Predictors of nonunion and reoperation in patients with fractures of the tibia: an observational study

We conducted an observational study of 200 tibial shaft fractures to determine what fracture characteristics are predictive of a nonunion and the need for a reoperation. Medical records (clinical notes and radiographs) of patients with tibial shaft fractures were reviewed. This study was approved by the Research Ethics Board (project number 10-595-C under the Hamilton Health Sciences/McMaster Health Sciences, Research Ethics Board).

Consecutive patients with tibial shaft fractures were identified at three sites within a university-affiliated teaching hospital. Patients were identified through their hospital medical records by the Health Records Department. The study included patients who were treated for acute fractures and excluded those that were referred for a complication. In addition, patients were required to fulfill the following eligibility criteria: (1) 19 years of age or older, (2) tibial fracture requiring operative fixation with internal fixation, (3) sufficient clinical information available within the patient’s medical record and (4) radiographs available for the assessment. Tibial fractures that extended into the joint, bilateral fractures and Gustilo-Anderson Type IIIc open fractures were excluded. Two individuals independently assessed each identified patient’s electronic medical records to determine if they met the pre-defined eligibility for this study. Any disagreements were resolved by a third individual and reasons for exclusion were documented.

Reoperation was defined as any invasive procedure completed to promote fracture healing. Such invasive measures included, but were not limited to, bone grafting, nail dynamization, or implant exchange. Nonunions were determined by the patient’s attending physician as dictated in their clinical records and identified during the data collection process.

Study specific case report forms were developed for this study. The case report forms were developed using previous literature and 15 surgeons in the field of orthopaedic trauma surgery. A sampling to redundancy strategy was utilized, by which additional experts were contacted until no new items were generated for the case report forms. A panel of experts rated their perceived importance of each identified factor on a scale of one to ten, with ten as the top priority to include in the analysis and one as the lowest priority to include in the analysis. The highest ranked factors were therefore included on the case report forms. The case report forms were piloted on ten sample patients before the study was initiated. The pilot test data was reviewed to confirm that the case report forms captured the information it was intending to collect.

Two individuals independently extracted the data from each patient’s electronic medical record. The two individuals reviewed all available medical records including initial hospital notes from the emergency room consultation, the surgical consultation, the operative report(s), in-hospital progress reports, the hospital discharge summary, all fracture clinic notes, subsequent operative report(s) (if applicable), and all re-hospitalization notes (if applicable) related to the patient’s included fracture. They recorded the data on the study-specific case report forms. Any disagreements were resolved by a third individual, who independently extracted the data in which there was a discrepancy.

An independent reviewer assessed the initial x-ray, the post-operative x-ray, and all available sequential x-rays. The independent reviewer assessed the radiographs and documented fracture characteristics, post-operative characteristics and any identified radiographic complications on the case report forms. Fracture characteristics included location of fracture, type of fracture, cortical continuity and bone loss. Post-operative characteristics included size of fracture gap and amount of cortical continuity (lack of bridging of tibial cortices), which were used as indicators of healing.

Data from both the available patient medical records and data from the case report forms completed during radiographic evaluation were consolidated and, in preparation for analysis, entered into a database and a subset of the data was reviewed. Reporting adverse events was not applicable, as the data sources used in this study did not contain physician attribution of causality of adverse events to any medicinal or surgical products.

Our previous randomized trial in patients with tibial shaft fractures identified an event rate of a secondary intervention of 14% (95% CI: 12-16%) in closed fractures and 27% (95% CI: 22-31%) in open fractures [26]. Given a sample size of 200 patients with tibial shaft fractures treated operatively, we anticipated to have 50 patients (25%) with one healing complication leading to an expected outcome event per predictor variable (EPV) of 10 for the logistic regression model. Based on simulation studies, EPV values of 10 (or greater) showed no major issues in terms of confidence interval coverage, type I error, relative bias and other model performance measures [27, 28].

Data was analyzed using SPSS Version 21.0 [SPSS, Chicago, IL]. Our primary analyses included looking at nonunions and reoperations.

Dichotomous data was reported as number of participants and proportions, with corresponding confidence intervals to estimate precision. Continuous data was presented as means and medians with standard deviations. Descriptive statistics describing the patients’ demographics, fracture characteristics, fracture complications and management of the documented fracture complications were also provided.

The primary analysis was a multivariable logistic regression using the primary outcome (need for reoperation) as the dependent variable looking for predictor variables in patients with tibial shaft fractures at baseline. The same analysis was completed to assess fracture characteristics and their predictive value using nonunion as the secondary outcome.

We also conducted an exploratory regression analysis comparing different aggregates of two of the three fracture characteristics included in a previously-developed prognostic index for their predictive value using reoperation and nonunion as outcomes.

A chi-square test was performed to test the null hypothesis of no association between nonunion and reoperation.

We identified 594 potentially eligible patients from the hospital database review. Of these, 200 were included in our analysis. Patients were excluded if they had an insufficient amount of follow-up clinical or radiographical data (141), a tibial fracture not fixed via operative internal fixation (108), a fracture extending into the joint (87), a non-tibial fracture (35), a bilateral fracture (17), or were under the age of 18 (6). The typical patient was a healthy male (69.0%), averaging 42 years of age, who consumed alcohol (73%), and had no previous injuries to the ipsilateral tibia (98%) (Table 1). Most tibial fracture injuries were accompanied by injuries to the ipsilateral fibula (92%). Of the fractures included, 43% were a result of a motor vehicle-related incident. Fracture injuries tended to be closed (78%) and were treated with reamed intramedullary nailing (81%). Post-operatively, patients did not suffer from any fracture complications during their hospital stay (97%), which was often less than a week (Table 2).

In our cohort of tibial fractures, 37 (18.5%) went on to nonunion, as identified by the attending physician. Of the 37 patients with nonunion, 62.2% of them were identified within the first six months of the initial injury. Additionally, 27 (13.5%) of all included patients underwent a reoperation, with 55.6% of them occurring within the first six months of the initial injury.

Univariable analysis identified six fracture characteristics associated with the incidence of nonunion (p < 0.05), which included fractures with less than 25% cortical continuity (odds ratio = 6.44 [95% CI 1.89, 21.95]; p = 0.003), open fractures (odds ratio = 2.56 [95% CI 1.24, 5.29]; p = 0.011), the presence of comminution (odds ratio = 2.21 [95% CI 1.05, 4.66]; p = 0.037), and an oblique (odds ratio = 2.94 [95% CI 1.32, 6.58]; p = 0.009) or segmental (odds ratio = 3.17 [95% CI 0.96, 10.46]; p = 0.058) fracture type (Table 3). Multivariable logistic regression analysis suggested that only cortical continuity remained predictive (odds ratio = 4.72 [95% CI 1.33, 16.76]; p = 0.02) (Table 4).

Alternatively, open fractures and transverse fractures were the only fracture characteristics that showed any significant predictive value for reoperation, as shown via univariable analysis (Table 5). All tibial fractures with less than 25% cortical continuity accounted for all 27 reoperations we identified in our sample. The presence of a transverse fracture was the only variable to approach significance for the incidence of reoperation in the multivariable logistic regression analysis (odds ratio = 3.03 [95% CI 1.00, 9.18]; p = 0.05) (Table 6).

Of the 37 nonunions reported in this study, 32.4% of them were treated with a noninvasive therapy option alone to promote bony union. Such options include, but are not limited to, ultrasound therapy, electrical stimulation, or medication. Additionally, 40.5% of the nonunions were treated with a reoperation alone and 24.3% were treated with a combination of both noninvasive therapies and a reoperation. There was only one case who did not receive any secondary treatment for nonunion.

A previous index identified open fractures, presence of a fracture gap post-fixation, and a transverse fracture type as variables in a predictive model for reoperation within one year following operative management of tibial fractures [25].

Using the same prognostic risk model (Table 7), our findings largely confirmed the incremental increase in the risk of reoperation with one, two, and three prognostic risk factors. Patients with at least two of the three risk factors were more likely to develop a nonunion, with the presence of both a fracture gap and open fracture having the highest predictive value for a nonunion over any other aggregate of two of the three variables included in the prognostic risk model (Table 8). Patients with at least two or all three risk factors were more likely to incur a reoperation. All possible aggregates of two of the three prognostic risk variables provided a predictive basis for reoperation (Table 9).

Chi-square analysis found an association between nonunion and reoperation, χ² (1, N = 198) = 101.4, p < 0.001. 12.1% of all included patients experienced both a nonunion and a reoperation and patients with a nonunion were 97 times (95% CI 25.8-366.5) more likely to have a reoperation.

Our multivariate regression analysis further corroborates the well-supported concept that a lack of cortical continuity (<25%) is a strong predictor of nonunion and reoperation [6, 11, 25, 29, 30]. Although the presence of an open fracture was not a significant covariate in the multiple logistic regression model, open fractures are known prognostic factors for complications following fracture fixation [31‐35] and this characteristic demonstrated a significant association to the incidence of both nonunion and reoperation in the univariate models, suggesting a possible study power issue. Our findings suggest a trend towards increased reoperations with transverse fractures compared to oblique or spiral fractures and is consistent with previous reports of their characteristic association to high-impact injuries and the need for reoperation to achieve bony union in such cases [11, 25].

Using our cohort of patients, we were able to independently evaluate the validity of a previous prognostic index. This index used three variables (open fracture, presence of a fracture gap post-fixation, and transverse fracture type) and was developed using a cohort of 200 patients [25]. Using our study patients as a validation cohort, we are able to see the increase in reoperation rates in patients who satisfied the prognostic variable criterion, thereby confirming the predictive value of the risk index.

Logistic regression analysis using different aggregates of two of the three previously-developed prognostic risk variables indicated that a grouping of both a fracture gap and open fracture together was the only significant predictor for the incidence of nonunion. Using this observation to build upon the predictive value of the previously-identified risk index, such patients can be considered as having fractures of high risk for the development of nonunion and can therefore be more accurately identified to facilitate treatment options more appropriate for achieving bony healing.

It is of interest to note that although only a combination of a fracture gap and an open fracture was found to be significantly attributed to the incidence of nonunion alone, all combinations of the three previously-identified prognostic variables demonstrate predictive value for reoperation. The results indicate that although all included patients who experienced a reoperation had a tibial fracture with less than 25% cortical continuity, not all patients with a diagnosed nonunion, as determined by their attending physician, satisfied this same criterion. This observation leads us to believe that all reported reoperations in this study were performed in an effort to achieve bony union in a fracture showing little to no progression towards healing on its own [36]. There is potential evidence here that surgeons may be performing reoperations on fracture patients with the expectation of developing nonunion rather than the incidence of it. There is still a general consensus across the orthopaedic community that prognostic variables predicting fracture complications such as nonunion or delayed union remain poorly defined [4] and if nonunions are indeed being treated prematurely, further studies on the relationship between reoperations and the proceeding development of fracture union could be warranted based on these findings to truly determine the therapeutic effect of a reoperation and prevent unnecessary and potentially problematic secondary surgery for fracture patients.

In addition to the validation of a previously-developed prognostic risk index for reoperation, we have identified three important findings: (1) tibial fractures with less than 25% cortical continuity have the highest predictive value of nonunion and reoperation than any other fracture characteristic included in this study, (2) the combination of a fracture gap post-fixation and an open fracture yields patients that are at high risk for developing nonunion (3) orthopaedic surgeons may be prematurely treating patients with anticipated nonunion rather than the incidence of it via invasive surgical methods.

Our findings are strengthened by (1) involving a number of experts during development of the case report forms and prognostic factor list; (2) using a comprehensive sampling of patients from three hospital sites within three university-affiliated academic centers; (3) including only patients with sufficient amount of follow-up data to identify healing or the occurrence of a bone healing complication; (4) choosing an outcome measure that is objective and of unequivocal importance to patients; (5) our thorough data collection and analytic methods, which included a compilation of data extracted from patient charts (completed independently by two reviewers) and separately from the radiographs.

The results have limitations, which include (1) our sample may not be completely representative of all tibial fracture patients, but only reflects those treated within three university affiliated academic centers; (2) the identified prognostic factors may only be generalizable to tibial fracture outcomes; (3) as a retrospective review, the data collected from patient charts may not always be reliable or complete; (4) our sample size may have been insufficient to definitively power our regression analyses.