This was a retrospective descriptive and case controlled observational study. The study was approved by the Institutional Review Board. Those who had undergone spinal surgeries in the Seoul Sacred Heart General Hospital between January 2005 and September 2012 were reviewed. The patients who received single or multiple PLIFs, or PLIFs for more than half of the fused segments, were included. All cases were consecutive. Repeated operations of the same patients were counted as different cases. Follow-up period was limited to 2 years. Those who had no complaints or clinical signs suggesting an infection during the follow-up period were regarded as non-infected cases. Image test review was not a prerequisite in all cases; however, all infection cases were reviewed. The overall rate and annual rate of SSI were investigated. Classification was performed as per the criteria of Center for Disease Control and Prevention (CDC) of USA: superficial incisional infection (SII) involves only skin or subcutaneous tissue of the incision, deep incisional infection (DII) involves deep soft tissues (e.g., fascial and muscle layers) of the incision, and organ/space infection (O/SI) involves any part of the anatomy, other than the incision, which was opened or manipulated during the operation. Age, sex, amount of blood loss, operation time, number of fused segments, association with diabetes mellitus, whether virgin or revision surgery, local bone irrigation and intradiscal irrigation, were other features analyzed as parameters of risk factors. Other subgroup analyses included difference of elapsed time to a diagnosis (ETD), local and general manifestations, blood test, loosening of implants, and the causative bacteria among the different types of infections. The time of diagnosis was determined as the time when infectious exudates were discovered from the wound in cases of incisional infection, and as the time when antibiotics were started on the evidences of image and laboratory findings in cases of spondylitis. There were two transferred cases that were excluded from the risk factor analysis and included in subgroup analysis of infection cases.

In the statistical analysis of risk factors, independent t-test for ratio variables and Fisher exact test for nominal variables were applied for single variable analysis, and logistic regression analysis was applied for multiple variables analysis. In the analysis of differences among types, chi-squared test, Fisher exact test and analysis of variance were applied. In subgroup analysis, linear by linear association was applied when expected value was less than 5. Significant p-value was defined as 0.1 in single variable risk factor analysis and 0.05 for the others. SPSS ver. 16.0 (SPSS Inc., Chicago, IL, USA) package was used for all analysis.

There is an increased application of PLIFs using pedicle screws and cages. It enables to reconstruct the anterior column through a posterior approach. Thus, it is more feasible to achieve sagittal reconstruction and bone union.12345) However, complications occur more frequently as the surgical procedure is more complex and extended. The incidence of SSI in instrumented PLIF has been reported over a wide range from 3.0% to 7.2%.67910) This was higher than instrumented PLF.678910) Yashiro et al.9) indicated that increased blood loss, longer operation time and tissue damage were attributable. While most types of SSI in PLF were incisional infection, majorly was found as spondylitis, especially in the vicinity of grafted bone and cage in PLIF.8) The current study implied the epidemiology of infection in PLIF would be different from PLF, by presenting that 67% of SSI were spondylitis. The local bone chips prepared for graft were usually contaminated by bacteria of the room air, and intradiscal space was contaminated more than the muscle layer after conventional wound irrigation.11) Furthermore, irrigation of local bone chips decreased the infection rate. Based on the above evidences, the mechanism of development of spondylitis was inferred as follows: the bacteria which contaminated local bone chips and intradiscal space adhere to the remaining avascular disc materials and cages. They are protected from antibiotics by a biofilm and thus develop a latent infection. Intradiscal irrigation with a narrow tip nozzle was expected to be effective to prevent spondylitis. However, this was not proved in multivariate analysis. It was thought that the incidence of SSI was remarkably reduced after the implementation of local bone irrigation and it also had colinearity with local bone irrigation.

Diagnosis and classification followed the CDC criteria. 12) A distinctive feature of the criteria is that it includes O/SI which has no infection at the approach route. In spine surgeries, spondylitis, discitis and meningitis are included in this class.12) Though the O/SI began from the disc space, it seemed appropriate to label them as 'spondylitis' because most of the disc was already removed in PLIF. The criteria extend the diagnostic time up to 12 months in case of implant related DII and O/SI.12) Richards13) diagnosed SSI at an average of 27 months after surgeries for adolescent idiopathic scoliosis. Clark and Shufflebarger14) evaluated at 31 months average. These usually showed no growth or low virulence normal flora of the human skin on microbiological exam.15) Clark and Shufflebarger14) succeeded to identify S. epidermidis by extending the culture period up to 10 days. All of their cases were DII. Anaerobic bacterial contamination of disc after spinal operation has also been reported.16171819) Anaerobic bacteria are usually difficult to identify. Therefore, the no growth cases in our study were presumed as anaerobic bacterial infections. In the current study, while SIIs and DIIs were diagnosed at about 9 days, O/SIs were diagnosed at an average of 165 days, and the most delayed case was at 410 days after the surgery. All latent SSIs were O/SIs, and all the causative bacteria were either having low virulence or were not identified. Thus, latent SSI caused by low virulent bacteria was considered as a particular feature of implant related SSI in PLIF.

Contamination during an operation is unavoidable. Dietz et al.20) reported that 58% of the elective orthopedic surgeries had bacterial contamination. According to an intraoperative experiment, surgical wound, local bone harvested from decompression sites, gloves of surgeons and implants were contaminated by the same bacteria which were cultured from the room air of the operation theater, and the degree of contamination increased proportionally to the exposed time.1121) Latent spondylitis in PLIF has been reported as an unusual case.2223) However, our study showed that it was the major type of SSI in PLIF and strongly suggests that it could be a cardinal reason why PLIF has more SSI than PLF. It made us hypothesize that the chance of contamination by low virulence bacteria is higher and there are more refuges for them. The vectors more liable in contamination are local bone chips and cages, and the refuges are avascular disc material and intradiscal space which are not irrigated effectively. The reasons why diagnosis of O/SI was delayed include the late development of clinical symptoms, and inappropriate workup because of vague clinical manifestation. All O/SIs of the current study had no general fever or local inflammatory symptoms. Local tenderness and pain, which increased with motility, were the only clinical findings. All of them had compatible hematologic findings. Therefore, diagnosis depended on suspicion through clinical findings, and was confirmed by hematological and image tests. However, these kinds of symptoms are not rare in spine fusion patients, and simple radiological findings, such as implant loosening or erosion of end plates, are akin to aseptic nonunion. Many of the current cases were treated mistakenly as usual chronic back pain and received epidural steroid injections, even though manifestations of infection had already begun. Thus, the first step of diagnosis is having a suspicion. If it is supported by hematological examinations and image tests such as magnetic resonance imaging (MRI) and computed tomography (CT) scan, diagnosis can be decisive.

There were several limitations in the current study. All potential variables for risk factors were not considered due to its retrospective design. Interbody graft materials were not unified. Even though auto-iliac bone, demineralized bone matrix, calcium phosphate, and fresh allergenic bone were added to extend the graft material, the major component was auto-local bone in all cases.

The rate of SSI in PLIF was 1.6%. The particular feature was O/SI, i.e., spondylitis, which did not have incisional infection. This kind of infection showed either no bacterial growth or bacteria with low virulence, and was not accompanied by general manifestations of infection. Contamination of local bone chips which were harvested while performing decompression was regarded as an important cause, and irrigation of these had a preventive effect. Longer operation time also increased the risk independently. Latent spondylitis caused by low virulence bacteria developed later. Furthermore, appropriate diagnosis and treatment were delayed due to its vague and confusing clinical and radiological findings. Thus, having a clinical suspicion should be followed by hematological tests and image tests such as MRI and CT to reach a confirmatory diagnosis. Given the causative organisms were normal floras of human skin, thorough control of operation environment is considered to reduce SSI in PLIF.