This study, and several preceding ones, have established several common features of deep infections of tumor endoprosthesis. These include the need for a considerably longer period of follow-up and infection monitoring for patients with endoprostheses than is required for those with conventional arthroplasty [2, 13, 14]; the dominant incidence of staphylococci as the infecting organisms [2, 9, 15]; the advantages of two-stage revision for infection control over other limb salvage modalities [2, 9, 13, 14, 16, 17]; the difficulty of conservative therapy for deep infection [2, 9]; and the need for more than 2 or more surgical interventions in order to control an infection [3, 9].
Because the diagnosis of infection was based on clinical findings, we expected some homogeneity of clinical symptoms among the cases included in our study. However, we detected some variations in patient condition that might have influenced their infection condition. Early infections  may be associated with discharge/pus at presentation, whereas late infections  may be associated with prosthetic loosening. In this study, extraarticular resection, resulting in a lack of soft tissue as a protective factor against infection, was associated with pus/discharge at presentation and was predictive of poor infection control period and limb salvage prognoses (Figure 2C and 4C). This suggests that in cases with extraarticular resection, both surgeons and patients should recognize the higher risk for poor treatment outcome. In the present study, infection control rate by soft-tissue flap was 42.9%, suggesting that soft tissue coverage without prosthesis removal is not sufficient for infection control. Combination of soft tissue coverage, especially microvascular free flap coverage, and two-stage revision has been reported to yield favorable results, highlighting the importance of both soft tissue and prosthesis management in endoprosthesis infection . We agree that such methods should be employed, especially in patients at high risk of poor infection control, such as those with extraarticular resection or discharge/pus at infection presentation.
S. aureus infections were associated with a lack of soft tissue (large defect of quadriceps muscle, extraarticular resection). Factors associated with the invasiveness of the initial operation (extraarticular resection and prolonged operation time) were especially associated with MRSA infection. In contrast, patients infected with coagulase-negative staphylococci, displayed considerably milder signs of infection (low body temperature and CRP) than those infected with other organisms. This suggests that infection symptoms might be, at least partly, regulated by the causal organisms. In the present study, however, negative cultures were considerably more frequent than in previous studies [2, 9]. Pathogens were detected in 75.4% of our cases, whereas in previous studies, the positive culture rates were reported to be 89.7% to 93%. In addition, the incidence of polymicrobial infection was lower in the present study than in previous studies (3.5% vs 20.7-26%). We also observed a higher incidence of S. aureus- induced infection in the present study than was reported previously. In previous reports, coagulase-negative Staphylococcus was reported to be dominant over S. aureus[2, 9]. The factors underlying such differences are unclear; however, they may be partially explained by differences in infection location, soft-tissue condition, antibiotic application, and definition of infection.
Several factors that may affect infection control status and numbers of surgical intervention were detected in the present study. Early amputation after the initial surgery may have been recorded as early infection control with fewer surgical interventions. Therefore, shorter control periods or fewer surgical interventions do not necessarily imply acceptable results for each patient. Nevertheless, the risk factors that emerged in the present study for failure of infection control or for more frequent surgery for infection control, such as tumor with the extracompartmental extension, comorbidities that increased the risk of infection, such as diabetes mellitus, and lack of a gastrocnemius flap in the tibia, still reinforce the importance of the immune system and soft-tissue condition in infection control. Unexpectedly, parameters that represent the intensity of inflammation such as body temperature or CRP levels did not predict the treatment outcome of infection. On the other hand, we emphasized the importance of other clinical key findings at presentation; discharge or pus at the infection presentation and the period of infection presentation, that might be useful for the prediction of treatment outcome such as the number of surgical intervention or treatment period for infection control.
The advantage of two-stage revision over other limb salvage modalities has been well established [2, 9, 13, 14, 16, 17]. Although the success rate of one-stage revision is less than 50% [2, 9], this modality can still be used for short-term infections or patients with extreme comorbidities [19, 20]. Debridement without prosthesis removal is not recommended for the control of infection, as it has <5% success rate [2, 9]. Because of the significant difference in the success of infection control between prosthesis removal and prosthesis preservation observed in the present study, we recommend early removal of prostheses for infection control.
If the conservative treatment can be successfully completed, patients can benefit greatly. The difficulty of conservative therapy is well accepted [2, 9], and it was attempted in less than 10% of the cases we examined. Our results suggest that conservative therapy may be attempted in cases without discharge/pus, in limited circumstances.
Although limb salvage or prosthesis survival rates have been extensively reported, few data have been published regarding factors that might influence these rates. The significant of soft tissue has been well established for limb salvage. Poor skin condition, repeated surgery, radiotherapy, and lack of soft tissue are factors that may have affected limb salvage in previous reports [2, 21]. In the present study, we found extraarticular resection and infection with discharge/pus to be novel risks for amputation, indicating the significance of soft tissue and early diagnosis and treatment.
When analyzing the infection of tumor endoprostheses, one should take into account the rarity of the condition, the wide variety of independent variables, and the diversity in the definition of infection. Compared with conventional arthroplasty, tumor endoprosthesis is less frequent, and the accompanying independent variables much more diverse. In addition, the definition of deep infection might be different from that in conventional arthroplasty. The guidelines of the Centers for Disease Control and Prevention  define deep infection or organ/space surgical site infection as infection that occurs within 1 year of surgery if an implant is in place, appears to be related to the operation, and involves deep soft tissues. However, in many previous studies, infections occurring more than 12 months after initial tumor endoprosthesis operations were interpreted as surgical site infections [2, 8, 9]. In order to overcome these bottlenecks, we selected a multicenter approach. In addition, we strictly defined deep infection at the beginning of the study to avoid confusion in data interpretation.
Nevertheless, the study has some limitations. First, it lacked a definite protocol for deep infection control, which resulted in wide variation in treatment modality among the institutions. The effect of each treatment modality might not have been precisely evaluated because its selection depended from the beginning on the infection status. Second, it lacked a definite protocol for administration of perioperative antibiotics. Third, the definition of infection used in the present study might have resulted in the collection of cases with a wide range of infectious conditions (i.e., from short-term infectious local findings with good general condition to fatal septic shocks). In addition, patients who died and thus did not reach the 24 months of follow-up were included; this increased the number of infection cases. However, this might introduce bias because those patients were no longer subject to the potential risk of infection. Finally, treatment decision is based on the intention of patients and surgeons; this might also introduce bias in treatment outcome and prosthesis/limb salvage status. For example, in this study, patients with metastatic bone tumor had a higher risk for amputation. This was, at least partly, due to poorer oncological prognosis, which led patients or surgeons to select less invasive treatment modalities. A future prospective study using a well-established protocol for infection control is necessary.