Introduction
Approximately 5–10 % of fractures fail to heal properly, qualifying as delayed union or persistent non-union fractures [
1]. Non-unions are radiologically classified as atrophic, oligotrophic, or hypertrophic, based in part on the particular pathogenesis of the non-union, which thus has an impact on further treatment options [
2]. The occurrence of infection at the site of non-union is a major issue in selecting appropriate treatment; infection is considered the worst prognostic factor in tibia fracture care [
3]. When stable fracture fixation can be confirmed, the presence of a bacterial infection is a possible reason for the persistence of a non-union [
4]. The treatment algorithm in such cases becomes more radical and aggressive debridement, with installation of an external fixator.
The presence of an acute infection may be obvious due to open draining wounds, erythema, or incomplete soft tissue coverage. However, swelling, erythema, and hyperthermia can be suggestive of an infection, while merely indicating instability and local inflammatory reaction. Furthermore, a bacterial infection at the site of non-union can be subacute or chronic, and consequently more difficult to diagnose. Besides the clinical aspects, the diagnosis of infection is supported by laboratory blood tests such as C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), and peripheral leukocyte count. However, these laboratory tests are neither highly sensitive nor specific for infection as a cause of non-union. The gold standard proof of infection is from histological examination and pathogen-positive cell cultures of the infected area, which requires invasive biopsy. Non-invasive diagnostic tests include plain radiography, CT, ultrasonography, infection scintigraphy, and MRI. It is the last approach that may be the first course of imaging work-up of a suspected infection [
5]. MRI can exclude osteomyelitis with high certainty [
6]. However, it cannot distinguish with the same certainty between osteomyelitis and noninfectious causes of abnormal marrow signal intensity, such as neuropathic arthropathy and reactive marrow edema [
6]. Furthermore, MRI can be hampered by the presence of screws, plates, or other metallic devices at the site of bone injury [
7].
In a previously published study, [
18F]FDG PET demonstrated high sensitivity in identifying the presence of osteomyelitis in orthopedic surgery patients with nonspecific clinical symptoms of infection [
8]. In a meta-analysis of studies on suspected osteomyelitis, PET/CT with [
18F]FDG has shown promising sensitivity, specificity, and accuracy in comparison to bone scintigraphy or leukocyte scintigraphy [
9]. Despite these findings, there has hitherto been no study systematically evaluating [
18F]FDG PET for diagnosis of suspected infections at non-union fractures. The aim of our study, therefore, was to investigate whether increased [
18F]FDG accumulation in non-union fracture sites affords a non-invasive imaging biomarker for infections in cases confirmed by intraoperative tissue culture or clinical follow-up.
Methods
Patient population
This study was approved by the institutional ethics committee.
From a large PET database at the University of Munich, we retrospectively identified 35 patients with suspected infection of non-union fractures who had been treated in a level I trauma center. The patients had undergone either [18F]FDG PET/CT (N = 24), [18F]FDG PET (N = 11) with additional CT (N = 8), or conventional X-ray (N = 3) between 2000 and 2015.
Patients presented with persistent pain at the fracture site and may have also reported abnormal movement or clicking at the level of the fracture. The geneses of the non-unions were trauma, wedge osteotomy, osseous segment transport, and arthrodesis or fatigue fractures. Clear clinical markers for acute local infection such as fistulas or pyrophoric wounds, erythema, and/or hyperthermia were absent. Laboratory findings (CRP, leukocytes), risk factors (atherosclerosis, alcohol, adiposity, vascular disease, history of infection, smoking or nicotine use, neurology, gout, and renal insufficiency), a surgically implanted metallic device, gender, age, body mass index (BMI), and location of injury were recorded. At the time of the [18F]FDG PET scan, patients were not being treated with antibiotics.
[18F]FDG PET and PET/CT scan
Partial-body PET scans were acquired in three-dimensional (3D) mode using different PET scanners (N = 5 GE Discovery 690 CT, N = 11 ECAT PET scanner, N = 14 Philips Gemini CT, N = 5 Siemens BioGraph 64). Among the 35 patients, 24 had undergone an [18F]FDG PET/CT, and 11 had undergone only standalone [18F]FDG PET. Based on the National Electrical Manufacturers Association (NEMA) NU2-2001 standard, phantom studies were conducted, and standardized uptake value (SUV) conversion factors were calculated to allow valid pooling of the results. After patients had fasted for at least 6 h and blood glucose levels were measured to exclude cases of blood glucose levels above 150 mg/dL (mean 97 ± 10 ng/mL prior to scan), a diuretic was administered intravenously (furosemid, Furorese 20 mg; Hexal AG, Holzkirchen, Germany), closely followed by bolus administration of [18F]FDG (mean 254 ± 72 MBq) according to body mass. The emission recording sequence was initiated 60 min after intravenous injection of [18F]FDG. Attenuation correction was based on either CT or an external rotating 68Ge source. CT was performed as high-dose CT (automated dose modulation, mean 220 mAs; 120 kV, CT slice 3 mm) with (N = 20) or without contrast agent (N = 4). When contrast agent was administered prior to the CT scan, it consisted of a mean weight-adapted volume of 120 mL iodine-containing contrast agent (iomeprol, Imeron® 350 mg iodine/mL, Bracco Imaging Deutschland GmbH, Konstanz, Germany), administered intravenously at a rate of 2.5 mL/s; the CT scan was initiated 50 s after the beginning of the infusion of contrast agent in order to depict the venous contrast medium phase. A dedicated software package was used for image reading (Hermes Hybrid Viewer, version 2.0; Hermes Medical Solutions, Stockholm, Sweden). In the present retrospective study, the images were interpreted by consensus of two experienced nuclear medicine physicians and a radiologist, who were blinded to bacteriological and surgical data and the patients’ clinical follow-up.
Image interpretation
For image interpretation, we performed quantitative SUV-based analysis, visual analysis, and a combination of both assessments.
1.
SUV-based analysis
Within the SUV based analysis, we performed further receiver operating characteristic (ROC) analysis to obtain a cut-off value. Consequently, all patients with SUVmax greater than the cut-off value were rated positive, and those with SUVmax less than the cut off-value were rated negative for infection.
2.
Visual analysis
We interpreted images positive for infection of the non-union using two different criteria.
3.
Combination of visual diagnosis and SUV-based diagnosis
In addition to these analyses, we performed a combination of the second set of visual analysis and the SUV-based diagnosis. In this approach, [18F]FDG PET was rated as positive only if visual assessment and SUV-based analysis were concordant.
Results validation
PET findings were verified by intraoperative tissue cultures (N = 25) or long term clinical follow-up for more than one year (N = 10). The clinical assessment and other follow-up data were registered during regular visits to the trauma center.
True positive (TP), true negative (TN), false positive (FP), and false negative (FN) were defined as follows:
-
TP:
-
A positive PET result according to the above definitions, and
-
positive intraoperative bone microbiology or, in the case of follow-up, a radiologically and clinically eventful follow-up with the occurrence of clinical symptoms such as pain at the infection site, drainage from the area, fever, tenderness, redness, and warmth or swelling around the affected bone, lost range of motion, or increasing laboratory parameters for infection (leukocytes, CRP level);
-
TN:
-
negative PET result according to our given definitions, and
-
negative intraoperative bone microbiology or, in the case of follow-up, a radiologically and clinically uneventful follow-up without clinical symptoms, and with normal infection parameters with no further therapy for at least one year;
-
FP:
-
positive PET result according to our given definitions, and
-
negative intraoperative bone microbiology or negative follow-up;
-
FN:
-
negative PET result according to our given definitions, and
-
positive intraoperative bone microbiology or negative follow-up.
Statistical analysis
Statistical analyses were performed using the SPSS statistical software package (version 22.0; IBM Corp., Armonk, NY, USA). A p value of less than 0.05 was considered statistically significant. Comparisons of variables between clinically infected and non-infected groups were performed using the t test for parametric data, the chi-square test for ordinal data, and the Mann–Whitney U test for nonparametric data. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated by comparing the PET results to intraoperative findings or clinical follow-up.
Discussion
Infection at the site of a fraction can delay union, or can result in a non-union, which is a potentially disabling persistent failure of fracture healing. By definition, a non-union will not heal unless properly treated in accordance with the cause of the failure. It is of the utmost importance, therefore, to rule out suspected infection in cases of non-union fractures so as to facilitate timely and appropriate surgical fixation treatments. While several imaging modalities have found use in the investigation of potentially infected non-union fractures, [18F]FDG PET has not hitherto been tested. We thus present the first systematic evaluation of [18F]FDG PET exclusively in patients with non-union fractures and suspected infections.
We tested several approaches for evaluating the imaging results. Applying the second set of the visual analysis, [18F]FDG PET showed high diagnostic sensitivity of 85 % and specificity of 86 %, which should afford a secure diagnosis of infected non-union fractures. Furthermore, the high NPV of 90 % should help to avoid inappropriate attempts to treat with antibiotics or surgical debridement when infection is absent. Thus, we demonstrate that [18F]FDG PET reliably differentiates infected and non-infected non-unions, and should then be a useful adjunct for therapeutic management.
Metallic implants are vulnerable sites of postoperative infection; in a previously published [
18F]FDG PET study, a subgroup of six of 22 patients with suspected metallic implant-associated infections had non-union fractures. These six patients were rated as having TN findings [
12], indicating that [
18F]FDG uptake at uninfected non-union sites is significantly lower than that at the sites of actual infections. Another study evaluated [
18F]FDG PET/CT in trauma patients with suspected chronic osteomyelitis [
13]; the authors found that six of 33 patients with suspected osteomyelitis had non-union fractures, of which three were rated true positive, two as true negative, and one as false positive, due to pronounced [
18F]FDG uptake within the area of the fracture, despite absence of infection. Our investigation in a larger series confirms these preliminary published results [
12,
13].
In the absence of infection, [18F]FDG accumulation may be elevated in non-union fractures due to imperfect fixation, with sterile inflammation arising from the resultant friction. However, we expected that the [18F]FDG uptake of the infected non-union should significantly exceed that seen with non-infected non-union. Indeed, clear discernment of the distinction between aseptic inflammation and infection did not pose a problem in the present study.
A standardized approach to the interpretation of [
18F]FDG PET for diagnosing infection in non-union fractures is presently lacking. In our study, no reliable SUV-based threshold for more accurate diagnosis in individual patients emerged. Likewise, it has been reported that there is no simple relation between the probability of infection and the intensity of periprosthetic activity [
14]. Considering merely the presence of activity at the bone–bone interface or the prosthesis–bone interface cannot be unambiguously associated with infection. These results are in line with previously published studies addressing periprosthetic [
18F]FDG uptake [
10,
15,
16]. We find that the particular pattern of pathological uptake must be considered. Applying the second set of visual analysis resulted in achieving high sensitivity of 85 % and specificity of 86 %. [
18F]FDG uptake at the infection sites was focally increased, and typically did not affect the entire surface of the non-union fracture. Locally increased stress and local friction can also lead to focally circumscribed uptake, which is generally less pronounced than that due to infection. In non-infected non-unions, [
18F]FDG uptake was more homogenous along the fracture line. These patterns are likely due to aseptic inflammation provoked by persistent movement and friction along the entire surface of the non-union fraction. Regarding the pathological pattern in this visual analysis, the number of false-positive findings was lower than in the SUV-based quantitative analysis. We thus concur with the findings of Chacko et al. [
16], who noted that the intensity of [
18F]FDG uptake is less informative than is the location of the increased signal. The combination of the visual and quantitative analysis seems to underperform the proposed visual method, because no diagnostically reliable SUV
max cut-off value could be identified.
Hypertrophic, oligotrophic, and atrophic radiographic appearance allows estimation of the degree of fracture stability and also the biologic viability of the fracture fragments. It is likely that different forms of fracture non-unions may be responsible for the differences in [18F]FDG uptake patterns. In particular, we suppose that hypertrophic non-unions may have a higher physiological [18F]FDG uptake than oligotrophic or atrophic non-unions, due to more distinct bone remodeling. In our study, hypertrophic non-union was radiologically diagnosed in six patients, of whom five were rated as true negative and one as false positive. In that individual, hypertrophy might have aggravated the diagnosis. However, the other five patients with hypertrophic non-union were rated correctly, thus indicating that hypertrophy is not a contraindication for the [18F]FDG PET method. Further studies are needed to examine the diagnostic accuracy in the relatively rare patients with hypertrophic non-union and suspected infections.
Contemporary nuclear medicine offers several imaging modalities for use in diagnosing infection in bone. At present, white blood cell (WBC)/bone marrow imaging is considered the best functional diagnostic imaging test for infected prosthetic joints [
14,
17]. The diagnostic value of combined imaging with
111In-labeled leukocytes and
99mTc methylene diphosphonate for the diagnosis of non-union fractures was discussed in an earlier study [
18], with the authors reporting sensitivity, specificity, PPV, NPV, and accuracy of 86 %, 84 %, 69 %, 94 %, and 82 %, respectively. Except for the lower PPV, these results seem comparable to those from our second set of [
18F]FDG PET visual analysis, which may call into question the additional value of single-photon emission computerized tomography (SPECT)/CT in WBC in patients with suspicion of infection at non-union fractures. Thus far, no multimodal imaging comparison has been performed regarding diagnostic accuracy in patients with non-unions, and the best imaging modality for non-union fractures remains to be identified. Images obtained with PET have much higher resolution than scintigraphy images [
19], and [
18F]FDG PET has an inherently high target-to-background ratio. Furthermore, the procedure is widely available. Diagnostic results from [
18F]FDG PET can be secured within 2 h, which is favorable for patient welfare. Our results indicate that [
18F]FDG PET/CT or [
18F]FDG PET with additional morphological imaging represents a plausible alternative for imaging infections of non-unions.