Introduction

Fracture-related infection (FRI) continues to be challenging for both patients and orthopaedic surgeons, with potentially devastating effects on patient function and long-term outcomes [1,2,3]. A significant proportion of patients are at risk as the incidence of infection following fracture fixation ranges from 1.8% in closed, low-energy fractures to 27% in high energy, open fractures [4]. Infection in the setting of orthopaedic implants can be especially difficult to treat, often requiring extended-course antibiotics, repeat surgical procedures, or even amputation [5,6,7].

Many concepts regarding the treatment of implant-related infection in fracture patients have been adapted from treatment algorithms for periprosthetic joint infection (PJI) in arthroplasty [1, 8] due to the comparatively scarce evidence specific to musculoskeletal trauma patients. There are some similarities in treatment approaches for these two groups of patients, including the potential need for aggressive debridement, antibiotic administration, and the decision between implant retention or explant followed by eventual replacement of the implant [5, 9, 10]. While some crossover exists between PJI treatment approaches and those involving fracture patients, many differences in the treatment of these two patient populations exist. Operative fixation in musculoskeletal trauma involves the added complexity of damaged overlying soft tissue and the potential for concurrent vascular structure compromise—factors that likely contribute to the higher rate of infection in orthopedic trauma surgery [11]. Additionally, microbial infiltration can weaken and disrupt normal bone healing [12]. Ultimately, the most important distinction is the presence of a fracture which compromises biomechanical stability. Failure to provide adequate stability at the zone of injury leads to a cycle of ongoing soft tissue damage, disruption of neovascularization, bone resorption, and persistent infection [8, 13]. Balancing the management of implant-related infection in proximity to actively healing bone and soft tissue escalates the difficulty that surgeons face when treating these patients.

While rates and predictors of treatment failure of PJI have been well described in arthroplasty [3, 14,15,16], evidence regarding the incidence and risk factors for treatment failure of FRI is more limited. For this reason, our goal was to determine the incidence of and identify risk factors for treatment failure in FRI based on the experience at three level 1 trauma centers.

Patients and methods

We conducted retrospective review of patients who underwent surgical intervention for FRI at three adult level 1 trauma centers. Institutional review board approval was obtained at each of the three centers. Subjects were identified from a trauma registry or billing database from January 2011 to December 2015. Patients met inclusion criteria if they were 16 years of age or older and underwent operative irrigation and debridement (I&D) for presumed FRI at any time after having undergone surgical fixation of an open or closed fracture. A minimum of 6 months of follow-up was required for inclusion. Patients were excluded if the presumed infection was associated with hip or knee prothesis or spine instrumentation, and if infection occurred without preceding fracture fixation. Patients were also excluded if an I&D was performed for the treatment of a post-operative hematoma. Only patients who underwent I&D for an infection diagnosed clinically by the following criteria were included.

We defined infection as erythema, purulent drainage, and/or exposed hardware presenting after definitive wound closure that required return to the operating room for I&D at the discretion of the treating surgeon. It should be noted that not all patients may have met confirmatory criteria by the newly defined FRI consensus definition [17]. Cultures were obtained at the time of index I&D in all patients. Antibiotic therapy was initiated in consultation with an infectious disease specialist, with the antibiotic regimen tailored according to culture sensitivity results. If the intraoperative cultures failed to identify a pathogen, then empiric antibiotics were selected per infectious disease recommendations. The specific antibiotic used and the total duration that extended antibiotic therapy was given were both determined in consultation with the infectious disease specialists.

The standard protocol at all three institutions was to repeat operative I&D every 2–3 days as deemed necessary by the surgical team. Based on the standard practice at these institutions, after 7 days had elapsed from the last planned surgical debridement, operative treatment of the infection was considered complete. The primary outcome was treatment failure, defined as the need for secondary surgical debridement or implant revision or removal due to recurrent infection occurring at least 7 days following the presumed final procedure during the initial phase of management.

Patient demographics and injury characteristics were extracted from the medical record. Operative reports were used to determine the number of I&D procedures and how implants were handled—retention or removal following the initial diagnosis of infection was also documented. Outcomes were documented through the last available follow-up visit for each patient.

Statistical analysis was performed using Stata software (College Station, Texas, USA). We summarized baseline characteristics by treatment success using means and standard deviations (SD) for continuous variables and proportions for categorical variables. A Student’s t test or Kruskal–Wallis test were performed as appropriate to compare continuous variables based on conditions of normality. A χ2 test was used to compare categorical variables. In both cases, statistical significance was determined by a p value ≤ 0.05. Multivariable logistic regression was conducted using hypothesis-driven variables with a p value < 0.20 according to our univariate analyses to identify independent associations with treatment failure. Based on our number of treatment failures (n = 51), we were able to include five variables in the regression model—Charlson Comorbidity Index (CCI), whether or not a fracture was open at presentation, how the implant was handled, the number of initial I&D procedures, and the culture result.

Results

We identified 134 patients with FRI that met study inclusion criteria (Table 1). The study population consisted of 61 females and 73 males with an average age of 54.5 years (SD 15.7 years) and an average CCI of 2.6 (SD 2.3). Most sustained fractures of the lower extremity (92%). Thirty of the fractures were classified as open injuries at presentation. Twenty-eight percent of patients were treated at the rural level 1 trauma center while the remainder were treated at two urban level 1 trauma centers.

Table 1 Demographics and patient characteristics
Full size table

A total of 51 patients (38.1%) failed the initial course of treatment (Table 1). Patients in the treatment failure group were approximately 4.1 years older on average, with a mean age of 57.0 (p = 0.14). The mean Charlson Comorbidity Index (CCI) score for patients among those who failed was 2.9 compared to 2.5 in the treatment success group was 2.5 (p = 0.11). Patients who failed were more likely to have had an open injury (31% versus 17%; p = 0.05), to have undergone implant removal as part of their initial treatment (p = 0.03), and to undergo additional index I&D procedures (3.3 versus 1.6; p < 0.001).

The majority of culture results identified a single organism (62%), whereas only 15% were culture-negative (Table 1). Treatment failure was slightly more common in those with cultures that failed to identify an organism (p = 0.08). Staphylococcal species were the most commonly identified microorganisms (Table 2). In cases where an organism was identified, methicillin-resistant Staphylococcus aureus (MRSA) was the most common organism responsible for treatment failure (29%; p = 0.08), while methicillin-sensitive Staphylococcus aureus (MSSA) was more common in the treatment success group (36%; p = 0.04). Additional details on culture results and microorganisms that were identified can be found in Table 2.

Table 2 Specific microorganisms identified
Full size table

Multivariate regression analysis demonstrated a statistically significant association with two or more I&Ds at index diagnosis of infection and treatment failure (odds ratio (OR) 13.22, 95% CI 4.77–36.62, p < 0.001). Culture-negative infection was also significantly associated with treatment failure (OR 4.74, 95% CI 1.26–17.83, p = 0.02). Treatment failure was also associated with elevated CCI, open fracture, and implant removal, though none achieved statistical significance (Table 3).

Table 3 Multivariate regression analysis on risk factors associated with treatment failure in FRI
Full size table

Discussion

FRI remains a challenging problem encountered in the care of the traumatized orthopaedic patient. Our investigation of 134 patients with FRI after surgical intervention demonstrated an unacceptably high rate of treatment failure at nearly 40%. Unfortunately, our rate of treatment failure is comparable to multiple prior studies that demonstrated excessive failure rates, ranging from 26 to 39%. [18,19,20,21,22,23,24] Among these studies, Gitajn et al. identified a 26% treatment failure rate (103/391 patients) while comparing culture-positive and culture-negative infections [21]. Berkes et al. and Rightmire et al. assessed treatment of FRI in the setting of implant retention, citing treatment failure rates of 29% (36/123 patients) and 32% (22/69 patients), respectively. [20, 24] Ovaska et al. assessed outcomes related to FRI after open reduction internal fixation of ankle fractures with 27% (26/97 patients) experiencing treatment failure [23]. Despite substantial heterogeneity amongst these investigations, most clinicians would likely agree that the incidence of treatment failure in this clinical scenario remains too high, and the potential to improve the management of FRI exists.

In our cohort, treatment failure was associated with open fracture, implant removal, culture-negative infection, and having undergone two or more I&Ds at initial diagnosis and treatment. Multivariate analyses reaffirmed the association of culture-negative infection and multiple I&D procedures with treatment failure. Gitajn et al. were among the first to highlight culture-negative infections as important clinical entities encountered in FRI. In contrast to our findings, they did not identify a meaningful difference in the rate of treatment failure among culture-positive and culture-negative infections [21]. The strength of our association may be tempered by the relatively low number of culture-negative infections in our study population, especially as the body of literature in arthroplasty has also demonstrated equivalent or better outcomes in culture-negative infection [25, 26]. When cultures were positive, MSSA was protective, whereas MRSA was the most common organism identified in treatment failure. A high incidence of MRSA infection overall and in treatment failure has been frequently documented in FRI [20, 21, 24, 27]. There are several factors that contribute to MRSA being a particular concern in trauma patients. There is a high prevalence of colonization in this population, and surgeons may have a limited ability to optimize the host prior to fracture surgery [28]. Additionally, perioperative cefazolin fails to provide sufficient coverage for MRSA [27]. Ultimately, our data support continued vigilance when treating MRSA-associated FRI but also suggests aggressive treatment is also important in the management of clinically diagnosed infections even in the absence of positive cultures.

Patients who develop FRI often undergo single or multiple I&Ds depending on a variety of clinical factors and primarily surgeon preference [29]. I&D of devitalized tissue has been described as critical in the eradication of infections associated with implants [30]. Previous studies have suggested that serial debridement is necessary to continually disrupt biofilm formation [31]. However, our finding that multiple I&D procedures was associated with treatment failure is supported by multiple studies [18, 23]. In our study, this may be a marker for more severely infected or more purulent wounds as these are the ones that often undergo additional repeat I&D procedures. It is also likely that the surgical protocol employed in our study period contributed to this finding. Regardless, our data clearly suggest a negative effect of recurrent debridement procedures, and instead, support recent recommendations that focus on a thorough initial debridement and early closure or soft tissue coverage, rather than repeated violation of the surgical wound and the fracture milieu, increasing the risk for contamination and delayed healing [10, 32]. This delicate balance between adequate debridement of infected tissue and avoiding disturbance of fracture healing and the tenuous soft tissue envelope adds complexity to the treatment of FRI.

Our univariate analysis suggested open fracture and implant removal may portend treatment failure; however, statistical significance was not reached in the regression model for these variables. This may be attributable to the limited number of open injuries in our study population, as open fractures are widely understood to be significant risk factors for infection [33], and other investigators have identified an associations between open fracture and treatment failure in FRI [18, 24]. Implant removal, particularly when removed prior to fracture union, is another variable that has been well-documented as detrimental in the management of FRI [18, 20, 23, 24]. The overwhelming majority of patients who did not experience treatment failure had implants maintained or removed only after fracture union in our study. Maintaining fracture stability not only fosters a favorable mechanical environment for osseus union, but it is critical to allow the soft tissues to rest and avoids ongoing damage [13]. Any foreign body effect attributed to the presence of an implant is essentially outweighed by the benefits of stability [8]. Balancing the concern for biofilm formation and the subsequent development of chronic infection or osteomyelitis with fracture stability is yet another challenging component of the management of FRI.

We must acknowledge several limitations of our study. First, this is a retrospective cohort study and is subject to the many weaknesses inherent to this form of study design. Second, a major limitation includes the lack of standardization in the diagnosis of infection, as the clinical judgement of the treating physician was allowed, consistent with multiple prior studies [34]. Recently, consensus recommendations on the diagnosis and management of FRI were published by a panel of experts subsequent to our investigation [8, 35, 36]. Use of this new information would have added strength to our work. Third, our dataset may neglect important patient factors that could increase the risk of treatment failure. As an example, prior authors highlighted the impact of smoking status on the risk of failure [20, 23, 24], while our dataset only included age, sex, CCI, and diabetes status. Fourth, FRI represents a complex clinical scenario with likely substantial heterogeneity among the cases included in our study. In a practical sense, this enabled the inclusion of a larger number of cases, albeit at the expense of standardization of treatment pathways. Although the broad treatment concepts were constant, patients received prompt surgical I&D followed by an extended antibiotic regimen formulated in consultation with infectious disease physicians. Additionally, the surgical protocol for multiple debridements employed during the study period may be considered somewhat outdated but was consistent with standards at the time. The protocol itself is likely a component of the strong association between repeat I&D and treatment failure. Fifth, the inclusion of patients with as few as 6 months of follow-up is not in accordance with recent expert recommendations [8]. Though the majority of the patients included in our study had a year or more of follow-up, our methodology may neglect to capture all significant treatment failure events. Sixth, we failed to include time to onset of symptoms and implant type in our dataset. Both variables may impact the association between implant removal and treatment failure in our study. Prior authors have identified these factors as important determinants of failure with duration of infection and implant surface area implicated in biofilm formation [1, 24, 37, 38]. There are several important strengths to our study. Multiple sites were included, both urban and rural level one trauma centers, providing increased demographic diversity and extending the generalizability and external validity of this study. Further, we were able to identify a large study population, particularly of patients experiencing treatment failure. To our knowledge, this represents one of the largest studies on treatment failure in FRI, and our investigation adds to the limited, but expanding, body of knowledge surrounding this topic.

Conclusion

In conclusion, we witnessed a rate of failure after treatment of FRI of 38.1% which lends further credence to the challenging nature of this clinical problem. Our investigation identified important risk factors for failure including open fracture, implant removal, and multiple debridements, which are in agreement with the work of prior authors. We are the first to highlight culture-negative infection as a potential risk factor for failure, suggesting aggressive treatment of clinically diagnosed cases remains critical even without positive culture data.