar to those identified in previous epidemiological studies2,3,11,12. It is therefore likely that the prevalence of nonunion in our study is representative of the prevalence of this complication among fractures initially treated nonoperatively.
Medial end fractures were comparatively rare, accounting for <5% of all fractures in this series. Estimating the relative importance of risk factors for nonunion, therefore, was difficult, and a detailed statistical analysis was not performed. For the small number of patients in this series who sustained a medial end fracture, the rate of nonunion was relatively high and the risk appeared to be increased for fractures with complete displacement compared with those that had residual cortical apposition; however, because of the small number of patients who sustained a fracture in this location, the difference was not significant.
Diaphyseal fractures were the most common fracture in this series, accounting for nearly 70% of the fractures. The majority of these fractures occurred in a younger, predominantly male population and were often caused by a sports injury or a traffic accident. A high proportion of the fractures sustained in traffic accidents occurred in individuals riding bicycles. Since this population was skewed toward the young, the majority of nonunions were encountered in this age-group. However, paradoxically, the risk of nonunion was highest in elderly female patients with a diaphyseal fracture. Most nonunions in younger individuals require surgical intervention because of the poor function associated with the development of this complication, while some of the more elderly patients may not require surgery because of their reduced functional demands13-15.
We were unable to predict with greater sensitivity the risk of nonunion after a diaphyseal fracture, when taking into account the severity of fracture displacement, as other studies have done1. It is possible that fractures with more displacement are associated with a poorer functional or cosmetic result or with the development of malunion, thoracic outlet syndrome, or other symptoms of brachial plexus irritation16,17, and the risk of these complications and the risk of nonunion may be reduced by performing surgery at an earlier stage. However, at present, most diaphyseal fractures of the clavicle are initially treated nonoperatively, and the most common reason for operative treatment remains nonunion. Therefore, we believe that additional modifications to the existing classification systems2,9 are unnecessary.
Lateral end fractures accounted for approximately one-third of the fractures in this series. The risk of nonunion was highest in elderly patients, and, because a greater number of elderly patients sustained a lateral end fracture, they had a relatively higher proportion of the nonunions than did patients with a diaphyseal fracture. Many patients with a nonunion are relatively asymptomatic, despite the nonunion, and they do not subsequently require surgical intervention2,8.
We developed multivariate models to estimate the risk of nonunion after diaphyseal and lateral end fractures. These models may be used clinically to counsel patients about the risk for the development of this complication immediately after the injury. Although these models can be used to assess the risk of nonunion following a fracture, their positive predictive power is still low, as most fractures have united by twenty-four weeks, irrespective of the number of risk factors for nonunion.
A potential weakness of our study is the relatively high proportion of patients who were lost to follow-up before the final assessment at twenty-four weeks. Many of these patients were frail and demented or had died. However, ninety-five patients who remained locally resident were lost during the twenty-four-week follow-up period, even though they were offered additional follow-up appointments, and forty-nine patients could not be traced. We were unable to confirm whether these fractures had united, and it could be argued that many of these patients probably failed to return for the follow-up examination because they were asymptomatic, with a healed fracture. The possibility exists that we have, therefore, overestimated the risk of nonunion in our series. However, the demographic data for the patients lost to follow-up were not significantly different from those for the remainder of the patients with a clavicular fracture, suggesting that the patients who completed follow-up were representative of the population as a whole. We also think that our use of a survival methodology to analyze our data was the most effective way of reducing the potential for confounding introduced by those lost to follow-up.
It could be argued that a watershed of twenty-four weeks for the diagnosis of nonunion is too late in the management of a patient with a clavicular fracture, and surgery should be considered for the treatment of a nonunion at an earlier stage. However, our results demonstrated that approximately 90% of diaphyseal fractures and 80% of lateral end fractures that are unhealed at twelve weeks will progress to union by twenty-four weeks. There is evidence from the Cox proportional-hazards model that patients with a lower prognostic index have worse prospects for progression to union if the fracture remains ununited at twelve weeks. For instance, a patient with a prognostic index of –2 (for example, a sixty-year-old woman with a displaced, comminuted diaphyseal fracture) has a 72.4% projected probability that the fracture will remain ununited at twelve weeks and a 38.3% probability of nonunion at twenty-four weeks, whereas a patient with a prognostic index of –1 (for example, a twenty-two-year-old man with a displaced but noncomminuted fracture) has a 41.6% probability that the fracture will remain ununited at twelve weeks and only a 7.4% probability of nonunion at twenty-four weeks. This suggests that earlier surgery might be beneficial for patients with a lower prognostic index. However, age is a major determinant of the prognostic index, and the lower functional expectations of the elderly may lead such patients to decline the surgery or it may be deemed to be inappropriate for many of them. It is also unclear whether earlier surgery with its attendant risks of infection, fixation failure, and persistent nonunion would guarantee an improved result compared with the traditional methods of nonoperative treatment in most patients18,19. A prospective, randomized, controlled trial of primary operative treatment compared with nonoperative treatment in stratified high-risk groups would be useful to evaluate this issue further.
Life tables of the clavicular fractures at the various time-intervals are available with the electronic versions of this article, on our web site at www.jbjs.org (go to the article citation and click on “Supplementary Material”) and on our quarterly CD-ROM (call our subscription department, at 781-449-9780, to order the CD-ROM).
In support of their research or preparation of this manuscript, one or more of the authors received grants or outside funding from Scottish Orthopaedic Research Trust into Trauma (SORT-IT). None of the authors received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, educational institution, or other charitable or nonprofit organization with which the authors are affiliated or associated.
Investigation performed at the Shoulder Injury Clinic, Orthopaedic Trauma Unit, Edinburgh, Scotland
- Copyright © 2004 by The Journal of Bone and Joint Surgery, Incorporated