Spinal infection is one of the most serious complications of spine surgery and the incidence of infection ranges from 0.7% to 12%17). Clinical presentations, laboratory testing, and imaging findings should all be considered for the diagnosis of postoperative spinal infection18). Magnetic resonance imaging (MRI) is the most important imaging modality for evaluating postoperative spinal infection1). To date, several studies have reported the findings of postoperative osteomyelitis, discitis, and epidural abscesses6,7). Boden et al.1) reported differential MRI findings between non-pathologic postoperative changes and discitis during the early postoperative period as well as signal intensity changes in the disc space and adjacent bone marrow on pre- and post-enhancement scans. To our knowledge, information regarding the relationship between the clinical onset of infection and the infection site, as seen on MRI, has rarely been published8). Reviewing the MRI of postoperative spinal infection, we suspected that the postoperative infection begin at the manipulation site and then spread to adjacent local structures. Therefore, we performed a retrospective review of the spine MRI results in patients with proven postoperative spinal infection in order to evaluate the relationship between the infection site seen on MRI and the elapsed time to postoperative spinal infection.

Characteristics of patients’ demographics and surgical procedures are summarized in Tables 1 and 2, respectively. All patients, except for one, underwent surgery using a posterior approach (Table 2). Infection part and MRI findings of each part are summarized in Table 3. 72.1% (31/43) of the patients presented with posterior infection seen on MRI (Fig. 1). In the posterior part, abscess or phlegmon in the back muscle (93.0%) is the most common finding seen on MRI. Ten patients were proven to have infection in both parts (Fig. 2), and two patients were proven to have anterior part infection (Fig. 3) on MRI. The relation between the types of surgical procedures and infection parts on MRI are summarized in Table 4. The clinical onset of each infection part and the relation between clinical onset and detailed site on MRI are summarized in Tables 5 and 6, respectively. Patients in the posterior and both parts groups showing clinically acute onset were 90.3%, and 70 %, respectively (Table 5). Abscess or phlegmon in the back muscle and laminectomy site are the most common MRI findings in the patients with acute and late onset (Table 6).

Follow-up MRI was performed in 12 of the posterior part group patients. The follow-up interval ranged from four to 118 days (mean, 26.3 days). In 50% (6/12) of the patients, infection spread into the anterior part and consequently both parts became infected, as seen on follow-up MRI (mean, 38.2 days). In 16.7% (2/12) of the patients, the infection site was spread into the more anterior portion of the posterior part but still presented with posterior part infection (mean, 8 days). 33.3% (4/12) of the patients showed no change in the infection site (mean, 19.5 days).

The clinical onset ranged from one to 730 days with a mean value of 33.4 days following surgery. The mean durations from surgery to the clinical onset, from the clinical onset to MRI, and from surgery to MRI of each infection part are summarized in Table 7. Fever and a wound were the most common symptoms (n=19, 19, respectively). Back pain (n=6), radiating pain (n=2), and myalgia (n=1) were also observed. ESR (mean, 72.45; range, 23–130) and CRP (mean, 9.80; range, 1.2–30.1) were high in all patients (100%). WBC counts were elevated in 22 patients (51%). The number of patients confirmed by positive cultures on aspirated pus, tissue biopsies, blood or cerebrospinal fluid was 26, nine, six, and two, respectively. The causative micro-organisms are summarized in Table 8. Thirty-six patients underwent surgical treatment with either removal or retaining of instruments, and seven patients underwent conservative management.

Pull ter Gunne and Cohen17) reported the mean time and the median time to the diagnosis of deep and superficial spinal infection. In their study, deep and superficial infection were defined by whether or not the infection extended beyond the fascia, whereas in our study, the location of the infection was determined based on MRI. This classification by anatomic location reflected surgical approach and procedure. Traditionally, postoperative spinal infections were classified into superficial or deep infection depending on the involvement of fascia. In the lumbar spine, deep infection occurred below the lumbodorsal fascia in posterior wounds and in the anterior abdominal fascia in anterior wounds. In the cervical spine, deep infection occurred below the ligamentum nuchae and the fascia layer in posterior wounds and in the platysma layer in anterior wounds3). From the anatomic point of view, both the anterior and posterior parts were included in the deep infection, as seen on the traditional clinical view. In our study, only one patient presented with posterior infection limited to skin and the subcutaneous layer, and which is relevant to the superficial infection. In the surgical debridement of deep, postoperative spinal infection, meticulous eradication of the infection is important, whether the instrumentation is loose or in a well-fixed state. Therefore, it is essential for successful patient management to decide whether the infection extends beyond the fascial plane as well as to identify any infectious foci before surgical treatment. In this respect, MRI has an important role regarding the evaluation of postoperative infection as it provides more accurate and detailed information regarding the infection site.

In general, infection is caused by inoculation at the time of surgery. The manipulation site is suitable for growth of bacteria owing to the injury of the endplate and small vascular structures and resultant hematoma and necrotic tissue formation during surgery11). Moreover, back muscles and the subcutaneous fat layer, both of which were retracted during surgery, might be at risk for tissue ischemia, necrosis, and desiccation. These could increase the risk of wound contamination16). According to the study by Pull ter Gunne and Cohen16), the risk of surgical site infection showed a direct relationship to the surgical approach taken for spine surgery. In their study, the surgical site infection rate was significantly lower using the isolated anterior surgical approach than any in surgery that included a posterior surgical approach (1.7% vs. 4.4%). In our study, abscess or phlegmon in the back muscle (93.0%), laminectomy site (67.4%) and subcutaneous fat layer (62.8%) are the most common MRI findings and this result might be explained by most operations were underwent using a posterior approach. A patient who underwent surgery using both anterior and posterior approaches, showed only posterior part (back muscle) infection on MRI. These findings are consistent with the results of the authors we mentioned. In our study, disc space manipulation such as cage insertion or discectomy were performed in 23 patients, which involving the both parts designated as anterior and posterior part. All patients showed posterior part infection and there were no anterior part infection without accompanied posterior part infection. All of 11 patients who underwent surgery involving only posterior part such as posterior decompression with or without interspinous process decompression device also showed same results. Moreover, 66.7% (8/12) of patients with posterior part infection showed anterior spreading pattern on the follow up MRI. According to these results, postoperative spinal infection undoubtedly occurred as a result of direct inoculation of the posterior part during surgery, and then undoubtedly spread into the anterior portion.

Pull ter Gunne et al. reported surgical site infection after spinal surgery occurred on average during 28.7 postoperative days (range, 5–730)17). These findings are similar to our results. In our study, 86% (37/43) of the patients showed clinical symptoms and signs in less than four weeks following surgery, and 72.1% (31/43) of the patients were proven to have an abnormality seen on MRI within four weeks following surgery. In the study of Hamdan11), the symptom of postoperative infection appeared between four days and three weeks following discectomy. In the study of Fang et al.9), 83.3% (40/48) of the postoperative deep wound spinal infections occurred within 90 days of surgery (range, 5–840 days). Therefore, careful follow-up is essential during the early postoperative period, especially within one month following spinal surgery, according to our study.

In our study, only six patients presented with late clinical onset, and in general, late onset infection was caused by low-virulence pathogens5). These low-virulence organisms usually do not result in clinically significant sepsis, unless it is an immunocompromised patient4). Infection caused by a low-virulence pathogen was difficult to culture5) and might have been excluded in our study, therefore resulting in a lower incidence. According to the study of Bose2), there were three possible mechanisms causing late infections: seeding of the surgical field during surgery; a sterile inflammatory reaction or stimulation of low-virulent organisms to fester caused by metal fretting; and hematogenous seeding. They reported four cases of delayed infection after spine surgery with instrumentation, two patients showed epidural or paraspinal abscess, and other patients showed a soft-tissue mass in the subcutaneous layer and fluid collection in the laminectomy site on computed tomography or MRI. They believed that late infection in their patients was caused by hematogenous seeding, due to the fact that all patients had a distant focus of infection before spinal infection. In our study, two anterior part cases presented with late onset infection. As these patients underwent posterior spine fusion using pedicle screw instrumentation, seeding of the surgical field during surgery could be excluded as the mechanism of infection. As the causative microorganisms in these patients were methicillin-resistant Staphylococcus aureus (MRSA), metal fretting could be excluded as the mechanism of infection. We could consider that these anterior infections might have been caused by hematogenous seeding, although none of the patients had a distant focus of infection.

It is well-known that Staphylococcus aureus is the most common causative organism and Staphylococcus epidermidis and β-hemolytic Streptococci are also commonly encountered pathogens14,15). In our study, the most commonly detected microorganisms were Staphylococcus epidermidis (n=19) and MRSA (n=11). The contamination rate of Staphylococcus epidermidis in blood culture varied widely from as little as 0.6% to over 6%10). Infection of Staphylococcus epidermidis occurred most commonly on intravenous catheters and on medical prostheses12). In our study, among the 19 patients who became infected with Staphylococcus epidermidis, 15 (78.9%) had instruments used in their surgery and all of these patients improved after using antibiotics. Therefore, we considered that isolated Staphylococcus epidermidis in these cases was caused by instrumentation and was the true causative pathogen rather than contamination.

Our study had some limitations. First, as it was a retrospective study, all of the clinical information was obtained through review of electronic medical records. Second, as our medical institution is a tertiary referral center, the possibility that the culture results had been affected by previous antibiotic therapy before a patient’s referral to our hospital, could not be excluded. Moreover, as we excluded culture-negative cases, there might have been a selection bias. Third, most operations were underwent using a posterior approach and these might affect infection sites. But as we have mentioned earlier, isolated anterior surgical approach is the factor decreasing infection risk, so there was much to be said for including a small number of anterior approach. Fourth, as the MRI findings of postoperative infection were investigated by radiologists and no controlled study was designed, some findings might have overlapped the normal postoperative findings. The last limitation is that a metal-induced susceptibility artifact might prohibit evaluating tissue surrounding a metallic implant on MRI. To reduce the metallic artifact, we obtained post-enhancement images without spectral fat saturation.

Postoperative spinal infection mainly occurred within four weeks after surgery in the posterior part and over time the infection was considered to spread into the anterior part, as seen on MRI. Abscess or phlegmon in the surgical area was the most common MR finding. Therefore, for the evaluation of postoperative spinal infection, on MRI, the posterior surgical field is more important than the anterior part such as vertebral body or the disc space.